Executive Political Leanings and Corporate Communications: Evidence from the COVID-19 Pandemic

Abstract

In the context of the COVID-19 pandemic, we investigate whether the U.S. partisan gap over the crisis affected how firms communicated its impact to investors. Specifically, we analyze the effect of executives’ political leanings on their firms’ qualitative disclosures regarding the pandemic. We find that firms led by Republican-leaning executives provided fewer COVID-related disclosures in quarterly financial reports and used language with a more positive tone compared to firms led by Democratic-leaning executives. Furthermore, the partisan effect on both the amount and tone of COVID disclosures was more pronounced when the Republican party held the U.S. presidency, consistent with the prediction of partisanship alignment theory. We also show that the market reacted more strongly to the tone of COVID disclosures from Republican-leaning executives. Our research provides valuable insights for investors and policymakers into the substantial influence of executives’ political ideology on corporate communications, particularly in the context of today’s increasingly polarized political environment and growing global political tensions.

Keywords: Political leanings; COVID-19; Corporate disclosures; Linguistic tone; 10-Qs
JEL: G41, M41, P16

Full text

Executive Political Leanings and Corporate Communications:
Evidence from the COVID-19 Pandemic
Yixing (Ivee) Che
Institute for Financial and Accounting Studies, Xiamen University
iveeche@xmu.edu.cn
Changling Chen*
School of Accounting and Finance, University of Waterloo
clchen@uwaterloo.ca
Victor Xiaoqi Wang
College of Business, California State University, Long Beach
victor.wang@csulb.edu
ABSTRACT:
In the context of the COVID-19 pandemic, we investigate whether the U.S. partisan gap over the
crisis affected how firms communicated its impact to investors. Specifically, we analyze the effect
of executives’ political leanings on their firms’ qualitative disclosures regarding the pandemic. We
find that firms led by Republican-leaning executives provided fewer COVID-related disclosures
in quarterly financial reports and used language with a more positive tone compared to firms led
by Democratic-leaning executives. Furthermore, the partisan effect on both the amount and tone
of COVID disclosures was more pronounced when the Republican party held the U.S. presidency,
consistent with the prediction of partisanship alignment theory. We also show that the market
reacted more strongly to the tone of COVID disclosures from Republican-leaning executives. Our
research provides valuable insights for investors and policymakers into the substantial influence
of executives’ political ideology on corporate communications, particularly in the context of
today’s increasingly polarized political environment and growing global political tensions.
Keywords: Political leanings; COVID-19; Corporate disclosures; Linguistic tone; 10-Qs
JEL: G41, M41, P16
We acknowledge financial support from the Social Sciences and Humanities Research Council of
Canada (SSHRC). We thank the insightful comments from Hans Christensen, Julia Jing Wang,
Kean Wu, Aner Zhou, and discussants (Ilona Bastiaansen and Ray Zhang) and participants at the
2022 Canadian Academic Accounting Association Annual Conference and the 2022 Haskayne and
Fox Accounting Conference.
*Address for correspondence: University of Waterloo, The School of Accounting and Finance,
200 University Avenue West, Waterloo, ON. N2T 3G1, Canada

1
Executive Political Leanings and Corporate Communications:
Evidence from the COVID-19 Pandemic
1. Introduction
The COVID-19 pandemic caused unprecedented damage to the world economy and generated
enormous uncertainties for business operations. In such an environment of extreme uncertainty, it
is crucial for firms to communicate to investors and about the impact of the pandemic on their
business outlooks (Chen et al., 2022). The upper echelon theory suggests that managerial
characteristics can affect their communication styles and have implications for their disclosure
behaviors (Li, 2011; Hanlon et al., 2021). Political ideology,
1
as a highly stable personal
characteristic (Alford et al., 2005; Hatemi et al., 2009), affects managerial decisions on firms’
policies and operations (Chin et al., 2013; Hutton et al., 2014; Elnahas & Kim, 2017), financial
reporting (Notbohm et al., 2019; Zhang, 2015), and voluntary earnings guidance (Elnahas et al.,
2024). In the polarized political environment of the U.S., the contentious COVID pandemic
provides a unique setting to investigate how executives’ political ideology affected their handling
of the pandemic at the corporate level. Specifically, in this paper, we examine the effect of
executives’ political leanings on COVID disclosures in firms’ corporate financial reports.
Republicans generally appear to have a lower assessment of the pandemic risk compared to
Democrats,
2
which leads to differential social responses to the pandemic (Allcott et al., 2020;
Barrios & Hochberg, 2020; Painter & Qiu, 2021). Our research extends this stream of literature by
1
Tedin (1987) defines political ideology as “an interrelated set of attitudes and values about the proper goals of society
and how they should be achieved” (p. 65, as cited in Jost, 2006, p. 653).
2
Numerous polls suggest a large partisan gap in perceived risks posed by COVID-19. For example, according to an
NBC News/Wall Street Journal poll conducted between March 11th and 13th of 2020, 68% of Democrats were
concerned that their family members might catch the novel Coronavirus, whereas only 40% of Republicans shared the
same fear (Hart Research Associates/Public Opinion Strategies, 2020). A June 2020 survey by the Pew Research
Center (a nonpartisan American think tank) shows that 61% of Republicans and Republican-leaning independents
believed that the worst stage of the pandemic was over, in contrast to just 23% of Democrats and pro-Democratic
respondents who shared this belief (Pew Research Center (PRC), 2020).

2
investigating how executives’ political leanings affect their disclosure decisions about the
pandemic in the corporate world. We conjecture that Republican-leaning executives are likely
more optimistic about the pandemic than their Democratic-leaning counterparts and would thus
downplay the COVID risk in their disclosures.
3
To test our conjecture, we collect a sample of 4,791 firm-quarter observations from 908 unique
S&P 1500 firms, which filed quarterly financial reports (10-Qs) over the period from the second
quarter of 2020 (when firms started to file their first 10-Qs after the outbreak of the pandemic) to
the first quarter of 2022. We analyze qualitative COVID disclosures in 10-Q reports, which are
mandated by the U.S. Securities and Exchange Commission (SEC) and provide comprehensive
and relatively timely reporting compared to annual 10-K reports.
4
We quantify COVID disclosures
by assessing the word count of pandemic-related keywords and measure the tone of these
disclosures by evaluating the net positivism of sentences containing COVID-related contents.
Following prior search (e.g., Christensen et al., 2015), we measure the political leaning of a firm
based on its top five executives’ political donations to the Republican versus the Democratic
party.
5
,
6
We first examine how executives’ political leanings affect the amount and tone of COVID
3
On the other hand, according to the value theory, political identities reflect individuals’ core values (Feather, 1979;
Goren et al., 2009) and Republican-leaning managers are usually more conservative and risk-averse than their
Democratic-leaning counterparts (Hutton et al., 2014; Zhang, 2015). Our research focuses on how partisanship
influences executives’ risk attitudes, which may make more risk-averse Republican executives perceive lower COVID
risk relative to their Democratic counterparts. An unconditional argument of Republican-leaning executives being
more risk-averse does not fit into the COVID setting. We thank the anonymous reviewer for this insight.
4
Bursztyn et al. (2020) show that other information outlets such as media news may provide misleading information
and alter people’s beliefs and behaviors in dealing with COVID-19 transmission.
5
A quarterly report is a long and comprehensive document, and usually many managers are involved in preparing this
type of document (Amel-Zadeh et al., 2019). It is more appropriate to use the political ideologies of the top five
executives, rather than only the CFO or CEO, for our research question.
6
We measure political ideology based on executives’ individual political contributions rather than a firm’s donations
to corporate Political Action Committees (PAC) as in Benton et al. (2021), and our textual measures capture multiple
aspects of COVID-19 disclosures, including their tones. Prior studies find that a firm’s donations to PACs usually go
to both parties and do not necessarily reflect executives’ own political ideology (e.g., Hutton et al., 2014; Elnahas &
Kim, 2017).

3
disclosures. We find that firms run by Republican-leaning executives provide a smaller amount of
COVID disclosures relative to firms run by Democratic-leaning executives, and they use language
of a more positive tone than Democratic-leaning firms. We next examine the effect of partisanship
alignment between executives and the U.S. president by utilizing the change of presidential party
affiliation at the 2020 election. We find that the effect of political leanings on both the amount and
the tone of COVID disclosures is more pronounced in the period when Republican President
Trump was in office than in the later period after Democratic President Biden took over the
presidency. This suggests that partisanship alignment with the president intensifies the lower
COVID risk perception of Republican-leaning executives.
In our consequence tests, we find that the tone of disclosures from Republican-leaning
executives elicits stronger stock market reactions, suggesting that the market cares more about the
sentiment of Republican-leaning executives regarding the pandemic. Moreover, we find that the
market reacts stronger, albeit more negatively, to the tone of COVID disclosures when executives’
political leanings align with that of the U.S. president. Our market reaction results hold after
controlling for other partisan-related biases such as managerial positive forecast errors (Stuart et
al., 2024). This suggests that the market understands that executives would be more optimistic in
their disclosures when they share the same political ideology with the U.S. president. Our results
hold for alternative measures of political leanings, and after controlling for other managerial
characteristics and a battery of other factors that may affect the amount and tone of COVID
disclosures. Our results are also robust to entropy balancing, which adjusts for firm characteristics
that are possibly correlated with executive political leanings.
Our focus on COVID disclosures in mandatory 10-Q filings distinguishes our paper from
Benton et al. (2021), which examines the partisan effect on COVID disclosures in voluntary

4
earnings conference calls. This key distinction in disclosure channels yields several important
insights. First, Benton et al. (2021) show that the partisan effect on earnings calls appears to be a
temporary phenomenon, which largely existed in the first quarter of 2020 and started to diminish
in late March 2020. In contrast, we show that the partisan effect on 10-Qs persists over a much
longer period until at least the end of 2020, which marks the transition of the U.S. presidency from
the Republican to the Democratic party. This temporal difference in partisan effects suggests that
mandatory and voluntary disclosure channels may be subject to different institutional and
regulatory pressures. The contrasting persistence of partisan effects between these disclosure
channels is particularly noteworthy given the SEC’s guidance issued at the end of March 2020.
Second, Benton et al. hypothesize that the diminishing partisan effect in conference calls might be
attributable to a crowding-out effect, where mandatory disclosures potentially substitute for
voluntary disclosures following the SEC guidance. Our analysis of 10-Q filings directly addresses
this open question. We do not find evidence supporting such a crowding-out effect; instead, we
document that disclosures from the two outlets are complementary. This complementarity suggests
that managers use these distinct disclosure channels to convey different types of information, with
voluntary conference calls potentially serving as a platform for more immediate, but perhaps less
persistent, partisan expression, while mandatory 10-Q filings reflect more enduring partisan
influences on corporate disclosure policies.
To examine the generalizability of our results, we extend our analysis to non-S&P 1500 firms,
which are excluded from our main sample due to the lack of political leanings data. Employing a
machine learning algorithm, we train a model on firms with political leanings data available, using
COVID disclosure amount and other firm attributes as explanatory variables and political leanings
as the response variable. We then use the model to predict the political leanings for firms lacking

5
political data based on their COVID disclosure amount and other related variables. Using the
predicted political leanings, we find consistent results for non-S&P 1500 firms. This test indicates
that our results based on S&P 1500 firms should be generalizable to non-S&P 1500 firms.
7
Our study contributes to the literature in several ways. First, we contribute to the literature on
the relation between executive characteristics and qualitative corporate disclosures. Prior research
shows that the tone of conference calls has a manager-specific component, which can be attributed
to managers’ observable characteristics, such as age, gender, educational, and professional
background (Davis et al., 2015). Frerich et al. (2018) find narcissism, a relatively less observable
characteristic trait, is linked to an abnormally optimistic tone in financial disclosures. In contrast
to prior research, we show that executives’ political preference (also a less observable personal
trait) is another significant determinant of the attributes of qualitative disclosures.
Second, we extend the literature on the relation between managerial political leanings and
corporate reporting. Prior research examines the impact of political leanings on the quality of
reported earnings (Notbohm et al., 2019; Zhang, 2015) and on managerial forecast behaviors
(Elnahas et al., 2024; Stuart et al., 2024). Our research expands this literature by focusing on the
link between political leanings and qualitative disclosures in financial reports.
Finally, we contribute to the debate on how partisanship may affect corporate reactions to the
pandemic by providing evidence on the impact of political ideology on pandemic-related
disclosures. In contrast to prior research on managerial operational decisions, such as store visits
(Bizjak et al., 2021) during the early pandemic period, we examine managerial communication
decisions in their COVID disclosures. In this regard, our research also differs from Benton et al.
(2021) in that we focus on mandatory disclosures, as mentioned earlier, and that we further identify
7
We thank the anonymous reviewer for this out-of-sample test suggestion.

6
a moderating effect of partisan alignment. We show that partisan alignment (non-alignment) can
amplify (mitigate) the effect of executives’ political leanings on their COVID disclosures. Our
research findings help investors and policy makers understand the considerable influence of
executives’ political ideology on corporate communications in an environment of increasing global
political tensions.
2. Literature Review and Hypotheses Development
2.1 Political Ideology and Managerial Decision Making
Our research lies at the intersection of two streams of literature, one related to the implications
of managerial political ideology for management decisions and the other to the determinants of
qualitative disclosures. Prior studies show that managers’ political ideologies affect many aspects
of their firms’ policies and operations.
8
For example, Chin et al. (2013) find that executives’
political ideologies strongly reflect their level of conservatism and have a significant impact on
management decisions. Firms headed by Republican managers have higher profitability but
maintain a lower level of debt, spend less on research and development, and pursue less risky
investments (Hutton et al., 2014). Republican CEOs are also associated with lower inherent risk
and control risk relative to their Democratic counterparts (Bhandari et al., 2020). Elnahas and Kim
(2017) find that Republican CEOs pursue fewer M&A activities compared to Democratic CEOs.
Closely related to our research, several prior studies find that the political ideologies of CEOs
affect financial reporting and voluntary disclosures of their firms. For example, firms with
Republican CEOs are less likely to restate their financial statements, report more conservatively,
8
More generally, evidence from political psychology shows that corporate executives’ political attitudes and views
are a major part of their identities (Tetlock, 2000). Political ideology has a pervasive effect on people’s working and
private life, from personal health-risk evaluations (Boeuf, 2019) to perceptions of climate change risk (Hu et al.,
2017).

7
and have a lower level of discretionary accruals (Notbohm et al., 2019; Zhang, 2015). Regarding
voluntary disclosures, Elnahas et al. (2024) find that Republican executives provide more timely
and accurate earnings guidance than Democratic executives do.
Overall, the above-mentioned literature suggests that Republican-leaning managers are usually
more risk-averse and conservative than their Democratic-leaning counterparts. However, in the
COVID pandemic setting, Republican-leaning individuals have lower perceptions of COVID risks
relative to Democratic-leaning individuals (Barrios & Hochberg, 2020; Dryhurst et al., 2020; Kerr
et al., 2021). This is seemingly a paradox. Kyung et al. (2021) reconcile this paradox by proposing
a model of identity-based risk perception (IRP), which posits that risk perception also depends on
people’s political identity and on whether the risk relates to their group identity or individual
identity. They suggest that as a group, Republican-leaning individuals focus on national pride and
underestimate COVID risks for the general public.
2.2 Executive Political Leanings and COVID-19 Disclosures (Hypothesis 1)
The upper echelons theory proposes that executives’ characteristics, such as experiences,
values, and personalities, have a great influence on corporate decision making (Hambrick, 2007;
Hambrick & Mason, 1984). Amongst various characteristics, executives' political ideologies play
a significant role in shaping their identities and tend to influence their behavior consistently across
different settings and domains (Epstein, 1980; Funder & Colvin, 1991; Tetlock, 2000; Sherman et
al., 2010). Consistent with the upper echelon theory, extant COVID-19 research suggests that
executives’ political leanings, as an important managerial personal characteristic, likely influence
their perceptions of COVID risks for their firms. A survey conducted by the Pew Research Center
(PRC) in June 2020 showed that 61 percent of Republican-leaning participants thought “the worst
is behind us”, while 76 percent of Democratic-leaning individuals said, “the worst is still to come”

8
(PRC, 2020). Individuals living in Republican-leaning counties perceived a lower risk of COVID-
19 and were less compliant with social distancing protocols and stay-at-home orders (Allcott et al.,
2020; Barrios & Hochberg, 2020; Painter & Qiu, 2021). Moreover, Republican counties are more
likely to underreport COVID-19 cases than Democratic counties (Eutsler et al., 2023).
The above survey and research findings suggest that Republican-leaning executives would
likely downplay the pandemic risk relative to Democrats and tend to demonstrate a more optimistic
assessment of the COVID impact on their businesses at large. If Republican-leaning executives
have a lower assessment of the pandemic risk, we expect that firms managed by Republican-
leaning executives would provide a smaller amount of COVID disclosures and use language of a
more positive tone, compared to firms managed by Democratic-leaning executives. Formally, we
state our first set of hypotheses in alternative form as follows:
H1a: Firms managed by Republican-leaning executives provide a smaller amount of
COVID-19 disclosures compared to firms managed by Democratic-leaning executives.
H1b: Firms managed by Republican-leaning executives provide COVID-19 disclosures
with a more positive tone compared to firms managed by Democratic-leaning executives.
2.3 Management Partisan Alignment and COVID Disclosures (Hypothesis 2)
The literature in political psychology (Bartels, 2002; Jerit & Barabas, 2012) defines partisan
alignment (partisan bias) as individuals’ subconscious interpretation of information in a manner
consistent with their partisanship ideology (i.e., Republican or Democrat)” (Stuart et al., 2024).
Abramowitz and Saunders (2006) find that the influence of party identification on individuals is
stronger than the influence of social identities and the increased alignment between individuals'
political ideology and party identification improves individuals’ party loyalty. The literature in
political science shows that partisans have both confirmation bias (bias for the information that is
consistent with their political leanings) and disconfirmation bias (bias against disconfirming
information) (Taber & Lodge, 2006). Therefore, the biased information processing may cause

9
individuals with different partisan preferences to view the same information differently (Gaines et
al., 2007; Jerit & Barabas, 2012; Shapiro & Bloch-Elkon, 2008). The impact of partisan biases
also extends to financial markets, influencing financial decision-making processes (Kempf et al.,
2021). Thus, understanding the role of political factors, such as partisan alignment, is crucial for
comprehending the dynamics of financial markets and investment behavior.
For the highly politicalized COVID pandemic, Barrios and Hochberg (2020) show that a higher
share of Trump voters in a county is associated with lower perceptions of pandemic risk. Painter
and Qiu (2021) find that residents’ compliance with state-mandated stay-at-home orders is affected
by the extent of their partisanship alignment with the officials issuing the orders. Residents in
counties with aligned political ideologies show quicker responses to state-mandated social
distancing orders relative to those in non-aligned counties. These research findings suggest that
Republican-leaning individuals are more likely to follow the federal government’s policy in
response to the pandemic when Trump from the Republican party served as the U.S. president.
In the corporate world, executives tend to have a more optimistic economic view when their
political ideology aligns with that of the U.S. president (Ramirez & Erickson, 2014). A survey
conducted by Gerber and Huber (2009) showed that individuals whose political ideology is aligned
with that of the president tend to hold higher expectations of current and future economic
performance. Empirically, Stuart et al. (2024) find that firms with CEOs whose political ideology
is aligned with that of the President are more likely to issue overstated management earnings
forecasts, consistent with their more optimistic economic outlook.
The research evidence discussed above implies that Republican-leaning executives likely hold
a more positive view of the economy and their firms’ future performance when the Republican
President Trump is in office because they agree more with Trump’s policies towards the pandemic

10
(Bizjak et al., 2021). After Joe Biden, a Democratic candidate, won the 2020 election and took
office on January 20, 2021 as the new president, Republican executives would likely become less
positive about their firms’ future performance because they agree less with Biden’s policies on
coping with the pandemic.
Therefore, we posit our second set of hypotheses focusing on management partisanship
alignment before and after the change of the U.S. presidency in the 2020 election. We conjecture
that Republican-leaning executives would no longer be more optimistic about the impact of the
pandemic on their firms relative to Democratic-leaning executives in the period when a
Democratic President is in office. Accordingly, they may not continue to provide a smaller amount
of COVID disclosures or use more positive language when discussing the impact of the pandemic
than Democratic-leaning executives do in the post-election period. In other words, we expect that
the partisan effect on executives’ COVID disclosure is more pronounced in the pre-election period
than the post period. Formally, we state our second set of hypotheses in alternative form as follows:
H2a: The negative relation between managerial Republican-leaning and the amount of
COVID-19 disclosures was more pronounced in the period when a Republican president
was in office.
H2b: The positive relation between managerial Republican-leaning and the tone of
COVID-19 disclosures was more pronounced in the period when a Republican president
was in office.
3. Research Design
3.1 Sample Selection
Our sample selection starts with the Compustat Quarterly and Edgar intersection for all 10-Qs
filed over the period from April 2020 to February 2022 for fiscal years 2020 and 2021. We exclude
10-Qs filed in the first quarter of 2020 because the COVID-19 pandemic had not significantly

11
affected firm operations in that initial period.
9
,
10
Firms have 40 days after the end of a fiscal quarter
to file their 10-Q reports for accelerated filers, and 45 days for non-accelerated filers (firms with a
public float of less than $75 million). We match Compustat quarterly data with 10-Q filings based
on CIK and use a cut-off of 60 days to account for the fact that the SEC extended filing deadlines
in 2020 for firms severely affected by the pandemic. Our initial sample consists of 27,296 firm-
quarter observations from 6,218 unique firms. The final sample comes to 4,791 firm-quarter
observations from 908 unique firms, after we exclude observations with missing values in stock
return data (5,004), Execucomp data of executives (13,821),
11
executives’ political contribution
data from the Federal Election Commission (FEC) (3,676),
12
and control variables (4). Panel A of
Table 1 outlines the sample selection process.
Panels B and C of Table 1 provide the sample distribution by fiscal quarter and Fama-French
38-industry classification, respectively. As shown in Panel B, observations for the fiscal year 2020
(2021) account for 50.53% (49.47%) of the final sample. The sample firms are evenly distributed
across the quarters. Panel C shows that sample observations are more concentrated in a few
industries, with the largest industry being Finance, Insurance, and Real Estate (27.61%), followed
by Services (13.80%), and Chemicals and Allied Products (6.64%).
13
[INSERT TABLE 1 HERE]
9
Our empirical results also hold when we include 124 10-Qs filed during the first quarter of 2020 in our analysis.
10
While financial and utility firms are in regulated industries, we keep these firms in our sample because there are
no specific regulatory rules on how these firms should disclose COVID information. We ran robustness tests using a
subsample of financial and utility firms. Our results (untabulated) hold.
11
Execucomp database only covers S&P 1500 firms.
12
We follow prior studies and exclude any CEOs’ contributions through their firms’ PACs because PACs usually
make contributions to multiple parties (Hutton et al., 2014; Elnahas & Kim, 2017).
13
Ramelli and Wagner (2020) identifies a list of industries such as retail, restaurants, and service industries, which
were more severely hit by the pandemic. We test whether the effect differs for firms from such industries. For this test,
we include an indicator variable for such industries and interact it with the political leaning variable. We do not find
a moderating effect (untabulated).

12
3.2 Empirical Models
To test our hypotheses, we estimate the following general model:
COVID_DISC_ATTRIBUTESit = β0 + β1REPUBLICANit + ΣCONTROLSit + ΣINDUSTRY
FE + ΣTIME FE + ΣSTATE FE +εit
(1)
where the dependent variables (COVID_DISC_ATTRIBUTES) are the attributes of COVID
disclosures, including their amount and tone. The test variable, REPUBLICAN, measures the
political leanings of the top-five executives towards the Republican party. As described earlier, we
use the political leanings of the top-five executives rather than that of the CEO or CFO alone
because many managers are involved in the preparation of comprehensive reporting packages such
as quarterly reports (Amel-Zadeh et al., 2019).
Following prior literature (e.g., Christensen et al., 2015), we use two measures, REP_INDEX
and REP_DUM, as the main proxies of executives’ pro-Republican tendency. For each executive,
we calculate the donation ratio as (Donations to Republican – Donations to Democratic) /
(Donations to Republican + Donations to Democratic) for each of the five most recent election
cycles (i.e., 2020, 2018, 2016, 2014, and 2012). We take the average of the ratios over the five
cycles as the Republican-leaning score of an executive.
14
We then aggregate the scores of the top
five executives into a firm-level measure by averaging the scores of these executives. We label
this firm-level measure as REP_INDEX. For our second measure of political leanings, REP_DUM,
we follow the same procedure, except that we assign one to an executive if the executive donated
only to Republican candidates or committees during an election cycle and zero otherwise.
14
This choice is consistent with prior studies showing that an individual’s political ideology remains relatively stable
over the individual’s adult life (Green et al., 2002). Multiple studies make the same assumption in measuring an
individual’s political leaning (e.g., Chin et al., 2013 and Christensen et al., 2015). The political contribution data
supports this assumption. For example, Christensen et al. (2015) find that the political orientation inferred from the
political donation behavior of an executive in different cycles is the same as the person’s lifetime political orientation
92% of the time.

13
Although an individual executive’s political orientation score remains the same during the person’s
tenure, the overall political orientation of the Top-5 executive team of a firm can change from year
to year as executives leave or join the company. Following Chin et al. (2013), we adopt a lagged
design where we match the disclosure measures of a fiscal year with the political leaning measure
of the previous fiscal year. Appendix I provides detailed definitions of all variables.
We include industry fixed effects to control for potentially different impacts of the COVID-19
pandemic on various industries. Since the impact of the pandemic varies widely across industries
(Ramelli & Wagner, 2020), we include Fama-French 38 industry dummies to account for the
nuanced impacts of the pandemic on specific industries. We also include time fixed effects using
a monthly indicator variable in which quarterly reports were filed. The COVID-19 pandemic was
evolving quickly, and the amount and tone of COVID disclosures can vary with the development
of the pandemic, which has undergone many twists and turns. Further, we include state fixed
effects because many COVID policies were implemented at the state level (Adolph et al., 2021).
We construct two measures to capture the amount of disclosure. The first measure,
COVID_DISC1, is the absolute frequency of COVID-related keywords. We read a large random
sample of 10-Qs to develop a list of keywords that include terms such as direct references to the
pandemic and the impact of the pandemic on firms’ operations. Specifically, the keywords include
“COVID 19, pandemic, disrupt, outbreak, closure, disease, coronavirus, travel restriction, public
health, virus, social, distancing, shutdown, quarantine, shelter-in-place, work-from-home, stay-at-
home, non-essential, epidemic, illness, health concern/crisis/threat, infect, remote working,
contagion/contagious, border closure, employee safety, confirmed case, and virtual meeting.”
“Disrupt” and "infect” include suffixes and we allow plural forms of nouns where appropriate. All
keywords are case-insensitive. We include variants for some keywords. For example, “travel

14
restriction” also includes “restrictions on travel”. For “virus”, we exclude “computer virus”, which
is common in 10-Qs when firms discuss cyber-security risk as a risk factor.
We use the following model to test H1a regarding the amount of COVID disclosures:
COVID_DISCit = β0 + β1REPUBLICANit + β2NON_COVID_WCit +
β3OVERCONFIDENCEit + β4SIZEit + β5BTMit + β6LEVit + β7FIRM_AGEit +
β8CASHit + β9EARNit + β10LOSSit +β11∆EARNit + β12STD_EARNit +
β13SALES_GROWTHit + β14BHRETit + β15RET_VOLit + β16BUSSEGSit +
β17GEOSEGSit + β18FORE_RATIOit + β19FOLLOWit + β20AFEit + β21AFit +
β22INST_OWNit + β23NUM_ITEMSit + β24DERIVATIVEit + β25TOTAL_CASESit +
β26NEW_CASESit + β27TELEWORKit + INDUSTRY FE + TIME FE + STATE FE +
εit
(2)
where the dependent variable is the amount of COVID disclosures, namely, COVID_DISC1 or
COVID_DISC2. To mitigate the skewness in the distribution of raw frequency, we calculate
COVID_DISC1 as the natural log of (1+ total frequency of COVID keywords in a 10-Q). The
second measure, COVID_DISC2, is the relative frequency of COVID-related keywords, calculated
as the absolute frequency scaled by the total number of words in a 10-Q. Our H1a predicts a
negative coefficient on β1.
In Equation (2), when the dependent variable is COVID_DISC1, we additionally control for
the length of non-COVID disclosures in 10-Qs by including the variable NON_COVID_WC, which
is the natural log of the word count of non-COVID disclosures. We identify non-COVID
disclosures as sentences not containing any COVID keywords in a 10-Q report. A large value of
NON_COVID_WC likely correlates positively with a firm’s business complexity, and/or the
tendency of the firm’s executives to provide longer disclosures. In either case, the amount of
COVID disclosures (COVID_DISC1) would increase with the amount of non-COVID disclosures
(NON_COVID_WC). We do not include NON_COVID_WC as an additional control when the
dependent variable is COVID_DISC2, since COVID_DISC2 is scaled by the length of the 10-Q.

15
We further include a battery of variables to control for managers’ decisions about COVID
disclosures, including executive characteristics, firm characteristics, firm’s financial performance,
information uncertainty, firm’s operating and supply chain complexities, outside monitoring, other
disclosure-related factors, and pandemic-related factors. Appendix II provides detailed
descriptions of these control variables and the rationales for their inclusion.
We use the following model to test the impact of political ideology on disclosure tone (H1b):
COVID_TONEit = β0 + β1REPUBLICANit + β2NON_COVID_TONEit +
β3OVERCONFIDENCEit + β4SIZEit + β5BTMit + β6LEVit + β7FIRM_AGEit +
β8CASHit + β9EARNit + β10LOSSit +β11∆EARNit + β12STD_EARNit +
β13SALES_GROWTHit + β14BHRETit + β15RET_VOLit + β16BUSSEGSit +
β17GEOSEGSit + β18FORE_RATIOit + β19FOLLOWit + β20AFEit + β21AFit +
β22INST_OWNit + β23TOTAL_CASESit + β24NEW_CASESit + β25TELEWORKit +
INDUSTRY FE + TIME FE + STATE FE + εit
(3)
where the tone of COVID disclosures, COVID_TONE, is measured as the net positivism of the
disclosures related to the pandemic. We identify COVID disclosures as sentences that contain any
of the keywords. Following prior studies, we calculate the variable COVID_TONE as the (number
of positive words – number of negative words) / (number of positive words + number of negative
words), based on the keyword lists of Loughran and McDonald (2011). In the regression, we
control for the tone of non-COVID disclosures, i.e., sentences not containing any COVID
keywords in the 10-Q. Our H1b predicts a positive coefficient on β1.
For this model, we follow Huang et al. (2014) for selecting control variables, except that we
include a few additional variables that are especially relevant to the COVID-19 pandemic and to
our research question. In Equation (3), we control for the tone of non-COVID disclosure contents
in 10-Qs by including NON_COVID_TONE in the model. NON_COVID_TONE is the net
positivism calculated on all the sentences in the 10-Q that do not contain any COVID keyword.
We also control for overconfidence of the top five executives in the tone analysis model as in our

16
disclosure amount models. Other variables that are additional to Huang et al.’s (2014) model
include LEV, CASH, SALES_GROWTH, FORE_RATIO, and INST_OWN. We discuss how these
variables may affect the tone of COVID disclosures in Appendix II.
4. Empirical Results
4.1 Descriptive Statistics
Table 2, Panel A presents descriptive statistics for all the variables. The Republican-leaning
variable, REP_INDEX, has a mean (median) of approximately 0.092 (0.131), indicating that top
executives’ political ideology leans more towards the Republican Party, as they donated, on
average, nine percentage points more to the Republican Party than to the Democratic Party.
15
This
is consistent with the findings of prior studies such as Christensen et al. (2015), suggesting that
top executives of public firms overall hold conservative values.
We transform raw COVID disclosure values to derive our textual variables. As our first
COVID disclosure measure, COVID_DISC1 is defined as the log of (1+ COVID keyword
frequency). The log transformation mitigates the skewness of distribution in the COVID
keywords—COVID_DISC1 has a mean (median) of 4.161 (4.263) after the transformation. Our
second COVID disclosure measure, COVID_DISC2 is the percentage of COVID keyword
frequency, i.e., absolute keyword frequency divided by the total number of words in the 10-Q and
then multiplied by 100. It has a mean (median) of 0.417 (0.374), suggesting that an average
(median) 10-Q contains 0.417% (0.374%) of COVID keywords.
16
15
A value of zero is a neutral reading, indicating that there is no difference in donations to the two parties.
16
In terms of their raw magnitudes (untabulated), an average (median) 10-Q contains 81 (70) COVID-related
keywords. At the sentence level, an average (median) 10-Q has 41 (35) sentences that contain COVID-related
keywords. The average (median) of the total number of words in these sentences is 1,635 (1,401) words. These high
word counts suggest that firms overall provide a significant amount of disclosure about COVID-19 in their 10-Qs,
ranging from 336 words at the 5th percentile to 3,803 at the 95th percentile across firms. Firms’ 10-Qs are usually long
documents with an enormous amount of textual disclosure—an average (median) 10-Q contains 685 (600) sentences

17
The overall tone of COVID-related disclosures (COVID_TONE) is rather negative, with a
mean (median) of –70.655 (–75.680). The reading of –70.655 indicates that, among all tone words
(negative or positive), the percentage of negative words exceeds that of the positive words by
70.665 percentage points. This attests to the severity of the negative impact of the COVID-19
pandemic on most firms. The overall tone of non-COVID disclosure (NON_COVID_TONE) in
10-Q filings is also negative, reflecting the massive market-wide negative news during the
pandemic period. SIZE has a mean of 8.615, because our sample consists of S&P 1500 firms
covered by Execucomp, which are mostly large firms.
[INSERT TABLE 2 HERE]
Table 3 summarizes the Pearson and Spearman correlation matrix of major variables. We
observe a negative correlation between executives’ pro-Republican tendency (REP_INDEX and
REP_DUM) and the amount of COVID-19 disclosures (COVID_DISC1 and COVID_DISC2), with
coefficients ranging from –0.070 to –0.154. This provides preliminary support for H1a, which
predicts that Republican-leaning executives provide a smaller amount of COVID-19 disclosures.
On the other hand, there is a positive association between executives’ pro-Republican tendency
(REP_INDEX and REP_DUM) and the tone of COVID-19 disclosures (COVID_TONE), with
coefficients of 0.023 and 0.025, respectively. Again, this provides preliminary evidence supporting
H1b’s prediction that Republican-leaning executives are more optimistic about the impact of
COVID-19. The amount of non-COVID disclosure (NON_COVID_WC) is negatively correlated
with executives’ Republican-leaning tendency, REP_INDEX and REP_DUM, with coefficients of
–0.110 and –0.122, respectively. This suggests that Republican managers generally provide a
smaller amount of disclosures in 10-Qs, indicating the importance of controlling for the overall
of 20,260 (17,777) words. As mentioned earlier, we focus on 10-Q filings because of their large volume of qualitative
textual disclosures, which should reveal managerial assessment of COVID risks.

18
length of 10-Qs in our later regression analyses.
[INSERT TABLE 3 HERE]
4.2 Regression Results
4.2.1 Executive Political Leaning and the Amount of COVID-19 Disclosures
Table 4 reports the results for the relation between executive political leanings and the amount
of COVID-19 disclosures in 10-Q filings. As predicted in H1a, we expect executives with stronger
Republican ideology to provide a smaller amount of COVID-19 disclosures in 10-Qs due to their
lower assessment of the pandemic risk. To test this hypothesis, we regress the amount of COVID-
19 disclosures (COVID_DISC1 or COVID_DISC2) on Republican-leaning tendency
(REP_INDEX or REP_DUM), controlling for factors that may affect the amount of COVID-19
disclosures and/or political ideology.
Overall, our results suggest that the amount of disclosure is negatively associated with
executives’ pro-Republican tendency, consistent with our H1a prediction that Republican-leaning
managers would provide a smaller amount of disclosure. For Columns (1) and (2), the dependent
variable (COVID_DISC1) is the log-transformed COVID-19 keyword frequency. For Columns (3)
and (4), the dependent variable (COVID_DISC2) is the relative COVID-19 keyword frequency.
Columns (1) and (3) present the regression results where REP_INDEX is the main explanatory
variable, and Columns (2) and (4) those where REP_DUM is the main variable of interest. In both
Columns (1) and (3), REP_INDEX has a negative coefficient, –0.0462 (t-stat = –1.96) and -0.0116
(t-stat = –1.53), respectively. The former is significant at 10% level (two-tailed) and at 5% level
for a one-tailed t-test on our directional prediction of H1a. The latter is significant at 10% for a
one-tailed t-test. The results are stronger in Columns (2) and (4), which are significant at levels of
5% and 10% respectively for two-tailed tests and are both significant at 5% levels for one-tailed

19
tests. For the magnitude of the effect, as shown in Column (4), the relative frequency of COVID
keywords in 10-Qs is approximately 2.65% less for firms run by Republican executives
(REP_DUM = 1) than firms run by Democratic executives ((REP_DUM = 0).
The control variable, NON_COVID_WC, has a positive and significant coefficient in Columns
(1) and (2), suggesting that a firm tends to provide more COVID disclosures if it provides more
disclosures of non-COVID contents. Across the two models in Table 4, we observe a few other
notable relations. Firms with higher BTM (and thus lower growth opportunities) provide a smaller
amount of COVID-19 disclosures, consistent with the notion that such firms are less affected by
the COVID-19 pandemic because they are more established and mature. Firms with greater
earnings volatility (STD_EARN) provide a smaller amount of COVID-19 disclosures probably
because such firms generally face higher risk, and they are likely better prepared for the pandemic.
Those firms with a better stock return (BHRET) over the fiscal quarter provide a smaller amount
of COVID-19 disclosures, consistent with the notion that stock returns reflect the impact of the
pandemic. Firms followed by more analysts (FOLLOW) provide a larger amount of COVID
disclosures, consistent with the notion that analysts demand such information from the company.
Notably, firms from industries with greater teleworkability (TELEWORK) provide a smaller
amount of COVID disclosures. This is consistent with the fact that such companies can more easily
reorganize their workforce to serve their customers in the face of social-distancing measures.
[INSERT TABLE 4 HERE]
4.2.2 Executive Political Leaning and the Tone of COVID-19 Disclosures
Table 5 reports the results for testing H1b, which predicts that Republican-leaning executives
provide COVID-19 disclosures of a more positive tone. The dependent variable is COVID
disclosure tone (COVID_TONE) and the main explanatory variable is the political leaning measure

20
(REP_INDEX or REP_DUM). Columns (1) and (3) present the results using the REP_INDEX
measure. In the first two columns, we do not control for the tone of non-COVID disclosures
(NON_COVID_TONE). We find that the coefficients on REP_INDEX are positive and statistically
significant at levels of 5% and 10% (two-tailed) respectively, and both significant at the 5% level
for one-tailed tests.
In Table 5, Columns (3) and (4), we control for non-COVID disclosures to account for the
possibility that firms may have a general tendency to provide disclosures of more positive or more
negative tones due to firm and/or executive characteristics. For the magnitude of the effect, as
shown in Column (4), COVID disclosures from Republican executives (REP_DUM = 1) have
approximately 2.75% more positive words out of all tone words (i.e., positive and negative words)
relative to Democratic executives (REP_DUM = 0). The coefficient on NON_COVID_TONE is
positive and significant, suggesting that the tone of non-COVID disclosures is positively
associated with the tone of COVID disclosures. After we control for the tone of non-COVID
disclosures, the results become slightly weaker, but are still significant at the 10% level (two-
tailed), or at the 5% level (one-tailed). For brevity, Table 5 and later tables report only the
coefficients on key variables and control variable(s) additional to those reported in Table 4.
Overall, the coefficients on REP_INDEX and REP_DUM are consistent with our directional
prediction in H1b that firms led by Republican-leaning executives provide COVID-19 disclosures
of a more positive tone. For the control variables, the coefficients are largely insignificant in both
columns. This suggests that executives’ political ideology is the predominant factor that
determines the tone of COVID-19 disclosures.
[INSERT TABLE 5 HERE]

21
4.2.3 Executive Political Leaning and the Amount of COVID Disclosures: Partisan Alignment
A presidential election occurred during our sample period, which changed the presidency from
Republican to Democratic. On January 20, 2021, Joe Biden took office as the U.S. president. In
the period before that, the political partisanship of the Pro-Republican executives was aligned with
that of the Republican president. Our second set of hypotheses predicts that Republican-leaning
executives are even more optimistic about the pandemic in the pre-period when their political
ideology is more aligned with that of the U.S. president relative to the post-period when such
alignment stops. To test these hypotheses, we run the same analyses for H1a and H1b, but
separately for periods before and after Joe Biden took office as the new president. We expect that
the relation between the pro-Republican leaning and the amount (tone) of disclosure was stronger
in the pre-period when a Republican served as the U.S. President.
Table 6, Panel A provides the results for testing H2a. The dependent variable is the amount of
COVID-19 disclosures (COVID_DISC1 or COVID_DISC2) and the main explanatory variable is
political leaning (REP_INDEX or REP_DUM). Columns (1) and (3) present the results when the
political leaning variable is REP_INDEX. Columns (2) and (4) provide the results for testing H2a
using REP_DUM as the measure of political leaning. Across all the columns except for Column
(3), the coefficients on the political-leaning variable are significant at the 5% level (two-tailed) in
the pre-period, but not in the post-period. In Column (3), the coefficient is significant at the 10%
level (two-tailed) and 5% (one-tailed) for the pre-period but is insignificant for the post-period.
Overall, the results are consistent with the prediction of H2a. We find that the negative relation
between Republican-leaning tendency and the amount of COVID-19 disclosures holds only in the
period when a Republican president was in office. This relation is fully mitigated after Joe Biden,
the Democratic candidate, took office as the president, due to the non-alignment in political leaning

22
for firms run by pro-republican managers.
[INSERT TABLE 6 HERE]
4.2.4 Executive Political Leaning and the Tone of COVID Disclosures: Partisan Alignment
We further examine the relation between executive political leaning and the tone of COVID-
19 disclosures in pre- and post-periods conditioning on the U.S. 2020 presidential election outcome.
As predicted in H2b, we expect that the positive relation between pro-Republican leaning and the
tone of COVID-19 disclosures would be stronger in the pre-period when the Republican party held
the presidency. Table 6, Panel B presents the results for testing this hypothesis. Columns (1) and
(3) and Columns (2) and (4) report the results, using the political leaning measures REP_INDEX
and REP_DUM, respectively. We report results both with and without controlling for the tone of
non-COVID disclosures. Across all columns except for Column 4, the coefficients on the political
leaning variable (REP_INDEX and REP_DUM) are positive and significant at the 5% level (two-
tailed) for the pre-period but are insignificant for the post-period. In Column (4), the coefficient
on REP_INDEX is significant at the 10% level (two-tailed) for the pre-period but is insignificant
for the post-period. Overall, these results support our prediction in H2b of a stronger relation in
the pre-period relative to the post-period. Combined with the results in Table 6, Panel A, our
findings suggest that the political alignment between executives and the President is a moderating
factor of the partisan effect on COVID disclosures.
4.3 Market Reactions to COVID-19 Disclosures
4.3.1 Models for Market Reaction Test
Partisan biases can have substantial effects on financial markets, which indicates the
importance of political factors for firms’ financial reporting decisions (Chin et al., 2013; Notbohm
et al., 2019; Zhang, 2015). As an additional analysis, we examine whether investors react

23
differently to COVID-19 disclosures depending on the political leaning of the executives who
provide such disclosures. For this analysis, we regress abnormal returns surrounding the filing
dates of 10-Qs on the amount and tone of COVID disclosures, political ideology, the interaction
between the disclosure measures and political ideology, and a group of control variables.
Specifically, we run the following model:
CARit or ABS_CARit = β0 + β1REPUBLICANit + β2COVID_DISCit + β3COVID_TONEit +
β4REPUBLICANit × COVID_DISCit + β5REPUBLICANit × COVID_TONEit +
ΣCONTROLSit + ΣINDUSTRY FE + ΣTIME FE +εit
(4)
where the dependent variable is the three-day cumulative abnormal return centered on the 10-Q
filing date, i.e., CAR [-1, 1], estimated using the Fama-French three-factor model plus momentum.
Because the stock market was rather turbulent during our sample period, we estimate the beta and
other risk factors using the data of 2019, the most recent year in the pre-COVID period. We obtain
stock returns from Compustat daily security file and use the return of the S&P 1500 index as the
market portfolio. Ex-ante, it is not clear whether there is a directional relation between stock return
and the amount or tone of COVID disclosures. As an alternative dependent variable, we use an
unsigned cumulative abnormal return, i.e., the absolute value of CAR (labeled as ABS_CAR),
following prior studies that use unsigned cumulative returns for market reaction tests (e.g., Beaver,
1968; Cready & Mynatt, 1991; Griffin, 2003; Brown & Tucker, 2011; Hope et al., 2016).
The key independent variables include political leaning of executives (REPUBLICAN), the
amount of COVID disclosures (COVID_DISC), the tone of COVID disclosures (COVID_TONE),
and the interactions between political leaning and amount/tone of COVID disclosures. The two
interaction terms, REPUBLICAN * COVID_DISC and REPUBLICAN * COVID_TONE are the
main variables of interest, which capture the extent to which market reactions to the amount or
tone of COVID disclosures vary with the political ideology of a firm’s top executives. As with our

24
previous models, we have two specifications for the REPUBLICAN variable, labeled as
REP_INDEX and REP_DUM, and two COVID_DISC variables, labeled as COVID_DISC1 and
COVID_DISC2. Appendix II provides detailed descriptions of control variables.
Stuart et al. (2024) examine the effect CEOs’ partisan leaning on managerial forecast using a
sample of S&P 500 firms over the 2005-2018 period. They find that firms with CEOs whose
political ideology is aligned with that of the U.S. president are more optimistic about the U.S.
economic outlook and are more likely to issue overstated management earnings forecasts. On the
other hand, Hope et al. (2022) document that management withdrew earnings guidance during the
early stage of the COVID-19 pandemic. Therefore, it is unclear ex ante how political alignment
affects managerial forecasts and therein, to what extent, the market reaction to COVID disclosures
in 10-Qs is affected by potential managerial forecast biases due to partisanship alignment. To
provide insight into this issue, we run an additional test as specified in the following model:
CARit or ABS_CARit = β0 + β1COVID_DISCit + β2COVID_TONEit + β3MF_BIASit +
β4ALIGNit + β5REPUBLICANit × COVID_DISCit + β6 COVID_DISCit × ALIGNit + β7
COVID_TONEit × ALIGNit + β8 MF_BIASit × ALIGNit + ΣCONTROLSit +
ΣINDUSTRY FE + ΣINDUSTRY FE +εit
(5)
For this test, we regress CAR and ABS_CAR on the COVID disclosure amount (COVID_DISC)
and tone (COVID_TONE), management earnings forecast bias (MF_BIAS), and political alignment
(ALIGN), as well as the interaction between political alignment (ALIGN) with these three
disclosure/reporting variables. MF_BIAS is defined as the management’s forecast of annual
earnings less actual earnings, then scaled by the stock price at the beginning of the fiscal year.
ALIGN is an indicator variable equal to one if the political ideology of the firm’s executives is
aligned with that of the U.S. president. All other variables are the same as those in Equation (4).
4.3.2 Empirical Results of Market Reaction Tests
Table 7 presents the results of market reactions to political leaning (REP_INDEX or

25
REP_DUM) and COVID disclosure attributes (COVID_DISC1, COVID_DISC2, and
COVID_TONE). A clear pattern emerges from these results, i.e., investors appear to react more
strongly to the tone of COVID disclosures from Republican managers. Across all the four
specifications, the coefficients on the interaction term, REPUBLICAN * COVID_TONE, are
positive and significant at 5% levels when the dependent variable is the unsigned abnormal return,
ABS_CAR. For examples, REP_INDEX * COVID_TONE has a coefficient of 0.0118 and t-stat of
2.84 in Column (2); REP_DUM * COVID_TONE has a coefficient of 0.0104 and t-stat of 2.64 in
Column (4). These results imply that investors appear to care more about the sentiment of
Republican managers over the pandemic than that of Democratic managers.
[INSERT TABLE 7 HERE]
Table 8 summarizes the regression results of Equation (5), using a merged sample of 1,194
firm-quarter observations between our sample and management earnings forecast data. Stuart et
al. (2024) do not find a significant coefficient on the interaction of forecast bias and political
alignment. Their findings suggest that investors do not recognize the greater forecast bias of firms
with partisan aligned CEOs. For our results shown in Table 8, the coefficients on the interaction
of forecast bias and political alignment (MF_BIAS * ALIGN) are not significant. This is consistent
with the findings of Stuart et al. (2024), suggesting that their findings are generalizable to the
period of the COVID-19 pandemic when the attitudes towards the pandemic were politically
polarized, and when many companies withdrew their earnings guidance (Hope et al. 2022).
17
The coefficients on the interaction between COVID disclosure tone and political assignment
17
Note that Stuart et al. (2024) find a significant and positive coefficient on the main effect of forecast bias, but not
on the interaction of forecast bias and alignment. In contrast, our results show positive, but insignificant coefficients
on the forecast bias variable. This could be due to the weaker power of our tests from a much smaller sample of 1,194
observations, relative to Stuart et al.’s sample of 13,864 observations. Our sample size for this test is much smaller
because our sample period is much shorter, and many firms became less willing to provide earnings forecast during
the early stage of the pandemic (Hope et al., 2022).

26
(COVID_TONE * ALIGN) is significant at the 10% level (two-tailed) and 5% level (one-tailed)
when the dependent variable is the signed return. This suggests that investors recognize that firms
would be more optimistic in their disclosures when the political leaning of their executives are
aligned with the U.S. president. Accordingly, they react negatively to the overly optimistic tone of
such disclosures. Different from Stuart et al. (2024), this finding suggests that investors are better
able to discern the bias in soft information (i.e., linguistic tone) than hard information (i.e., earnings
forecast) when it becomes extremely difficult to forecast earnings due to heightened uncertainty
during the pandemic. When the dependent variable is the unsigned return, the coefficient on
COVID_TONE * ALIGN is also significant at 5% level (two-tailed). This suggests that investors
respond more strongly to the disclosure tone from companies whose executives are more aligned
with the U.S. president in political ideology. The stronger reaction is consistent with the result
from the test using signed return, also suggesting that investors understand that the tone may have
been biased.
[INSERT TABLE 8 HERE]
Overall, our results suggest the market appears to understand the effect of political ideology
on a firm’s disclosure in the COVID setting and reacts to disclosures differently based on
executives’ political leanings. Specifically, as shown in Table 8, for one standard deviation
increase in COVID tone (i.e., 20% more positive words than negative words in firms’ COVID
disclosure in their 10-Qs), the three-day cumulative abnormal returns (CAR) surrounding 10-Q
filing dates are 2.66 basis points less for firms whose executives are aligned with the U.S. president
in political ideology relative to the firms without such alignment. The economic significance of
the differential market reactions to COVID disclosure tone between firms aligned and non-aligned
suggests that investors anticipated the impact of executives’ political alignment on the tone of

27
COVID disclosures and appeared to be less “surprised” when they notice a more positive tone in
COVID disclosures from politically aligned executives.
5. Additional Test
5.1 Mandatory (10-Qs) versus Voluntary (Earnings Conference Calls) Disclosures
Benton et al. (2021) find that firms run by pro-democratic managers provide more disclosures
about the risk of the COVID-19 at earnings conference calls. Moreover, they find that this effect
becomes muted in late March of 2020. This timing coincides with the SEC’s call for more
disclosure of the COVID-19 risk. Theoretical studies on voluntary disclosure literature show that
mandatory disclosure crowds out voluntary disclosure when the manager’s private information
signals firm value (Bagnoli & Watts, 2007). Therefore, it is possible that mandatory disclosure in
10-Qs may crowd out voluntary disclosure in earnings calls, which in turn leads to Benton et al.’s
diminishing partisan effect on earnings calls.
We investigate in two ways whether a crowding-out effect can explain the diminishing relation
documented by Benton et al. (2021). First, we compare the time trends of COVID disclosures in
10-Qs versus earnings calls. As shown in Panel A of Figure 1, the amount of COVID disclosures
in earnings calls increases initially from April 2020 to June 2020, and then decreases over the rest
of the sample period. In Panel B, we observe a similar temporal pattern for COVID disclosures in
10-Q filings, except that the decrease occurs slightly later and at a rate much smaller than that of
COVID disclosures in earnings calls. A “crowding-out” effect would imply that the decrease in
COVID disclosure in earnings calls should coincide with an increase in 10-Qs. However, as shown
in the figures, the two trendlines do not move in opposite directions—they both first increase then
decline over time, albeit at different rates. Therefore, the time trends do not support a crowding-

28
out effect of mandatory COVID disclosure (in 10-Qs) on voluntary disclosure (in earnings calls).
For the second approach, we include the amount of COVID disclosures in earnings calls as an
additional control variable in Equation (2) to examine whether there is a substitutive relation
between disclosures from these two venues. A negative coefficient would indicate a substitutive
relation (implying a crowding-out effect), whereas a positive coefficient would indicate a
complementary relation. As shown in Table 9, the coefficients on COVID_DISC_CALL are
significantly positive across all regressions, suggesting a complementary relation. The regression
results do not support a crowding-out effect. Overall, based on our trend analysis and regression
results, it is not evident that firms reduce voluntary COVID disclosures in earnings conference
calls due to the increase in mandatory COVID disclosures in 10-Qs. In other words, we do not find
evidence that the diminishing partisan effect on COVID disclosures in earnings calls is due to the
crowding-out effect of mandatory 10-Q disclosures.
[INSERT TABLE 9 HERE]
To investigate whether there are also signs of a diminishing partisan effect on COVID
disclosures in 10-Qs, we follow the approach of Benton et al. (2021, page 19) and estimate the
coefficients and standard errors (SEs) of the partisan leaning variable using recursive regressions
on overlapping 90-day sub-samples at 5-day increments from the start of the second quarter of
2020 to the end of 2021. If the partisan effect on 10-Q disclosures diminishes over time, we would
expect to see the SE bars crossing zeros. As shown in the Figure 2, the SE bars (of two standard
deviations in length) are largely below zero throughout the 2020 period and begin to frequently
cross zeros in the 2021 period. This pattern corroborates our finding of a significant partisan effect
on COVID disclosures in the pre-period when Republican-leaning executives’ partisanship aligns
with the Republican President Trump, and an insignificant partisan effect in the post-period when

29
such alignment no longer exists as Biden, a Democrat, took over the U.S. presidency. Combined
with our regression results on political alignment tests, the evidence suggests that the political
alignment between executives and the U.S. president is a moderating factor of the partisan effect
on COVID disclosures.
18
5.2 Entropy Balancing
We conduct entropy balancing analysis to control for potential differences in firm
characteristics that are correlated with executive political leanings. Our political leaning variables
(REP_INDEX and REP_DUM) are the average of the top five executives of the company.
Mathematically, they are continuous variables, which better capture the variation in the political
leaning of a company. As a robustness test, we create a dummy variable, which equals one if on
average the top five executives donated more to Republican candidates than to Democratic
candidates and zero otherwise. We then balance the two groups of firms using entropy balancing
proposed by Hainmueller (2012) and operationalized by Hainmueller and Xu (2013) to ensure that
our results are not due to the observable difference in firm characteristics between the two groups.
Our results hold for the entropy balancing approach. We find (untabulated) that the coefficients on
the political leaning variable are significant for both COVID_DISC1 and COVID_DISC2 at the 5%
level (two-tailed).
5.3 Non-S&P 1500 Firms
In our sample selection, we drop non-S&P 1500 firms because Execucomp does not cover
non-S&P 1500 firms and thus we cannot gather executive political leaning data for them. To
18
We also test whether the political leanings of the first disclosing firms in a quarter can moderate the partisan effect.
We go through firms’ filing dates for each industry-quarter and code a dummy variable (FIRST_DEM) as one if a firm
with Democratic-leaning executives files first, and as zero otherwise. We include the FIRST_DEM and its interaction
term with political leaning variables. We do not find significant effects for FIRST_DEM and its interaction term in
COVID disclosure amount and tone regressions (untabulated).

30
explore whether our results are generalizable to these non-S&P 1500 firms,
19
we use machine
learning to predict the political leaning of a firm based on its COVID disclosures and factors that
may affect their disclosure decisions. To do this, we train a Support Vector Machine (SVM) model
on a random sample of 80% of firms in our sample and then validate the model by predicting the
political learnings for the remaining 20% test sample. The independent variables (i.e., features) for
training the model consist of the amount of COVID disclosures (COVID_DISC1) and all the
control variables in Equation (2), except for the political leaning variable (REPUBLICAN), which
becomes the response variable. We find that the SVM model can predict the political leaning of
the held-out sample at an accuracy of 85.4%.
We then use the model to predict the political learning for these firms excluded from our
sample. Using the predicted political learning data, we run COVID disclosure tone regressions for
these 10,204 firm-quarter observations. As shown in Table 10, we find significantly positive
coefficients on the predicted Republican learning variable (REP_PREDICTED). The results
suggest that our findings based on S&P 1500 firms should be generalizable to non-S&P 1500 firms.
[INSERT TABLE 10 HERE]
6. Conclusion
This study examines the relation between executive political leanings and how corporates
communicate to investors through their COVID disclosures in financial reports. Using 4,791 firm-
quarter observations from 908 unique firms, we find that Republican-leaning managers provide a
smaller amount of COVID disclosures and use languages of a more positive tone. We further
19
We compare the descriptive statistics for our selected sample to the statistics for the firms dropped from the sample
(untabulated), and as expected, these non-S&P 1500 firms are generally smaller. However, they are similar to the
sample firms in terms of the COVID disclosure measures.

31
explore the impact of partisanship alignment on the relation between executive political leanings
and corporate COVID disclosures. Overall, we find a more pronounced effect of Republican-
leaning on both the amount and tone of COVID disclosures during the Republican relative to the
Democratic presidency. Additional tests show that the tone of Republican-leaning executives'
disclosure elicits stronger market reactions, indicating that the market put a greater emphasis on
the sentiment of Republican executives.
In contrast to the prior research showing that Republican-leaning managers tend to be more
conservative in financial reporting (Zhang, 2015) and forward-looking disclosures (Elnahas et al.,
2024), we provide evidence that in the context of the highly politicalized COVID pandemic,
Republican-leaning executives appear to be more optimistic in their COVID disclosures than their
Democratic-leaning peers. Our research suggests that executives’ partisanship-motivated views on
a highly contentious event like the COVID-19 pandemic can override their general attitudes
towards risk, especially when their political leanings are aligned with the ruling party of the
country.
In addition, our focus on mandatory disclosures provides unique insights into how partisan
influences manifest under the constraints of SEC regulations and formal disclosure requirements.
The regulated nature of 10-Q filings creates a distinct environment where partisan effects must
operate within established boundaries of materiality, completeness, and accuracy requirements.
Unlike the relatively unconstrained nature of conference calls studied by Benton et al. (2021), 10-
Q filings must adhere to specific regulatory guidelines while still potentially reflecting partisan
influences. This tension between regulatory compliance and partisan expression manifests in
several ways. First, managers must find ways to incorporate their political beliefs within the formal
structure of risk factor discussions and Management's Discussion and Analysis (MD&A) sections,

32
perhaps leading to more subtle variations in tone, emphasis, and framing rather than overt political
statements. Second, the requirement for factual accuracy and completeness in 10-Q filings means
that partisan influences likely appear through selective emphasis on certain risks or uncertainties,
rather than through omission of material information. Third, the formal certification requirements
under the Sarbanes-Oxley Act create personal liability for officers, potentially leading to more
carefully crafted expressions of partisan views that must balance political beliefs with legal
obligations.
Overall, our research highlights the role of executives’ political ideology in shaping their
disclosure communication under the regulatory context of 10-Q disclosures, which helps
demonstrate how political beliefs influence corporate communications even within the bounds of
formal disclosure requirements. While earnings calls allow managers to express partisan views
more freely through tone, word choice, and topic selection during unscripted Q&A sessions, 10-Q
disclosures require a more nuanced integration of partisan influences within a highly regulated
framework. The persistence of partisan effects in 10-Q filings, despite these regulatory constraints,
suggests that political beliefs are deeply embedded in how managers interpret and communicate
business conditions, risks, and uncertainties.

33
Appendix I: Variable Definitions
Variable
Definition
Test Variables
REP_INDEX
Firm-level Republican index where, for each executive, we calculate the donation
ratio as (Donations to Republican – Donations to Democratic)/(Donations to
Republican + Donations to Democratic) for each of the five most recent election
cycles (i.e., 2020, 2018, 2016, 2014, and 2012). We calculate the Republican-leaning
score for an executive by averaging the donation ratio over the five election cycles.
We aggregate the Republican-leaning scores of the top five executives into a firm-
level measure by taking the average of the scores of these executives.
REP_DUM
Firm-level Republican dummy variable, where we assign one to an executive if the
executive donated only to Republican candidates or committees during an election
cycle and zero otherwise. We calculate the Republican-leaning dummy variable for an
executive by averaging the donation dummy over the five most recent election cycles.
We aggregate the Republican-leaning dummy variables of the top five executives into
a firm-level measure by taking the average of the dummy variables of these
executives.
REPUBLICAN
An indicator variable that equals one if REP_INDEX is greater than zero and zero
otherwise.
Dependent Variables
COVID_DISC1
Amount of COVID disclosures, measured as the natural log of (1+ total frequency of
COVID keywords in a 10-Q). Keywords are provided in Section 3.2.
COVID_DISC2
Amount of COVID disclosures, measured as the absolute frequency of COVID
keywords in a 10-Q, scaled by the total number of words in a 10-Q.
COVID_TONE
Tone of COVID disclosures, measured as the (number of positive words – number of
negative words)/(number of positive words + number of negative words), based on
the keyword lists of Loughran and McDonald (2011). COVID disclosures are
identified as sentences containing the keywords provided in Section 3.2.
Control Variables
NON_COVID_TONE
Net positivism calculated on all the sentences that do not include COVID keywords in
the 10-Q.
NON_COVID_WC
Length (word count) of non-COVID disclosures measured as the natural log of the
word count of non-COVID disclosures.
OVERCONFIDENCE
Overconfidence bias of the top executives of the firm based on the tendency to hold
deep-in-the-money stock options (Campbell et al., 2011; Hirshleifer et al., 2012;
Ahmed & Duellman, 2013).
SIZE
Log of the market value at the end of a fiscal quarter.
BTM
Ratio of book value to market value at the end of a fiscal quarter.
LEV
Sum of short-term and long-term debts, scaled by total assets at the end of a fiscal
quarter.
FIRM_AGE
Log of the number of years that the firm has existed in the Compustat annual file.
CASH
Amount of cash and cash equivalents, scaled by total assets at the end of the fiscal
quarter.
EARN
Income before extraordinary items of a fiscal quarter, scaled by total assets at the
beginning of the fiscal quarter.
LOSS
Dummy variable that equals one if EARN is negative and zero otherwise.
∆EARN
Change in earnings for the quarter, measured as the change in net income before
extraordinary items over the same quarter last fiscal year scaled by the beginning total
assets of the quarter.

34
Appendix I: Variable Definitions (Cont.)
STD_EARN
Standard deviation of earnings over the past five fiscal years, ending 2019 (inclusive),
calculated with at least three years’ non-missing data.
SALES_GROWTH
Percentage change in sales over the same quarter of the last fiscal year.
BHRET
Buy-hold raw return over the quarter.
RET_VOL
Stock return volatility over the quarter.
BUSSEGS
Number of business segments, measured as the log of (1+number of business
segments).
GEOSEGS
Number of geographical segments, measured as the log of (1+ number of geographic
segments).
FORE_RATIO
Percentage of revenue from foreign countries, measured as sales from non-domestic
segments of the firm divided by total sales for a quarter.
FOLLOW
Number of analysts that follow the firm, measured as the log (1+number of estimates
for the current quarter) before the end of the quarter from the I/B/E/S summary file.
AFE
Forecast error for the current quarter, measured as (the actual EPS of the quarter –
most recent median forecast before the end of the quarter), scaled by the stock price at
the quarter-end.
AF
Consensus analyst forecast for the following quarter, measured as the most recent
median analyst forecast for the following quarter, scaled by the stock price at the
quarter-end.
INST_OWN
Percentage of institutional ownership at the calendar quarter closest to the fiscal
quarter end from 13F filings available with WRDS SEC Analytics.
NUM_ITEMS
Number of non-missing line items in the Compustat quarterly file.
DERIVATIVE
Dummy variable that equals one if the firm uses derivatives and zero otherwise.
WORD_COUNT_10Q
Log (1 + word count of a 10-Q).
FOG_10Q
Fog index of a 10-Q.
SUE
Actual EPS minus consensus analyst forecast one day before the earnings
announcement date, scaled by the stock price at quarter end.
TAC
Total accruals calculated from cash flow statements, scaled by total assets at quarter
end.
TELEWORK
Teleworkability of each industry, i.e., the capacity to work from home, obtained from
Dingel and Neiman (2020).
CASES_NEW
Log transformed number of new COVID cases per thousand people over a fiscal
quarter in the county where the company is headquartered.
CASES_DATE
Log transformed number of new COVID cases per thousand people at the end of a
fiscal quarter in the county where the company is headquartered.
CASES_FDATE
Log transformed number of new COVID cases per thousand people on the filing date
of a 10-Q in the county where the company is headquartered.
Other Variables
CAR
Three-day [-1, 1] cumulative abnormal return centered on the 10-Q filing date
estimated using the Fama-French 3-factor model plus momentum, with factors
estimated using stock returns of 2019.
ABS_CAR
Absolute value of CAR.
MF_BIAS
The signed value of annual management earnings forecast errors (forecast – actual)
scaled by the stock price at the beginning of the fiscal year.
ALIGN
An indicator variable that equals one if the executives of a firm are aligned with the
U.S. president in office and zero otherwise.
COVID_DISC_CALL
Percentage of COVID disclosure contents in the presentation session of an earnings
conference call.

35
Appendix II: Descriptions of Control Variables (See Appendix I for variable definitions)
Control variables in Equations (1) – (4) are as follows:
• Executive overconfidence (OVERCONFIDENCE). Overconfident executives may provide fewer COVID
disclosures due to their more optimistic assessment of the pandemic. They could also provide more COVID
disclosures to signal their ability to navigate their organizations through the crisis.
• Firm characteristics such as size (SIZE), book-to-market (BTM), leverage (LEV), and age (FIRM_AGE). We
expect that larger firms provide more COVID disclosure due to their greater business complexity. Firms with
higher age (FIRM_AGE) and greater BTM are usually more established and may be less susceptible to the
pandemic. On the other hand, highly leveraged firms (LEV) can be more negatively affected by the pandemic due
to the higher level of debt in their capital structure. In a crisis such as the COVID-19 pandemic, it is important
for firms to have the liquidity to meet obligations when there is a decrease or even a total loss of revenue over an
extended period. Therefore, we control for firms’ cash holdings (CASH).
• Firms’ financial performance such as earnings (EARN), loss (LOSS), the change in earnings (∆EARN), and the
standard deviation of earnings (STD_EARN). A firm with higher EARN or still profitable (LOSS = 0) may provide
few COVID disclosures because its operations are probably less negatively impacted by the pandemic. On the
other hand, firms with high earnings volatility (STD_EARN) may provide more COVID disclosures because their
operations are riskier and may be more negatively affected by the pandemic. It could also be the case that such
firms provide a smaller amount of disclosure because their managers are more experienced in managing
risks/uncertainties due to the volatile nature of their business. For the same reason as with ∆EARN, we control for
sales growth (SALES_GROWTH).
• Stock returns such as buy-and-hold quarterly raw return (BHRET) and quarterly return volatility (RET_VOL). The
BHRET reflects the impact of the pandemic on the firm, as well as the market expectation about the development
of the pandemic. The RET_VOL reflects information uncertainty, which creates a demand for disclosures.
• Operating and supply chain complexities proxied by the number of business segments (BUSSEGS), number of
geographical segments (GEOSEGS), and percentage of revenue from foreign countries (FORE_RATIO). Firms
with more business/geography segments or more revenues from foreign countries may face a greater challenge to
manage their supply chains when flows of goods and people are restricted due to the pandemic.
• Analyst forecast variables including the number of analysts that follow the firm (FOLLOW), forecast error (AFE),
the most recent median analyst forecast for the following quarter (AF). As information intermediaries, financial
analysts are important users of corporate disclosures, and they demand information from the management. AFE
reflects the extent to which a firm meets or beats analyst forecasts and may affect managers’ disclosure decisions.
Disclosures in 10-Qs contain forward-looking information, and a firm’s future performance captured by AF may
affect the amount of its disclosure. For the same reason as with financial analysts, we include institutional
ownership (INST_OWN) to control for a firm’s information environment.
• Number of non-missing line items (NUM_ITEMS) in the Compustat quarterly file and whether a firm uses
derivatives (DERIVATIVE). Li (2008) finds that the number of line items is a determinant of the length of annual
reports (10-Ks). Cazier and Pfeiffer (2016) find that the use of derivatives increases the length of 10-Ks (which
is in proportion to the length of 10-Qs because quarterly and annual reports share key contents such as financial
statements and footnote disclosures, MD&A, and risk factor disclosures.
• The cumulative number of cases at the end of the quarter (CASE_DATE) as well as the new number of cases
(CASE_NEW). To account for the differential impacts of the pandemic on various industries, we further control
for the teleworkability (TELEWORK) of each industry, i.e., the capacity to work from home. Work from home is
critical to firms during economic watersheds and disruptions and can help mitigate the adverse effect of events
like COVID-19 (Nguyen et al. 2023). We obtain this measure from Dingel and Neiman (2020), who construct
this industry-wise teleworkability measure based on a survey administered by O*NET, an organization sponsored
by the U.S. Department of Labor. This measure captures the percentage of jobs that can be done at home for each
industry.
Controls variables in Equations (5) – (6): We control for the tone of non-COVID disclosures (NON_COVID_TONE)
and OVERCONFIDENT. Following Hope et al. (2016), we also include control variables such as the length and
readability of 10-Q (WORD_COUNT_10Q, FOG_10Q), and earnings news or financial performance (EARN, LOSS,
∆EARN, SALES_GROWTH), and standardized unexpected earnings (SUE). Other controlled firm characteristics
include SIZE, BTM, LEV, FIRM_AGE, BUSSEGS, and GEOSEGS. We also control for accruals (TAC), stock return
volatility (RET_VOL), analyst following (FOLLOW), institutional ownership (INST_OWN), the number of non-
missing line items in the Compustat quarterly file (NUM_ITEMS), and the number of new COVID cases in the county
on the filing date of the 10-Q (CASE_FDATE).

36
References
Abramowitz, A. I., & Saunders, K. L. (2006). Exploring the Bases of Partisanship in the American
Electorate: Social Identity vs. Ideology. Political Research Quarterly, 59(2), 175–187.
https://doi.org/10.1177/106591290605900201
Adolph, C., Amano, K., Bang-Jensen, B., Fullman, N., & Wilkerson, J. (2021). Pandemic Politics:
Timing State-Level Social Distancing Responses to COVID-19. Journal of Health Politics, Policy
and Law, 46(2), 211–233. https://doi.org/10.1215/03616878-8802162
Ahmed, A. S., & Duellman, S. (2011). Evidence on the role of accounting conservatism in monitoring
managers’ investment decisions. Accounting & Finance, 51(3), 609–633.
https://doi.org/10.1111/j.1467-629X.2010.00369.x
Alford, J. R., Funk, C. L., & Hibbing, J. R. (2005). Are Political Orientations Genetically Transmitted?
American Political Science Review, 99(2), 153–167. https://doi.org/10.1017/S0003055405051579
Allcott, H., Boxell, L., Conway, J., Gentzkow, M., Thaler, M., & Yang, D. (2020). Polarization and
public health: Partisan differences in social distancing during the coronavirus pandemic. Journal of
Public Economics, 191, 104254. https://doi.org/10.1016/j.jpubeco.2020.104254
Amel-Zadeh, A., Scherf, A., & Soltes, E. F. (2019). Creating Firm Disclosures. Journal of Financial
Reporting, 4(2), 1–31. https://doi.org/10.2308/jfir-52578
Bagnoli, M., & Watts, S. G. (2007). Financial Reporting and Supplemental Voluntary Disclosures.
Journal of Accounting Research, 45(5), 885–913. https://doi.org/10.1111/j.1475-679X.2007.00258.x
Barrios, J. M., & Hochberg, Y. (2020). Risk Perception Through the Lens of Politics in the Time of the
COVID-19 Pandemic (Working Paper 27008; Working Paper Series). National Bureau of Economic
Research. https://doi.org/10.3386/w27008
Bartels, L. M. (2002). Beyond the Running Tally: Partisan Bias in Political Perceptions. Political
Behavior, 24(2), 117–150. https://doi.org/10.1023/A:1021226224601
Beaver, W. H. (1968). The Information Content of Annual Earnings Announcements. Journal of
Accounting Research, 6, 67–92. https://doi.org/10.2307/2490070
Benton, R. A., Cobb, J. A., & Werner, T. (2021). Firm partisan positioning, polarization, and risk
communication: Examining voluntary disclosures on COVID-19. Strategic Management Journal,
n/a(n/a). https://doi.org/10.1002/smj.3352
Bhandari, A., Golden, J., & Thevenot, M. (2020). CEO Political Ideologies and Auditor-Client
Contracting. Journal of Accounting and Public Policy, 39(5), 1-23.
https://doi.org/10.1016/j.jaccpubpol.2020.106755.
Bizjak, J. M., Kalpathy, S. L., Mihov, V. T., & Ren, J. (2021). CEO Political Leanings and Store-Level
Economic Activity during COVID-19 Crisis: Effects on Shareholder Value and Public Health
(SSRN Scholarly Paper ID 3674512). Social Science Research Network.
https://doi.org/10.2139/ssrn.3674512
Boeuf, B. (2019). Political ideology and health risk perceptions of food. Social Science & Medicine, 236,
112405. https://doi.org/10.1016/j.socscimed.2019.112405
Brown, S. V., & Tucker, J. W. (2011). Large-Sample Evidence on Firms’ Year-over-Year MD&A
Modifications. Journal of Accounting Research, 49(2), 309–346. https://doi.org/10.1111/j.1475-
679X.2010.00396.x
Bursztyn, L., Rao, A., Roth, C. P., & Yanagizawa-Drott, D. H. (2020). Misinformation During a
Pandemic (Working Paper 27417). National Bureau of Economic Research.
https://doi.org/10.3386/w27417
Campbell, T. C., Gallmeyer, M., Johnson, S. A., Rutherford, J., & Stanley, B. W. (2011). CEO optimism
and forced turnover. Journal of Financial Economics, 101(3), 695–712.
https://doi.org/10.1016/j.jfineco.2011.03.004
Chen, C., Stratopoulos, T. C., Wang, V. X., & Xing, B. (Bin). (2022). Do Firms Provide Informative
Disclosures in An Environment of Extreme Uncertainty? Evidence from the COVID-19 Pandemic
(SSRN Scholarly Paper 4004649). https://doi.org/10.2139/ssrn.4004649

37
Chin, M. K., Hambrick, D. C., & Treviño, L. K. (2013). Political Ideologies of CEOs: The Influence of
Executives’ Values on Corporate Social Responsibility. Administrative Science Quarterly, 58(2),
197–232. https://doi.org/10.1177/0001839213486984
Christensen, D. M., Dhaliwal, D. S., Boivie, S., & Graffin, S. D. (2015). Top management conservatism
and corporate risk strategies: Evidence from managers’ personal political orientation and corporate
tax avoidance. Strategic Management Journal, 36(12), 1918–1938. https://doi.org/10.1002/smj.2313
Cready, W. M., & Mynatt, P. G. (1991). The Information Content of Annual Reports: A Price and
Trading Response Analysis. The Accounting Review, 66(2), 291–312.
Davis, A. K., Ge, W., Matsumoto, D., & Zhang, J. L. (2015). The effect of manager-specific optimism on
the tone of earnings conference calls. Review of Accounting Studies, 20(2), 639–673.
https://doi.org/10.1007/s11142-014-9309-4
Dingel, J. I., & Neiman, B. (2020). How many jobs can be done at home? Journal of Public Economics,
189, 104235. https://doi.org/10.1016/j.jpubeco.2020.104235
Dryhurst, S., Schneider, C. R., Kerr, J., Freeman, A. L. J., Recchia, G., van der Bles, A. M., Spiegelhalter,
D., & van der Linden, S. (2020). Risk perceptions of COVID-19 around the world. Journal of Risk
Research, 23(7–8), 994–1006. https://doi.org/10.1080/13669877.2020.1758193
Elnahas, A., Gao, L., Hossain, M. N., & Kim, J.-B. (2024). CEO political ideology and voluntary
forward-looking disclosure (SSRN Scholarly Paper ID 3637608). Forthcoming at Journal of
Financial and Quantitative Analysis, forthcoming.
Elnahas, A. M., & Kim, D. (2017). CEO political ideology and mergers and acquisitions decisions.
Journal of Corporate Finance, 14.
Epstein, S. (1980). The stability of behavior: I. On predicting most of the people much of the time.
Journal of Personality and Social Psychology, 37(7), 1097. https://doi.org/10.1037/0022-
3514.37.7.1097
Eutsler, J., Kathleen Harris, M., Tyler Williams, L., & Cornejo, O. E. (2023). Accounting for partisanship
and politicization: Employing Benford’s Law to examine misreporting of COVID-19 infection cases
and deaths in the United States. Accounting, Organizations and Society, 101455.
https://doi.org/10.1016/j.aos.2023.101455
Feather, N. T. (1981). Value correlates of conservatism. Journal of Personality and Social Psychology,
37(9), 1617. https://doi.org/10.1037/0022-3514.37.9.1617
Frerich, B., Reemda, J., Kerstin, L., & Maas, K. (2018). The use of optimistic tone by narcissistic CEOs.
Accounting, Auditing & Accountability Journal, 31(2), 531–562. https://doi.org/10.1108/AAAJ-11-
2015-2292
Funder, D. C., & Colvin, C. R. (1991). Explorations in behavioral consistency: Properties of persons,
situations, and behaviors. Journal of Personality and Social Psychology, 60(5), 773.
https://doi.org/10.1037/0022-3514.60.5.773
Gaines, B. J., Kuklinski, J. H., Quirk, P. J., Peyton, B., & Verkuilen, J. (2007). Same Facts, Different
Interpretations: Partisan Motivation and Opinion on Iraq. The Journal of Politics, 69(4), 957–974.
https://doi.org/10.1111/j.1468-2508.2007.00601.x
Gerber, A. S., & Huber, G. A. (2009). Partisanship and Economic Behavior: Do Partisan Differences in
Economic Forecasts Predict Real Economic Behavior? American Political Science Review, 103(3),
407–426. https://doi.org/10.1017/S0003055409990098
Goren, P., Federico, C. M., & Kittilson, M. C. (2009). Source Cues, Partisan Identities, and Political
Value Expression. American Journal of Political Science, 53(4), 805–820.
https://doi.org/10.1111/j.1540-5907.2009.00402.x
Green, D., Palmquist, B., & Schickler, E. (2002). Partisan Hearts and Minds: Political Parties and the
Social Identities of Voters. Yale University Press. https://www.jstor.org/stable/j.ctt1npp6m
Griffin, P. A. (2003). Got Information? Investor Response to Form 10-K and Form 10-Q EDGAR Filings.
Review of Accounting Studies, 8(4), 433–460. https://doi.org/10.1023/A:1027351630866
Hainmueller, J. (2012). Entropy Balancing for Causal Effects: A Multivariate Reweighting Method to
Produce Balanced Samples in Observational Studies. Political Analysis, 20(1), 25–46.

38
Hainmueller, J., & Xu, Y. (2013). ebalance: A Stata Package for Entropy Balancing. Journal of Statistical
Software, 54, 1–18. https://doi.org/10.18637/jss.v054.i07
Hambrick, D. C. (2007). Upper Echelons Theory: An Update. Academy of Management Review, 32(2),
334–343. https://doi.org/10.5465/AMR.2007.24345254
Hambrick, D. C., & Mason, P. A. (1984). Upper Echelons: The Organization as a Reflection of Its Top
Managers. The Academy of Management Review, 9(2), 193–206. https://doi.org/10.2307/258434
Hanlon, M., Yeung, K., & Zuo, L. (2021). Behavioral Economics of Accounting: A Review of Archival
Research on Individual Decision Makers†. Contemporary Accounting Research, n/a(n/a).
https://doi.org/10.1111/1911-3846.12739
Hart Research Associates/Public Opinion Strategies. (2020). 200149 NBCWSJ March Poll Final (3 14 20
Release)—DocumentCloud. https://www.documentcloud.org/documents/6810602-200149-
NBCWSJ-March-PollFinal-3-14-20-Release.html
Hatemi, P. K., & McDermott, R. (2012). The Political Psychology of Biology, Genetics, and Behavior.
Political Psychology, 33(3), 307–312.
Hirshleifer, D., Low, A., & Teoh, S. H. (2012). Are Overconfident CEOs Better Innovators? The Journal
of Finance, 67(4), 1457–1498. https://doi.org/10.1111/j.1540-6261.2012.01753.x
Hope, O.-K., Hu, D., & Lu, H. (2016). The benefits of specific risk-factor disclosures. Review of
Accounting Studies, 21(4), 1005–1045. https://doi.org/10.1007/s11142-016-9371-1
Hope, O.-K., Li, C., Ma, M. S., & Su, X. (2022). Is silence golden sometimes? Management guidance
withdrawals during the COVID-19 pandemic. Review of Accounting Studies.
https://doi.org/10.1007/s11142-022-09698-w
Hu, S., Jia, X., Zhang, X., Zheng, X., & Zhu, J. (2017). How political ideology affects climate perception:
Moderation effects of time orientation and knowledge. Resources, Conservation and Recycling, 127,
124–131. https://doi.org/10.1016/j.resconrec.2017.09.003
Huang, X., Teoh, S. H., & Zhang, Y. (2014). Tone Management. The Accounting Review, 89(3), 1083–
1113. https://doi.org/10.2308/accr-50684
Hutton, I., Jiang, D., & Kumar, A. (2014). Corporate Policies of Republican Managers. The Journal of
Financial and Quantitative Analysis, 49(5/6), 1279–1310.
Jerit, J., & Barabas, J. (2012). Partisan Perceptual Bias and the Information Environment. The Journal of
Politics, 74(3), 672–684. https://doi.org/10.1017/S0022381612000187
Jost, J. T. (2006). The end of the end of ideology. American Psychologist, 61(7), 651–670.
https://doi.org/10.1037/0003-066X.61.7.651
Kempf, E., & Tsoutsoura, M. (2021). Partisan Professionals: Evidence from Credit Rating Analysts. The
Journal of Finance, 76(6), 2805–2856. https://doi.org/10.1111/jofi.13083
Kerr, J., Panagopoulos, C., & van der Linden, S. (2021). Political polarization on COVID-19 pandemic
response in the United States. Personality and Individual Differences, 179, 110892.
https://doi.org/10.1016/j.paid.2021.110892
Kyung, E., Thomas, M., & Krishna, A. (2021). How Political Identity Influences COVID-19 Risk
Perception: A Model of Identity-Based Risk Perception (SSRN Scholarly Paper 3915780).
https://doi.org/10.2139/ssrn.3915780
Li, F. (2011). Textual Analysis of Corporate Disclosures: A Survey of the Literature. Journal of
Accounting Literature, 40.
Loughran, T., & Mcdonald, B. (2011). When Is a Liability Not a Liability? Textual Analysis,
Dictionaries, and 10-Ks. The Journal of Finance, 66(1), 35–65. https://doi.org/10.1111/j.1540-
6261.2010.01625.x
Notbohm, M., Campbell, K., Smedema, A. R., & Zhang, T. (2019). Management’s personal ideology and
financial reporting quality. Review of Quantitative Finance and Accounting, 52(2), 521–571.
https://doi.org/10.1007/s11156-018-0718-5
Nguyen, H., Pham, M. H., & Truong, C. (2023). Work from Home Suitability and Credit Risk
Assessment. European Accounting Review, 1–34. https://doi.org/10.1080/09638180.2023.2239865

39
Painter, M., & Qiu, T. (2021). Political beliefs affect compliance with government mandates. Journal of
Economic Behavior & Organization, 185, 688–701. https://doi.org/10.1016/j.jebo.2021.03.019
Pew Research Center. (2020, June 25). Republicans, Democrats Move Even Further Apart in Coronavirus
Concerns. Pew Research Center - U.S. Politics & Policy.
https://www.pewresearch.org/politics/2020/06/25/republicans-democrats-move-even-further-apart-
in-coronavirus-concerns/
Ramelli, S., & Wagner, A. F. (2020). Feverish Stock Price Reactions to COVID-19*. The Review of
Corporate Finance Studies, 9(3), 622–655. https://doi.org/10.1093/rcfs/cfaa012
Ramirez, M. D., & Erickson, N. (2014). Partisan Bias and Information Discounting in Economic
Judgments: Partisan Bias and Information Discounting. Political Psychology, 35(3), 401–415.
https://doi.org/10.1111/pops.12064
Shapiro, R. Y., & Bloch‐Elkon, Y. (2008). Do The Facts Speak for Themselves? Partisan Disagreement
as A Challenge To Democratic Competence. Critical Review, 20(1–2), 115–139.
https://doi.org/10.1080/08913810802316373
Sherman, Nave, & Funder. (2010). Situational Similarity and Personality Predict Behavioral Consistency.
Journal of Personality and Social Psychology, 99(2), 330–343. https://doi.org/10.1037/a0019796
Stuart, M.D., Wang, J. & Willis, R.H. CEO partisan bias and management earnings forecast bias. Rev
Account Stud (2024). https://doi.org/10.1007/s11142-024-09846-4
Taber, C. S., & Lodge, M. (2006). Motivated Skepticism in the Evaluation of Political Beliefs. American
Journal of Political Science, 50(3), 755–769. https://doi.org/10.1111/j.1540-5907.2006.00214.x
Tedin, K. L. (1987). Political ideology and the vote. Research in Micropolitics, 2(1), 63–94.
Tetlock, P. E. (2000). Cognitive Biases and Organizational Correctives: Do Both Disease and Cure
Depend on the Politics of the Beholder? Administrative Science Quarterly, 45(2), 293–326.
https://doi.org/10.2307/2667073
Zhang, Y. T. (2015). CEO political ideology and firms’ financial reporting practice.
https://theses.lib.polyu.edu.hk/handle/200/9049

40
Figure 1: Trend of COVID Disclosures
Panel A: Trend of COVID Disclosures in Earnings Conference Calls
Note: This panel plots the trend of COVID disclosures in the presentation session of earnings conference calls, defined
as the word count of COVID-related sentences as a percentage of the total word count of the presentation session. The
trend line in red is estimated using a six-order polynomial function.
Panel B: Trend of COVID Disclosures in 10-Qs
Note: This panel plots the trend of COVID disclosures in 10-Qs, defined as the word count of COVID-related
sentences as a percentage of the total word count of the 10-Q. The trend line in red is estimated using a six-order
polynomial function.

41
Figure 2: Coefficients and Standard Errors Estimated Using Recursive Regressions
Note: This figure plots the coefficients and their standard errors (SEs) for the partisan leaning effect on the amount of COIVD disclosures in 10-Qs estimated
using recursive regressions on overlapping 90-day sub-samples at 5-day increments from the start of Q2 2020 to the end of 2021. The dates on the horizontal axis
indicate the start date of each window. The blue curve indicates the coefficients, and the length of each bar indicates two standard errors.
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
04/01/2020
04/11/2020
04/21/2020
05/01/2020
05/11/2020
05/21/2020
05/31/2020
06/10/2020
06/20/2020
06/30/2020
07/10/2020
07/20/2020
07/30/2020
08/09/2020
08/19/2020
08/29/2020
09/08/2020
09/18/2020
09/28/2020
10/08/2020
10/18/2020
10/28/2020
02/05/2021
02/15/2021
02/25/2021
03/07/2021
03/17/2021
03/27/2021
04/06/2021
04/16/2021
04/26/2021
05/06/2021
05/16/2021
05/26/2021
06/05/2021
06/15/2021
06/25/2021
07/05/2021
07/15/2021
07/25/2021
08/04/2021
08/14/2021
08/24/2021
09/03/2021
09/13/2021
09/23/2021
10/03/2021
10/13/2021
10/23/2021

42
TABLE 1: Sample Selection and Distribution
Panel A:
Sample Selection Process
Firm-Quarters
All Compustat firms with 10-Qs filed over Q2, 2020 – Q1, 2022
27,296
Less: Firm-quarter observations missing stock return data
(5,004)
22,292
Less: Firm-quarter observations missing Execucomp data
(13,821)
8,471
Less: Firm-quarter observations missing political leaning data
(3,676)
4,795
Less: Missing control variables
(4)
Final sample
4,791
Note: The sample consists of 4,791 firm-quarter observations (representing 908 unique firms) for 10-Qs filed from
April 01, 2020, to February 28, 2022. In terms of fiscal quarters, these firm-quarter observations are from Q1 2020
through Q3 2021 (excluding Q4 2020, for which firms filed annual reports, i.e., 10-Ks).
Panel B: Sample Distribution by Fiscal Quarter
Fiscal Quarter
Observations
Percent
Cum. Percent
2020 Q1
735
15.34
15.34
2020 Q2
824
17.20
32.54
2020 Q3
862
17.99
50.53
2021 Q1
793
16.55
67.08
2021 Q2
791
16.51
83.59
2021 Q3
786
16.41
100.00
Total
4,791
100.00

43
TABLE 1: Sample Selection and Distribution (Cont.)
Panel C: Industry Distribution
Frequency
Percent
Cum.
Mining
29
0.61
0.61
Oil and Gas Extraction
151
3.15
3.76
Nonmetalic Minerals Except Fuels
22
0.46
4.22
Construction
109
2.28
6.49
Food and Kindred Products
116
2.42
8.91
Tobacco Products
12
0.25
9.16
Textile Mill Products
9
0.19
9.35
Apparel and other Textile Products
53
1.11
10.46
Lumber and Wood Products
12
0.25
10.71
Furniture and Fixtures
22
0.46
11.17
Paper and Allied Products
21
0.44
11.61
Printing and Publishing
8
0.17
11.77
Chemicals and Allied Products
318
6.64
18.41
Petroleum and Coal Products
64
1.34
19.75
Rubber and Miscellaneous Plastics Products
54
1.13
20.87
Leather and Leather Products
9
0.19
21.06
Stone, Clay and Glass Products
8
0.17
21.23
Primary Metal Industries
54
1.13
22.35
Fabricated Metal Products
63
1.31
23.67
Machinery, Except Electrical
215
4.49
28.16
Electrical and Electronic Equipment
191
3.99
32.14
Transportation Equipment
157
3.28
35.42
Instruments and Related Products
235
4.91
40.33
Miscellaneous Manufacturing Industries
11
0.23
40.56
Transportation
136
2.84
43.39
Telephone and Telegraph Communication
27
0.56
43.96
Radio and Television Broadcasting
33
0.69
44.65
Electric, Gas, and Water Supply
235
4.91
49.55
Sanitary Services
24
0.5
50.05
Wholesale
114
2.38
52.43
Retail Stores
272
5.68
58.11
Finance, Insurance, and Real Estate
1,323
27.61
85.72
Services
661
13.8
99.52
Public Administration
11
0.23
99.75
Other
12
0.25
100
Total
4,791
100
Note: This table provides the sample distribution of firm-quarter observations by the Fama-French 38-industry
classification.

44
TABLE 2: Descriptive Statistics
Panel A: All Variables of Sample Firms
N
Mean
SD
P5
P25
P50
P75
P95
REP_INDEX
4,791
0.092
0.816
-1.000
-0.999
0.131
1.000
1.000
REP_DUM
4,791
0.491
0.417
0.000
0.000
0.500
1.000
1.000
COVID_DISC1
4,791
4.161
0.746
2.833
3.714
4.263
4.691
5.231
COVID_DISC2
4,791
0.417
0.232
0.104
0.241
0.374
0.556
0.880
NON_COVID_WC
4,791
9.712
0.486
8.929
9.391
9.698
10.022
10.551
COVID_TONE
4,791
-70.655
21.881
-100.000
-85.190
-75.680
-61.060
-27.780
NON_COVID_TONE
4,791
-40.247
18.671
-66.550
-54.440
-41.540
-28.570
-5.830
FOG_10Q
4,791
19.589
1.044
17.950
18.840
19.580
20.270
21.360
OVERCONFIDENCE
4,791
0.289
0.301
0.000
0.000
0.200
0.500
0.833
CAR
4,786
0.188
6.941
-9.998
-3.209
-0.213
2.889
11.415
ABS_CAR
4,786
4.779
5.394
0.274
1.361
3.071
6.249
14.817
SIZE
4,791
8.615
1.686
6.046
7.436
8.422
9.814
11.702
BTM
4,791
0.566
0.575
0.012
0.210
0.427
0.756
1.623
LEV
4,791
0.336
0.217
0.025
0.169
0.330
0.467
0.707
CASH
4,791
0.134
0.140
0.007
0.037
0.089
0.174
0.435
EARN
4,791
0.009
0.028
-0.027
0.001
0.008
0.020
0.051
LOSS
4,791
0.206
0.405
0.000
0.000
0.000
0.000
1.000
∆EARN
4,791
0.003
0.030
-0.035
-0.005
0.001
0.010
0.042
STD_EARN
4,791
0.039
0.058
0.002
0.010
0.021
0.043
0.133
SALES_GROWTH
4,791
0.121
0.576
-0.415
-0.078
0.040
0.185
0.787
FIRM_AGE
4,791
3.404
0.544
2.303
3.135
3.434
3.807
4.111
BUSSEGS
4,791
1.079
0.404
0.693
0.693
1.099
1.386
1.792
GEOSEGS
4,791
1.071
0.463
0.693
0.693
0.693
1.386
1.946
FORE_RATIO
4,791
0.203
0.247
0.000
0.000
0.082
0.371
0.686
FOLLOW
4,791
2.223
0.661
1.099
1.792
2.303
2.773
3.219
AFE
4,791
0.002
0.020
-0.014
0.000
0.002
0.005
0.023
AF
4,791
0.012
0.025
-0.011
0.006
0.012
0.020
0.040
BHRET
4,791
0.047
0.274
-0.419
-0.083
0.035
0.168
0.494
RET_VOL
4,791
0.032
0.019
0.012
0.018
0.026
0.042
0.070
INST_OWN
4,791
0.726
0.159
0.460
0.624
0.729
0.847
0.969
NUM_ITEMS
4,791
5.837
0.059
5.733
5.799
5.841
5.875
5.935
DERIVATIVE
4,791
0.663
0.473
0.000
0.000
1.000
1.000
1.000
TAC
4,789
-0.012
0.031
-0.058
-0.022
-0.009
0.000
0.029
SUE
4,696
0.319
1.561
-0.904
0.011
0.157
0.515
2.227
TELEWORK
4,791
0.395
0.232
0.140
0.220
0.250
0.720
0.760
CASES_NEW
4,791
2.361
1.189
0.164
1.662
2.507
3.187
4.029
CASES_DATE
4,791
3.294
1.687
0.185
2.144
3.729
4.654
5.047
CASES_FDATE
4,791
3.600
1.434
0.777
2.712
4.032
4.713
5.115

45
TABLE 3: Correlation Matrix
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[1]
REP_INDEX
0.970
-0.144
-0.072
-0.110
0.023
0.020
-0.105
0.043
0.013
[2]
REP_DUM
0.959
-0.154
-0.070
-0.122
0.025
0.026
-0.114
0.039
0.012
[3]
COVID_DISC1
-0.134
-0.152
0.735
0.447
-0.136
-0.295
0.333
0.029
-0.006
[4]
COVID_DISC2
-0.061
-0.064
0.693
-0.220
-0.034
-0.042
0.131
0.073
0.018
[5]
NON_COVID_WC
-0.116
-0.136
0.440
-0.256
-0.148
-0.394
0.306
-0.062
-0.029
[6]
COVID_TONE
0.038
0.042
-0.220
-0.104
-0.181
0.262
-0.010
0.053
0.009
[7]
NON_COVID_TONE
0.016
0.025
-0.301
-0.039
-0.379
0.282
-0.110
0.034
0.023
[8]
FOG_10Q
-0.109
-0.118
0.330
0.115
0.317
-0.053
-0.114
0.059
-0.023
[9]
OVERCONFIDENCE
0.046
0.040
0.008
0.054
-0.076
0.053
0.043
0.073
0.004
[10]
CAR
0.008
0.009
0.012
0.035
-0.023
0.014
0.013
-0.011
-0.003
Note: Pearson and Spearman correlations of the key variables are reported below (above) the diagonal. Correlation
coefficients with significance at the 5% level are boldfaced. Appendix I provides the definitions of all variables.

46
Table 4: Relation between Executive Political Leaning and Amount of COVID Disclosures
DV = COVID_DISC1
DV = COVID_DISC2
(1)
(2)
(3)
(4)
REP_INDEX
-0.0462*
-0.0116
(-1.96)
(-1.53)
REP_DUM
-0.1035**
-0.0265*
(-2.31)
(-1.83)
NON_COVID_WC
0.7671***
0.7662***
(18.53)
(18.55)
OVERCONFIDENCE
-0.0033
-0.0038
-0.0003
-0.0004
(-0.05)
(-0.06)
(-0.01)
(-0.02)
SIZE
-0.0351*
-0.0361*
-0.0168***
-0.0170***
(-1.82)
(-1.88)
(-2.70)
(-2.75)
BTM
-0.0434
-0.0451
-0.0280**
-0.0285**
(-1.30)
(-1.35)
(-2.46)
(-2.50)
LEV
-0.0678
-0.0696
0.0059
0.0055
(-0.69)
(-0.71)
(0.18)
(0.17)
FIRM_AGE
0.0252
0.0238
0.0114
0.0111
(0.74)
(0.7)
(0.95)
(0.92)
CASH
-0.1039
-0.1000
-0.0053
-0.0045
(-0.69)
(-0.66)
(-0.11)
(-0.09)
EARN
-0.3588
-0.3327
0.2048
0.2129
(-0.50)
(-0.47)
(0.86)
(0.90)
LOSS
0.0184
0.0184
0.0161
0.0161
(0.55)
(0.55)
(1.29)
(1.29)
∆EARN
-0.3201
-0.3302
-0.2091
-0.2125
(-0.69)
(-0.71)
(-1.35)
(-1.37)
STD_EARN
-0.5778*
-0.5767*
-0.1835
-0.1832
(-1.79)
(-1.80)
(-1.60)
(-1.60)
SALES_GROWTH
0.0233
0.0228
0.0028
0.0026
(1.30)
(1.27)
(0.48)
(0.46)
BHRET
-0.1370***
-0.1375***
-0.0366***
-0.0367***
(-3.34)
(-3.36)
(-2.68)
(-2.69)
RET_VOL
1.4674
1.4052
0.2227
0.2057
(1.38)
(1.33)
(0.57)
(0.53)
BUSSEGS
-0.0336
-0.0322
-0.0106
-0.0103
(-0.78)
(-0.75)
(-0.79)
(-0.76)
GEOSEGS
-0.0279
-0.0284
-0.0065
-0.0066
(-0.54)
(-0.55)
(-0.34)
(-0.35)
FORE_RATIO
-0.1707*
-0.1679*
-0.0563*
-0.0556
(-1.83)
(-1.81)
(-1.66)
(-1.64)
FOLLOW
0.0832**
0.0826**
0.0231*
0.0230*
(2.02)
(2.00)
(1.69)
(1.68)

47
Table 4: Relation between Executive Political Leaning and Amount of COVID Disclosures
(Cont.)
AFE
0.6425*
0.6414*
0.2706*
0.2704*
(1.74)
(1.74)
(1.86)
(1.86)
AF
-1.2257**
-1.1990**
-0.2700
-0.2632
(-2.19)
(-2.14)
(-1.56)
(-1.52)
INST_OWN
0.0013
0.0028
0.0195
0.0199
(0.01)
(0.02)
(0.41)
(0.42)
NUM_ITEMS
-0.1462
-0.1448
0.0642
0.0651
(-0.43)
(-0.43)
(0.56)
(0.57)
DERIVATIVE
0.0201
0.0192
-0.0011
-0.0014
(0.53)
(0.51)
(-0.09)
(-0.11)
CASES_NEW
-0.0349
-0.035
-0.0086
-0.0087
(-1.24)
(-1.25)
(-1.49)
(-1.49)
CASES_DATE
0.011
0.0128
0.0079
0.0083
(0.38)
(0.44)
(0.77)
(0.81)
TELEWORK
-0.5475***
-0.5407***
-0.2743***
-0.2727***
(-3.32)
(-3.29)
(-4.79)
(-4.76)
FE (Time, Industry, and State)
Yes
Yes
Yes
Yes
Adjusted R-squared
0.483
0.483
0.386
0.386
N
4,791
4,791
4,791
4,791
Note: This table reports the results for the relation between executive political leanings and the amount of COVID-19
disclosure in 10-Q filings. Columns (1) and (2) present the regression results where COVID_DISC1 (the log-
transformed COVID-19 keyword frequency) is the dependent variable. Column (1) examines the association between
REP_INDEX and COVID_DISC1. Column (2) examines the association between REP_DUM and COVID_DISC1.
Columns (3) and (4) present the regression results where COVID_DISC2 (the relative COVID-19 keyword frequency)
is the dependent variable. Column (3) examines the association between REP_INDEX and COVID_DISC2. Column
(4) examines the association between REP_DUM and COVID_DISC2. All t-statistics are reported in parentheses
below the coefficients, with robust standard errors clustered at the firm level. *, **, *** denote significance at 10%,
5% and 1%, respectively (two-tailed). Time fixed effects are based on the filing month of 10-Qs.

48
Table 5: Relation between Executive Political Leaning and Tone of COVID Disclosures
DV = COVID_TONE
Not controlling for Non-COVID
Controlling for Non-COVID
(1)
(2)
(3)
(4)
REP_INDEX
1.5871**
1.4678*
(1.97)
(1.89)
REP_DUM
3.0199*
2.7469*
(1.93)
(1.83)
NON_COVID_TONE
0.2578***
0.2576***
(7.40)
(7.40)
Controls
Yes
Yes
Yes
Yes
FE (Time, Industry, and State)
Yes
Yes
Yes
Yes
Adjusted R-squared
0.149
0.149
0.188
0.188
N
4,791
4,791
4,791
4,791
Note: This table reports the results for the relation between executive political leanings and the tone of COVID-19
disclosure in 10-Q filings. Columns (1) and (2) do not control for the tone of non-COVID disclosure. Columns (3)
and (4) control for the tone of non-COVID disclosure. Columns (1) and (3) examine the association between
REP_INDEX and COVID_TONE. Columns (2) and (4) examine the association between REP_DUM and
COVID_TONE. All t-statistics are reported in parentheses below the coefficients, with robust standard errors clustered
at the firm level. *, **, *** denote significance at 10%, 5% and 1%, respectively (two-tailed). Coefficients and t-
statistics for other control variables are suppressed for brevity. Time fixed effects are based on the filing month of 10-
Qs.

49
Table 6: Relation between Executive Political Leaning and COVID Disclosures: Partisan Alignment
Panel A: Disclosure Amount
DV = COVID_DISC1
DV = COVID_DISC2
(1)
(2)
(3)
(4)
Pre-Period
Post-Period
Pre-Period
Post-Period
Pre-Period
Post-Period
Pre-Period
Post-Period
REP_INDEX
-0.0539***
-0.0464
-0.0173*
-0.0109
(-2.62)
(-1.44)
(-1.94)
(-1.30)
REP_DUM
-0.1263***
-0.0993
-0.0428**
-0.0206
(-3.20)
(-1.62)
(-2.50)
(-1.30)
Controls and FE (Time, Industry, and State)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Adj R2
0.389
0.383
0.391
0.383
0.323
0.268
0.325
0.268
N
2,434
2,357
2,434
2,357
2,434
2,357
2,434
2,357
Panel B: Disclosure Tone
DV = COVID_TONE
Not controlling for Non-COVID
Controlling for Non-COVID
(1)
(2)
(3)
(4)
Pre-Period
Post-Period
Pre-Period
Post-Period
Pre-Period
Post-Period
Pre-Period
Post-Period
REP_INDEX
1.5001**
1.7045
1.4230**
1.4881
(2.16)
(1.46)
(2.16)
(1.30)
REP_DUM
2.6931**
3.3422
2.5197*
2.8747
(1.99)
(1.47)
(1.96)
(1.30)
Controls and FE (Time, Industry, and State)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Adj R2
0.107
0.109
0.107
0.109
0.150
0.148
0.149
0.148
N
2,434
2,357
2,434
2,357
2,434
2,357
2,434
2,357
Note: This table reports the results for the relation between executive political leanings and the amount (Panel A) and tone (Panel B) of COVID-19 disclosures in
the pre-period (post-period) when a Republican (Democratic) served as the U.S. president. All t-statistics are reported in parentheses below the coefficients, with
robust standard errors clustered at the firm level. *, **, *** denote significance at 10%, 5% and 1%, respectively (two-tailed). Coefficients and t-statistics for
control variables are suppressed for brevity. Time fixed effects are based on the filing month of 10-Qs.

50
Table 7: Market Reactions to COVID-19 Disclosures
DV =
CAR (1)
DV =
ABS_CAR (2)
DV =
CAR (3)
DV =
ABS_CAR (4)
DV =
CAR (5)
DV =
ABS_CAR (6)
DV =
CAR (7)
DV =
ABS_CAR (8)
REP_INDEX/_DUM
-0.1244
-0.1663
0.0381
0.4896
0.4521
-0.0444
0.3212
1.0406
(-0.17)
(-0.28)
(0.08)
(1.35)
(0.32)
(-0.04)
(0.36)
(1.47)
COVID_DISC1
0.1469
0.0744
0.2297
-0.0645
(0.71)
(0.40)
(0.87)
(-0.28)
COVID_DISC2
0.5939
0.2761
1.2408
0.2415
(0.99)
(0.58)
(1.43)
(0.35)
REP_INDEX/_DUM
-0.0009
0.2163
-0.5155
0.3453
-0.1658
0.3161
-1.4362
0.1224
* COVID_DISC1/2
(-0.01)
(1.49)
(-0.82)
(0.76)
(-0.50)
(1.21)
(-1.17)
(0.14)
COVID_TONE
0.0036
0.0067*
0.0037
0.0070**
0.0051
-0.0029
0.0054
-0.0016
(0.74)
(1.94)
(-0.77)
(2.01)
(0.67)
(-0.53)
(0.71)
(-0.30)
REP_INDEX/_DUM
-0.0023
0.0118***
-0.003
0.0104***
-0.0035
0.0218***
-0.0042
0.0194**
* COVID_TONE
(-0.42)
(2.84)
(-0.57)
(2.64)
(-0.33)
(2.67)
(-0.40)
(2.51)
Controls
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
FE (Time, Industry)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Adjusted R-squared
0.050
0.176
0.050
0.176
0.050
0.176
0.051
0.176
N
4,689
4,689
4,689
4,689
4,689
4,689
4,689
4,689
Note: This table reports the results for market reactions to COVID-19 disclosures conditional on the political leaning of the executives. The main variables of
interest are the interaction terms of political leaning (REP_INDEX or REP_DUM) with COVID disclosure attributes (COVID_DISC1, COVID_DISC2, and
COVID_TONE), indicating whether the market reacts differently to COVID disclosures depending on the political leaning of the executives who provide the
disclosures. The dependent variable in Columns (1), (3), (5), and (7) is CAR. The dependent variable in Columns (2), (4), (6), and (8) is ABS_CAR. Columns (1)
and (2) examine the abnormal return related to REP_INDEX, COVID_DISC1, and COVID_TONE. Columns (3) and (4) examine the absolute value of abnormal
return related to REP_INDEX and COVID_DISC2 and COVID_TONE. Columns (5) and (6) examine the abnormal return related to REP_DUM, COVID_DISC1,
and COVID_TONE. Column (7) and (8) examine the absolute value of abnormal return to REP_DUM, COVID_DISC2, and COVID_TONE. All t-statistics are
reported in parentheses below the coefficients, with robust standard errors clustered at the firm level. *, **, *** denote significance at 10%, 5% and 1%, respectively
(two-tailed). Coefficients and t-statistics for control variables are suppressed for brevity. Time fixed effects are based on the filing month of 10-Qs.

51
Table 8: Market Reactions to COVID-19 Disclosures: Partisan Alignment
DV =
CAR (1)
DV =
ABS_CAR (2)
DV =
CAR (3)
DV =
ABS_CAR (4)
COVID_DISC1
-0.0249
0.4882
(-0.04)
(1.29)
COVID_DISC2
0.2430
1.5624
(0.15)
(1.42)
COVID_TONE
0.0258*
-0.0036
0.0262*
-0.003
(1.89)
(-0.42)
(1.94)
(-0.35)
MF_BIAS
5.951
5.5142
5.9191
5.4522
(0.74)
(1.07)
(0.73)
(1.07)
ALIGN
-2.5681
0.8333
-2.3589*
1.0252
(-1.11)
(0.56)
(-1.87)
(1.12)
COVID_DISC1 * ALIGN
0.1133
0.1094
0.4955
0.5923
(0.21)
(0.30)
(0.28)
(0.38)
COVID_TONE * ALIGN
-0.0266*
0.0249**
-0.0272*
0.0248**
(-1.84)
(2.44)
(-1.92)
(2.47)
MF_BIAS * ALIGN
-9.8337
-5.6066
-9.6622
-5.3063
(-1.11)
(-0.78)
(-1.10)
(-0.74)
Controls
Yes
Yes
Yes
Yes
FE (Time and Industry)
Yes
Yes
Yes
Yes
Adjusted R-squared
0.080
0.209
0.080
0.210
N
1,194
1,194
1,194
1,194
Note: This table reports the results for market reactions to COVID-19 disclosures controlling for the effect of partisanship alignment and management forecast
bias. The main variables of interest are the interaction terms of partisan alignment (ALIGN) with COVID disclosure attributes (COVID_DISC1, COVID_DISC2,
and COVID_TONE) as well as management forecast bias (MF_BIAS). The dependent variable in Columns (1) and (3) is CAR. The dependent variable in Columns
(2) and (4) is ABS_CAR. Columns (1) and (2) examine the abnormal return COVID_DISC1. Columns (3) and (4) examine the absolute value of abnormal return to
COVID_DISC2. All t-statistics are reported in parentheses below the coefficients, with robust standard errors clustered at the firm level. *, **, *** denote
significance at 10%, 5% and 1%, respectively (two-tailed). Coefficients and t-statistics for control variables are suppressed for brevity. Time fixed effects are based
on the filing month of 10-Qs.

52
Table 9: Relation between Mandatory and Voluntary COVID Disclosures
DV = COVID_DISC1
DV = COVID_DISC2
(1)
(2)
(3)
(4)
REP_INDEX
-0.0457**
-0.0122
(-2.03)
(-1.63)
REP_DUM
-0.1011**
-0.0275*
(-2.36)
(-1.92)
COVID_DISC_CALL
0.1696***
0.1695***
0.0457***
0.0457***
(10.88)
(10.89)
(8.24)
(8.24)
Controls
Yes
Yes
Yes
Yes
FE (Time, Industry, and State)
Yes
Yes
Yes
Yes
Adjusted R-squared
0.526
0.527
0.415
0.415
N
4,438
4,438
4,438
4,438
Note: This table reports the results for the relation between the amount of mandatory and voluntary COVID-19
disclosures. Columns (1) and (2) present the regression results where COVID_DISC1 is the dependent variable.
Columns (3) and (4) present the regression results where COVID_DISC2 is the dependent variable. All t-statistics are
reported in parentheses below the coefficients, with robust standard errors clustered at the firm level. *, **, *** denote
significance at 10%, 5% and 1%, respectively (two-tailed). Coefficients and t-statistics for control variables are
suppressed for brevity. Time fixed effects are based on the filing month of 10-Qs.

53
TABLE 10: Impact of Political Leaning on COVID Disclosure Tone for Non-S&P 1500 Firms
using Predicted Values
DV = COVID_TONE
Not controlling for Non-COVID
Controlling for Non-COVID
(1)
(2)
REP_PREDICTED
3.1899***
2.5773**
(3.01)
(2.52)
Controls
Yes
Yes
FE (Time, Industry, and State)
Yes
Yes
Adjusted R-squared
0.125
0.165
N
10,124
10,124
Note: This table reports the results for the impact of political leaning on COVID disclosure tone for non-S&P 1500
firms using predicted political leaning. We train a Support Vector Machine (SVM) model on firms with political
leaning and use the model to predict the political leanings of firms lacking political leaning data. REP_PREDICTED
is the predicted value of political leaning for firms for which we do not have actual political leaning data. All t-statistics
are reported in parentheses below the coefficients, with robust standard errors clustered at the firm level. *, **, ***
denote significance at 10%, 5% and 1%, respectively (two-tailed). Coefficients and t-statistics for control variables
are suppressed for brevity. Time fixed effects are based on the filing month of 10-Qs.