How Can a Firm Suppress Shareholders’ Punitive Reaction to Its Disengagement from a Geopolitically Uncertain Market?

Corporate Disengagement from South Africa

The first event characterizes the pressure on U.S. firms during the 1980s to help end apartheid in South Africa. Although U.S. firms and their stakeholder groups agreed that apartheid should end—like the broader disapproval of Russia's invasion of Ukraine—C-level officers and directors were divided on how corporate disengagement from South Africa could help expedite apartheid's demise (Meznar, Nigh, and Kwok 1994). Despite the lack of clarity, hundreds of U.S. firms disengaged from South Africa.

On the one hand, disengagement hurt a firm's sales and prevented it from salvaging its assets (Meznar, Nigh, and Kwok 1994). Therefore, the intuitive hypothesis is that shareholders reacted negatively. On the other hand, once the firm announced its disengagement, protestors removed it from their “hit lists,” allowing the firm to save the resources it expended to manage the issue (Posnikoff 1997). Additionally, the firm became attractive to government customers and socially conscious investors (Meznar, Nigh, and Kwok 1994). In such cases, the benefits outweighed the costs, and shareholders rewarded the firm (Meznar, Nigh, and Kwok 1994). Consistent with the two-sided arguments, the empirical evidence is split, with some studies reporting a negative average reaction (Meznar, Nigh, and Kwok 1994; Wright and Ferris 1997), others positive (Posnikoff 1997), and still others documenting no reaction (McWilliams and Siegel 1997). This conflicting evidence led Meznar, Nigh, and Kwok (1998, p. 729) to conclude that “the question of the impact of announcements of withdrawal from South Africa on shareholder wealth has not been entirely resolved.” The mixed evidence thus falls short of guiding managers on whether they should exit a geopolitically uncertain market.

Corporate Disengagement from Designated “State Sponsors of Terrorism”

The second shock relates to the U.S. Department of State designating some countries as “State Sponsors of Terrorism.” ⁴ Currently, four countries have this designation: Cuba, North Korea, Iran, and Syria. Countries formerly on this list include Iraq, Libya, South Yemen, and Sudan. This designation represents the broader phenomenon of governmental resolutions restricting trade with countries that violate international standards. U.S. firms operating in countries that received this designation came under stakeholder pressure to exit these countries (Ganesan and Mallapragada 2025). For example, U.S. state pension funds and university endowment funds—two important institutional shareholders—started divesting from firms operating in designated countries. Such divestment was a tactic to influence these countries’ attitudes toward their violation of international norms (Connelly et al. 2010). In response to such pressures, several U.S. firms exited the designated countries. Interestingly, these firms received a reward from their shareholders in the short term (Breuer, Felde, and Steininger 2017); however, firms that chose to stay received abnormally positive stock returns in the long term (Breuer, Felde, and Steininger 2017). The theoretical insight is that shareholder reactions may not be consistent over time (Guo and Yu 2024).

Corporate Disengagement from Russia

The third shock is Russia's invasion of Ukraine on February 24, 2022. Table WA1 in Web Appendix A summarizes the empirical evidence on shareholder reactions to firm announcements of disengagement from Russia. Like the evidence on firm withdrawal from South Africa, the evidence on disengagement from Russia is split.

Choy, Lai, and Wan (2023), Glambosky and Peterburgsky (2022), and Kovalenko, Rangaswamy, and Sorescu (2023) have reported that shareholders punish firms that leave Russia. In contrast, Kiesel and Kolaric (2023) found that firms that announced their exit experienced positive reactions, whereas firms that chose to stay received no impact. Interestingly, Tosun and Eshraghi (2022) found that shareholders penalize firms that stay in Russia. Balyuk and Fedyk (2023) have found that U.S. firms’ exit announcements elicit an insignificant reaction from their shareholders. We attribute the divergent evidence to the authors’ (1) theoretical reasoning, (2) events of interest, and (3) methodological focus.

First, prior research has suggested distinct theoretical explanations for shareholders’ reactions to disengagement announcements. On the one hand, disengagement from a geopolitically uncertain market signals a loss of future revenue and, by extension, a loss of shareholder value (e.g., Wright and Ferris 1997), eliciting shareholder penalty. On the other hand, announcing disengagement from uncertain markets will lower the odds that protestors will publicly denounce and boycott the firm, thus enabling the firm to save resources that would have been expended to manage the issue. Additionally, the firm may become attractive to government customers and socially conscious investors (Posnikoff 1997). Of course, the two opposite theoretical perspectives may cancel each other out, leading to insignificant shareholder reactions (McWilliams and Siegel 1997).

Second, some authors’ event of interest is the onset of the war, not a firm's disengagement announcement. Consequently, these authors have measured the impact of the event on the stock market–level returns rather than firm-specific returns (Boungou and Yatié 2022; Diaconaşu, Mehdian, and Stoica 2023).

Third, prior studies also vary in methodological choices. Some (e.g., Tosun and Eshraghi 2022) have not conducted an event study or estimated a cross-sectional regression to explore heterogeneity in firm-specific returns (e.g., French, Gurdgiev, and Shin 2023; Pajuste and Toniolo 2022; Sonnenfeld et al. 2022). Others have checked neither sample-selection bias nor confounded announcements (e.g., Balyuk and Fedyk 2023; Basnet, Blomkvist, and Galariotis 2022; Berninger, Kiesel, and Kolaric 2022). Among the studies that have explored heterogeneity in firm-specific returns, some have used longer event windows with few covariates (Berninger, Kiesel, and Kolaric 2022; Huang et al. 2024). The set of commonly explored heterogeneity-explaining covariates includes the level of disengagement (Basnet, Blomkvist, and Galariotis 2022; Kovalenko, Rangaswamy, and Sorescu 2023), corporate social responsibility (CSR) performance scores (Basnet, Blomkvist, and Galariotis 2022; Choy, Lai, and Wan 2023), delay in disengagement announcement (Balyuk and Fedyk 2023; Kovalenko, Rangaswamy, and Sorescu 2023), the country or industry of sampled firms (Basnet, Blomkvist, and Galariotis 2022; Berninger, Kiesel, and Kolaric 2022; Huang et al. 2024), and the firm's stock returns when Russia invaded Ukraine (Kiesel and Kolaric 2023; Kovalenko, Rangaswamy, and Sorescu 2023). We control for these and other relevant variables and thus attempt to reconcile the divergent evidence while focusing on the impression management tactics that a firm can use to mitigate shareholder penalty.

Summary

Three points summarize the prior literature on these three shocks to the geopolitical certainty of a market. First, the dominant evidence from this literature is that shareholders react punitively to disengagement announcements. Therefore, we expect corporate announcements of disengagement from Russia to elicit—on average—a negative reaction from shareholders because the disengagement can harm a firm's short-term revenue and shareholders’ wealth. Second, shareholder reactions change over time. They may react punitively in the short term but become accommodating in the longer term. Third, the literature lacks evidence of how managers can mitigate shareholders’ punitive reactions. Consequently, using as context firms’ disengagement from Russia, we theorize how the linguistic framing of disengagement announcements can shape the shareholders’ punitive reactions.

Market Emphasis and Social Emphasis as Linguistic Frames

A firm can frame its announcement from two vantage points, emphasizing (1) its product-market activities and stakeholders or (2) societal activities and stakeholders.

First, several countries in the West imposed sanctions on Russia in response to its invasion of Ukraine (Maloney and Gryta 2020). For example, Western governments removed Russian banks from the SWIFT system facilitating international payments (SWIFT 2022). Russian aircraft and ships were banned from U.S. ports (Shepardson 2022) and most European airspace (Partridge 2022). Some governments even imposed surcharges on firms that earned revenues from Russia (Temnycky 2022). These governmental sanctions imply that firms have trouble sourcing supplies, making payments, and distributing their offerings in Russia (Gershkovich, Boston, and Kantchev 2022; Stiff 2023). Relatedly, staying in Russia may be fraught with political uncertainty. A firm's disengagement announcement can emphasize the difficulty of conducting business in Russia. For example, AAK AB stated, “It has become very difficult to secure compliance to sanctions related to logistics and trade flows as well as third parties” (AAK 2022, p. 1). Relatedly, the firm can mention consumer backlash in its announcements. For example, U.S. consumers were less willing to buy French-sounding products when the relationship between the United States and France deteriorated in 2003 (Pandya and Venkatesan 2016; see Web Appendix WG for concrete examples). We use the term market emphasis for disengagement announcements’ mentions of product-market stakeholders (specifically, customers, suppliers, and rivals) and activities (e.g., sourcing supplies and distributing products).

Second, some firm announcements mention societal activities and stakeholders. Specifically, firms may reference the problems their employees and families face due to geopolitical conflicts, condemn the environmental impact of the war, and emphasize their efforts to provide safety and security to the community. These firms believe that disengaging from Russia is a social imperative even though their Russian markets may not have been directly disrupted. For example, in its disengagement announcement, McDonald's Corporation reasoned that “continued ownership of the business in Russia is no longer tenable, nor is it consistent with McDonald's values” (McDonald's Corporation 2022, p. 1) (see Web Appendix WG). We use social emphasis to refer to disengagement announcements’ mentions of societal stakeholders (specifically, employees, environment, and community) and activities (e.g., promoting employee welfare and community safety).

A firm's product-market activities and stakeholders directly impact its financial performance. Therefore, we contend that the more the announcement mentions product-market activities and stakeholders, the stronger the shareholders’ interpretations that the firm is disengaging to minimize potential negative implications for its shareholders’ wealth (Bhagwat et al. 2020; DesJardine, Zhang, and Shi 2023). More concretely, the announcement's market emphasis may lead shareholders to infer or assume that the managers have analyzed the costs and benefits of continuing (vs. discontinuing) operations in Russia and concluded that discontinuing is an economically wiser decision, helping the firm avoid future losses. A disengagement announcement with market emphasis will lower the firm's costs of managing the issue, particularly activists’ threats of boycotting the firm (Hydock, Paharia, and Weber 2019). As a result, the announcement's market emphasis would be positively associated with shareholder reactions.

A firm's societal activities and stakeholders do not directly influence its financial performance (Hydock, Paharia, and Weber 2019). Therefore, the greater the announcement's social emphasis, the stronger the shareholders’ interpretation that the managerial decision was triggered by nonmarket reasons instead of preventing immediate financial losses (Bhagwat et al. 2020; Ganesan and Mallapragada 2025). Thus, a disengagement announcement's social emphasis will be negatively associated with shareholder reactions to the announcement. However, we acknowledge a competing perspective: A firm's customers and employees may interpret its disengagement announcement as a sacrifice, which is a testament to its values. As a result, shareholders may believe that the announcement's social emphasis may improve the firm's outcomes in product and labor markets (Ganesan and Mallapragada 2025; Tsougkou et al. 2025). In addition, social emphasis can help a firm avoid a public relations crisis (Ganesan and Mallapragada 2025) and strengthen its long-term competitiveness (Vishwanathan et al. 2020). These explanations suggest that the announcement's social emphasis may not necessarily elicit shareholders’ punitive reactions. However, research has shown that these gains are uncertain and polarizing (Hydock, Paharia, and Weber 2019; Mukherjee and Althuizen 2020; Weber et al. 2023). Therefore, shareholders will likely interpret social emphasis’ costs to outweigh its benefits (Meznar, Nigh, and Kwok 1994), leading to a negative association between social emphasis in disengagement announcements and shareholder reactions.

Social Emphasis Weakens Market Emphasis’ Positive Association with Shareholder Reactions

A disengagement announcement's social emphasis could bolster consumers’ and employees’ perceptions of the firm (Ganesan and Mallapragada 2025). Thus, high social emphasis may lend a “halo” to an announcement characterized by high market emphasis, leading shareholders to interpret the announcement as rationally holistic. This logic suggests that social emphasis may strengthen market emphasis’s positive effect on shareholders’ punitive reaction to disengagement announcements. Alternatively stated, social emphasis may generate a negative effect unless paired with market emphasis (Bhagwat et al. 2020).

However, we argue that a disengagement announcement's social emphasis can weaken the positive association between the announcement's market emphasis and shareholders’ punitive reaction to it. Accounting research has shown that shareholders respond negatively to linguistically complex disclosures (e.g., Bushee and Huang 2024), particularly when the information relates to uncertainty (e.g., You and Zhang 2009). The evidence suggests that shareholders are likely to interpret a disengagement announcement's dual emphasis as obfuscation rather than as complementary information (Bushee and Huang 2024). Consider disengagement announcement A, which unambiguously mentions the firm's product-market activities and stakeholders. Now, juxtapose announcement A with its counterpart B, where the firm mentions product-market and societal activities and stakeholders. We reason that shareholders will find announcement A more informative and less complex than announcement B (Miller 2010). Thus, the complexity of a dual emphasis announcement can lead to a stronger punitive reaction. These reasons lead us to predict:

Delay's Weakening of Market Emphasis’ Positive Association with Shareholder Reactions

Firms differ in terms of not only how they frame their disengagement in announcements but also when to announce the disengagement. For example, on February 25, 2022, TSMC—the world's largest semiconductor manufacturer—led the global computer chip industry by announcing halting its sales to Russia. Similarly, on February 28, 2022, Swedish automaker Volvo Cars announced the suspension of its car shipments to the Russian market, thus becoming the first international carmaker to announce disengagement from Russia (Reuters 2022a). Other firms announced their disengagement when the costs of conducting business with Russia grew. For instance, Dell waited until August 27, 2022, to announce that it had closed all its Russian operations (Reuters 2022b), and Kia Corporation announced on October 25, 2022, that it was considering winding down its exposure in Russia (Automotive News 2022). Next, we explore how the delay in announcing disengagement moderates the main effects of the linguistic frames of market emphasis and social emphasis.

As reasoned in H₁, a high level of market emphasis can suppress shareholders’ punitive reactions. A delayed announcement with high market emphasis may indicate that managers have taken the time to carefully prepare for the exit (Basnet, Blomkvist, and Galariotis 2022; Glambosky and Peterburgsky 2022; Woolridge and Snow 1990). However, a delay could indicate that managers were ill-prepared to manage the event and waited to assess its product-market consequences. Therefore, shareholders may react punitively toward a reactive (vs. proactive) disengagement announcement characterized by high market emphasis. For instance, Kia Corporation announced disengagement after its suppliers or customers exited the market (Automotive News 2022). Thus, although Kia exhibited high market emphasis in its disengagement announcement, the associated delay suggested that the firm did not actively safeguard shareholders’ interests and was instead forced out of the market (Dalton et al. 2007; Jensen and Meckling 1976). In contrast, if a firm announces disengagement early and the announcement emphasizes the product market, shareholders may perceive that the firm has proactively undertaken actions to protect the former's interests.

Delay's Strengthening of Social Emphasis’ Negative Association with Shareholder Reactions

H₂ suggests that a disengagement announcement's social emphasis is negatively associated with shareholders’ reactions. Next, we reason why a delayed announcement with social emphasis is worse for the firm than its early counterpart—a hypothesis consistent with the preceding discussion on a delayed announcement with market emphasis performing worse than its early counterpart.

As time passes and geopolitical conflict takes a greater toll on local communities, the firm faces greater social and political pressure to disengage as a geopolitical conflict persists (Posnikoff 1997). This pressure may come from customers, suppliers, employees, and communities (Ganesan and Mallapragada 2025; Pajuste and Toniolo 2022). Specifically, these stakeholders could perceive a delayed announcement but with social emphasis as reactive and inauthentic rather than proactive and responsible (Ganesan and Mallapragada 2025). Under such a scenario, if firms exhibit strong social emphasis in their disengagement announcements, stakeholders may perceive that the firm is forced to disengage, and they may not believe that the social emphasis is authentic. Relatedly, the longer the firm takes to make a disengagement announcement with social emphasis, the weaker its anticipated rewards from these stakeholders. Thus, the firm's shareholders are less likely to consider these potential rewards when reacting to a late (vs. early) disengagement announcements with strong social emphasis.

Sample Construction

We obtained our list of firms and their disengagement announcements (if made) from three sources: (1) The Chief Executive Leadership Institute at the Yale School of Management (“the Yale Institute” hereinafter), (2) Kyiv School of Economics (KSE) Institute, and (3) The Good Lobby's and the Progressive Shopper's Ukraine Corporate Index (“The Good Lobby” hereinafter). ⁶ We obtained the list on August 8, 2023, thus considering about 17 months following Russia's invasion. We aim to measure stock returns to disengagement announcements by firms whose common stock is traded on a major U.S. or non-U.S. stock exchange. Because the Yale Institute and the KSE Institute provide a list of firms that did not disengage, we used these firms for the Heckman selection model. Next, we followed five steps to select our “announcement sample.” The aforementioned three sources provided us with a population of 1,980 announcements. First, we excluded announcements from sports and nonprofit organizations, such as Fédération Internationale de Football Association (FIFA). This step led us to 1,894 announcements.

Second, we searched for the names of the remaining firms on Compustat and Yahoo! Finance to determine whether the firm was publicly traded. Of the 1,894 announcements, 1,003 were made by private firms (i.e., either the stock was not publicly traded in any country or the stock prices were not available at the time of announcement), and we thus excluded these announcements from our sample. This procedure led to 891 announcements by 813 firms. Seventy-eight firms made between one and six announcements about their subsidiaries’ disengagement on the same date or very close dates. We kept the first announcement from such firms—that is, we excluded their announcements after the first. This exclusion led to 813 announcements by 813 firms (i.e., one announcement per firm).

Third, the data file from the Yale Institute listed the dates of most (but not all) announcements. We supplemented the Institute's effort by reading the press releases and newsroom section on the website of each of the 813 firms. We also searched the Nexis Uni (formerly LexisNexis Academic) database and popular news sources, such as the Wall Street Journal, New York Times, Reuters, and Bloomberg. At the end of this step, we had the exact date for each of the 813 announcements. Of these 813 announcements, 298 were made by 298 U.S. firms listed on either Nasdaq or the New York Stock Exchange (NYSE), and 515 announcements were made by 515 firms listed on stock exchanges across 45 countries in Europe, Asia, Africa, Oceania, South America, and North America (excluding the United States).

Fourth, following prior research (Boyd, Chandy, and Cunha 2010; Warren and Sorescu 2017), for each announcement, we checked for the presence of any confounding news specific to the focal firm five days before and five days after the focal disengagement announcement. We considered confounding news of CEO appointments, earnings conference calls, and new product announcements. Of the 813 announcements, 141 were confounded—that is, they had at least one piece of confounding firm-level news in the [−5,5] window centered on the date of the focal disengagement announcement—leaving us with a sample of 672 announcements. We report robustness to including confounded announcements—that is, a sample of 813 announcements.

Fifth, for each of the 672 firms, we checked Compustat North America and Compustat Global on whether the database reports data on the announcing firm's main and control variables for the year 2021 and/or 2022 (e.g., current assets, book value of equity, and sales). One hundred eight-five firms did not have the data for these variables. This procedure resulted in 487 announcements by 487 firms, 208 of them traded on either the NYSE or Nasdaq, and 279 public firms traded on stock exchanges in 24 countries (Table WB1 lists the names of countries in our sample). We report robustness to the inclusion of the 185 announcements. The 487 public firms in our sample operate in 53 industries based on their two-digit Standard Industrial Classification (SIC) codes. The industries with a substantial number of firms in our sample are Business Services (73), Industrial and Commercial Machinery and Computer Equipment (35), Chemical and Allied Products (28), and Electronic and Other Electrical Equipment and Components (36), with 66, 52, 46, and 35 announcements, respectively (Table WB2 lists the names of industries and the number of firms in each industry in our sample). Table 1 reports the measure and source for each variable in our specification.

OPEN IN VIEWER

Table 1.

Variable Key.

Variable	Measure and Data Source	Logic and Reference Research
Outcome Variables
Shareholders’ short-term reaction	The sum of abnormal stock returns to the stock of disengagement-announcing firms on each of the seven days in the event window of [−3,3]	CAR captures shareholders’ short-term reaction to the news that may materially affect the firm (Wiles et al. 2010).
	Source: CRSP, Compustat Global, Bloomberg (MSCI Country Indices)
	Method: Event study
Shareholders’ long-term reaction	The abnormal returns on the portfolio of disengagement-announcing firms during the 13-month event window of [0,12]	Calendar-time portfolio (CTP) return captures the long-term impact on the equity value of (a portfolio of) firms that decided to disengage with Russia (Feng, Morgan, and Rego 2017; Fotheringham and Wiles 2023; Homburg, Theel, and Hohenberg 2020; Kovalenko, Sorescu, and Houston 2022)
	Source: CRSP, Compustat Global, Bloomberg (MSCI Country Indices)
	Method: Calendar-time portfolio
Three Explanatory Variables (i.e., Variables of Theoretical Interest)
Market emphasis (in the announcement)	We count the terms shared in the focal announcement and the product-market dictionary. Next, we divide this number by the number of terms in the announcement text. Term in the product-market dictionary refers to n-grams and, thus, includes unigrams, bigrams, and so on.	A firm's disengagement announcement's market emphasis is the firm's emphasis on its product-market activities and stakeholders. Announcements with market emphasis may highlight the difficulty of obtaining supplies (e.g., Merck), packaging (e.g., PPG), customer service (e.g., Hapag-Lloyd), product development (PepsiCo), logistics (e.g., Mitsubishi Electric), and distribution (e.g., Twin Disc) (see Web Appendix WG for example announcements).
	See Web Appendix G (Table WG1) for examples of disengagement announcements with low and high market and social emphasis values.
	Sources for the dictionary of terms related to product-market activities (e.g., buying, selling) and stakeholders (i.e., customers, suppliers, and rivals): Armstrong et al. (2017); Bhandari et al. (2020); Bhattacharya, Misra, and Sardashti (2019); Feng, Patel, and Xiang (2020); Homburg, Theel, and Hohenberg (2020); Jedidi et al. (2021); Kashmiri, Nicol, and Arora (2017); Key and Keel (2020); Noble, Sinha, and Kumar (2002); Pride et al. (2017); Reed (2014); Saboo, Chakravarty, and Grewal (2016); Srivastava, Kashmiri, and Mahajan (2023); Wang, Gupta, and Grewal (2017); Zachary et al. (2011)
Social emphasis (in the announcement)	We count the terms shared in the focal announcement and the social emphasis dictionary. Next, we divide this number by the number of terms in the announcement text. Term in the social emphasis dictionary refers to n-grams and, thus, includes unigrams, bigrams, and so on.	A firm's disengagement announcement's social emphasis is the firm's emphasis on its societal activities and stakeholders. Announcements with social emphasis highlight the firm's societal activities and stakeholders, such as aiding employees and their families (e.g., Konica Minolta Group), condemning the environmental impact of the war (e.g., Kinross), and promoting a safe and secure society (e.g., Yokohama Rubber).
	See Web Appendix G (Table WG1) for examples of disengagement announcements with low and high market and social emphasis values.
	Sources for the dictionary of terms related to a firm's societal activities (e.g., aiding employee well-being, promoting a safe and secure community) and stakeholders (i.e., employees, environment, and community): Olsen, Sofka, and Grimpe (2016); Pencle and Mălăescu (2016)
Moderator
Delay (in announcing the disengagement)	The natural logarithm of 1 + the number of days between February 24, 2022, and the date when the focal announcement was made.	Later (vs. early) announcements may suggest that the firm has understood the benefits and costs of disengagement and crafted a well-thought-out disengagement plan minimizing the potential damage. Thus, we expect the delay to impact the shareholders’ reaction to the disengagement announcement positively.
	See Web Appendix G (Table WG2) for examples of disengagement announcements with short and long delays.
	Source: Nexis Uni, other news sources, press releases
Four Announcement-Specific Controls
Strength of disengagement	Five levels: digging in (Group 0), buying time (Group 1), scaling back (Group 2), suspension (Group 3), and withdrawal (Group 4)	The shareholders' reactions to disengagement become more negative when firms choose a stronger level of disengagement.
	In the Heckman selection regression, “digging in” is our proxy for the firm choosing not to disengage, whereas the other four levels indicate disengagement. In the second-stage regression, “Buying time” (i.e., the weakest level of disengagement) serves as our base level.
	Sources: Yale Institute, KSE Institute, The Good Lobby, and Progressive Shopper
Announcement sentiment	The number of positive terms in the announcement text minus the number of negative terms, divided by the total number of terms in the announcement text.	The shareholders' reactions to disengagement become less negative when the announcement's sentiment (net positivity) increases.
	Sources of dictionaries: Abrahamson and Park (1994); Henry (2008); Loughran and McDonald (2011)
	Software program: https://www.knime.com/ (Chung and Low 2022)
Upstream	We count the times the terms supply, vendor, parts, components, production, manufacturing, investment, divestment, and project appear in the focal announcement. Then, we divide this number by the total number of terms in the announcement text. Finally, we multiply this number by one hundred.	We conjecture that shareholders may react differently if the firm's disengagement concerns its upstream (i.e., not customer-facing) activities.
Downstream	We count the number of times the terms licensing, distribution, logistics, shipment, franchise, (alliance or business) partner, store, export, sale, selling, and (close) office appear in the focal announcement. Then, we divide this number by the total number of terms in the announcement text. Finally, we multiply this number by one hundred.	We conjecture that shareholders may react differently if the firm's disengagement concerns its downstream (i.e., customer-facing) activities.
13 Firm-Specific Controls (in the Year Before the Disengagement Announcement)
CSP rating	The focal firm's social responsibility performance (CSP) rating in the year before the announcement (2021 or 2022).	The announcing firm's prior CSP rating sets shareholders’ expectations that the firm may make decisions based on stakeholders’ concerns and not necessarily shareholders’ interests. Such expectations may be positively associated with their reaction.
	Source: Refinitiv ESG
Number of public statements on invasion	The number of public statements the firm made on the war before its disengagement announcement.	The higher the firm's number of statements on war, the lower the surprise in the firm's subsequent disengagement announcements and, thus, the less negative the shareholder reaction.
	Source: KSE Institute, firm's websites
Firm-specific risk environment	Natural logarithm of 1 + the number of incidents that involved the focal firm and related to (1) supply chain and marketing risks, (2) abuse of human rights, and (3) violation of national legislation.	The riskier the firm's informational environment, the more likely the firm is to announce disengagement (for the first-stage Heckman regression). Additionally, we expect the risk level of the firm's information environment to negatively impact the shareholders’ reaction (Wong and Zhang 2022).
	Source: RepRisk
Dependence on Russia	= 1 if the firm separately reported its revenue from the Russian market in its annual report in year t − 1 (where t is the year of announcement) or KSE Institute reported revenue from the Russian market for the firm, and 0 otherwise.	The more the firm's business depends on Russia, the more pressure it would receive to announce disengagement (for first-stage Heckman regression) (Basnet, Blomkvist, and Galariotis 2022). Shareholders may react more harshly if the firm depends (vs. does not depend) on Russia for sales revenue.
	Source: Mergent Online and KSE Institute
Profitabilitya	Ratio of net income (NI) to total assets (AT)	Profit boosts the firm's slack to cope with revenue loss due to disengagement from Russian business operations. Therefore, we expect profitability to suppress shareholders’ negative reactions (Fang, Palmatier, and Steenkamp 2008).
	Source: Compustat
Firm size	The natural logarithm of the number of employees (EMP)	Larger firms are less impacted by disengagement from a geopolitically uncertain market (Rao, Chandy, and Prabhu 2008; Vadakkepatt, Shankar, and Varadarajan 2021). Therefore, we expect the firm's size to increase the odds that it will announce disengagement. Shareholders’ reactions to disengagement become less negative as firm size increases.
	Source: Compustat and Mergent
Financial leveragea	Ratio of the firm's sum of long-term (DLTT) and short-term debt (DLC) to its market capitalization, where market capitalization is the product of stock price and the number of outstanding shares (CSHO × PRCC_C)	We expect a firm's financial leverage to boost its odds of disengagement (King, Loncan, and Khan 2021). Further, shareholders’ reactions to disengagement become more negative as the firm's financial leverage increases (Homburg, Vollmayr, and Hahn 2014).
	Source: Compustat
Sales growtha	The natural logarithm of the current year's sales (SALE)t divided by the last year's sales (SALE)t−1	On average, shareholders’ reactions to disengagement are less negative when firms have higher sales growth (Tuli, Bharadwaj, and Kohli 2010).
	Source: Compustat (for U.S. firms) and Worldscope (for non-U.S. firms)
Manufacturing (vs. service)	=1 if the firm operates in one of the two-digit SIC codes of 20–39, and 0 otherwise (service)	Manufacturing firms incur higher sunk costs than their service counterparts (Vadakkepatt, Shankar, and Varadarajan 2021) on announcing disengagement from Russia. Thus, shareholders’ reaction to disengagement becomes more negative for manufacturing firms compared to service counterparts.
USA	=1 if the firm's common stock is traded on a U.S. exchange (specifically, NYSE or Nasdaq) and 0 otherwise.	Because the United States has led the state sanctions against Russia, we expect U.S. firms to be more likely to disengage than their non-U.S. counterparts. Also, shareholders in U.S. firms react more unfavorably to disengagement than those in non-U.S. firms (Cheah, Chan, and Chieng 2007).
B2B	=1 if the firm operates in a business-to-business (B2B) industry (where we define industry by four-digit SIC codes) and 0 otherwise (Srinivasan, Lilien, and Sridhar 2011).	Two competing logics guide whether shareholders will be more punitive toward a B2B or B2C firm. On the one hand, because B2B (vs. B2C) firms have stronger and deeper ties with their suppliers and customers, they stand to lose more from their disengagement. Thus, we expect shareholders’ reactions to disengagement to become more negative toward B2B firms compared to their B2C counterparts. On the other hand, some consumer sectors (e.g., those sensitive to the prices of agricultural products/commodities) are expected to be impacted more severely during the war (Yudaruddin et al. 2023). Thus, we expect shareholders’ reactions to disengagement to become more negative for B2C firms than for B2B ones.
	Examples of B2B SIC codes: 2221 (Broadwoven Fabric mills), 2861 (Gum and Wood Chemicals), 2754 (Commercial Printing, Gravure), and 7311 (Advertising Agencies)
	Examples of B2C SIC codes: 2095 (Roasted Coffee), 2384 (Robes and Dressing Gowns), 7011 (Hotels and Motels), and 7241 (Barber Shops)
CAR to invasion[−10,0]	The sum of abnormal stock returns to the firm on each of the 11 days in the [−10,0] window, ending with the date when Russia invaded Ukraine.	We control for the invasion's initial impact on stock prices. At the onset of the invasion, the market may have penalized firms with greater exposure to the Russian market in the expectation of an eventual disengagement. On the one hand, the initial reaction may alleviate the negative abnormal return to the disengagement decision. On the other hand, disengagement by such firms may still be punished more severely due to a larger expected revenue loss (Basnet, Blomkvist, and Galariotis 2022).
	Source: CRSP, Compustat Global, and Bloomberg (MSCI Country Indices)
	Method: Event study
Precedence	The number of firms (in the sample) in the same 2-digit SIC industry that announced disengagement before the focal firm.	The prior announcement by industry peers (with the same two-digit SIC code as the focal firm) may attenuate the surprise in the focal firm's announcement and thus weaken the shareholder reaction (Lim, Tuli, and Dekimpe 2018; Warren and Sorescu 2017).
	Source: Yale Institute, The Good Lobby, and KSE Institute
Three Variables Specific to the Focal Firm's Country of Headquarters or Country of Stock Exchange
Sanction	= 1 if the firm announced its disengagement decision after its country of headquarters’ sanction decision and 0 otherwise.	The country's sanction of Russia may raise the odds of the firm announcing a disengagement. The sanction lowers shareholders’ surprise and, thus, their reactions.
	Source: Atlantic Council’s Russia Sanction database (https://www.atlanticcouncil.org/blogs/econographics/russia-sanctions-database/)
Country-level CSR sentiment	The average sentiment of all corporate social responsibility-related news involving firms in the country of the focal firm's stock exchange in the year before the disengagement announcement (2021 or 2022). The values range from −1 to +1, indicating negative and positive sentiments. Thus, higher values indicate positivity.	The average sentiment of CSR-related news in the country of the focal firm's stock exchange sets shareholders’ expectations. When the average sentiment of CSR-related news in the country of the focal firm's stock exchange is negative (positive), shareholders’ reactions to disengagement become less negative (more negative).
	Source: RavenPack Analytics
Country-level news volume	The natural logarithm of 1 + the sum of the number of invasion-related news in the country of headquarters from the month before the invasion until the day before the disengagement announcement.	News coverage of the invasion may set shareholders’ expectations such that a higher volume could dampen shareholders’ negative reaction to the focal disengagement announcement.
	Source: RavenPack Analytics
Excluded Variable for the First-Stage Selection (Heckman) Regression
Political disagreement	Consider firm i headquartered in country c. We measure political disagreement between c and Russia by Euclidean distance between c's estimated ideal point of votes (yea, abstain, nay) in the United Nations General Assembly and Russia's estimated ideal point of votes. We estimated these points using voting data from 2022 for firms that had reason to disengage with Russia but did not announce disengagement. For a firm that announced disengagement in year t (where t = 2022 or 2023), we used voting data from year t − 1. We obtained the ideal points and computed the Euclidean distance between the two points from Bailey, Strezhnev, and Voeten (2017). The higher the distance, the greater the divergence between c's and Russia's foreign policy.	Votes are mostly symbolic and indicate a country's national interest in global affairs (Stone 2004). Moreover, countries that vote together (differently) are likely to have a good (bad) relationship and share close (distant) views and understanding about global issues (Bertrand, Betschinger, and Settles 2016).
	Source: Voeten, Strezhnev, and Bailey (2009) and Bailey, Strezhnev, and Voeten (2017)

^a

Indicates winsorization at the 1^st and 99^th percentiles.

Notes: CRSP = Center for Research in Security Prices. We list our variables in the following order: (1) the two outcome variables that measure shareholders’ reaction to the announcement in the short-term of [−3,3] (daily) and the longer-term of [0,12] (monthly) event windows, (2) explanatory variables and moderator, (3) announcement-specific control variables, (4) announcing firm-specific control variables, (5) control variables specific to the announcing firm's country of headquarters or stock exchange, and (6) the one excluded variable in our first-stage Heckman selection regression. Zero in the short-term window refers to the date of announcement, whereas zero in the longer-term window refers to the month of announcement. Asterisk suffixed to names of financial ratio variables indicates winsorization at the 1^st and 99^th percentiles.

Event Study Method

We used the event study method (Brown and Warner 1985; McWilliams and Siegel 1997) to measure shareholders’ short-term reaction to a firm's announcement of disengagement from Russia. That is, the announcement is our event of interest. The event study method assumes that an unexpected event that may materially affect a firm can lead its shareholders to instantly update their expectations about the firm's future earnings (Van den Broek, Langley, and Hornig 2017). The updated expectations are reflected in a change in the price of the firm's common stock (Fama 1970).

Following Sorescu, Warren, and Ertekin’s (2017) advice, we used the market model ⁷ to estimate the expected return on the date when a firm announced its disengagement from Russia. Following event studies that sample firms from multiple exchanges (e.g., Klöckner, Schmidt, and Wagner 2022), we used Morgan Stanley Capital International's (MSCI's) country-specific index to measure the (value-weighted) market return for our sampled firms (MSCI 2022). ⁸ For example, consider a disengagement-announcing firm whose stock is traded on the Tokyo Stock Exchange. We used MSCI's index for Japan to compare the focal firm's stock return with the announcement relative to the Japanese market.

The market model assumes a linear relationship between a firm's expected return on date t, r_it, and the return for the market index m, r_mt (Brown and Warner 1985):

r it = α i + β i × r mt + ε it .

(1)

For each event i, we considered a 255-day estimation window that ends 42 trading days before the event/announcement date. Next, we estimated the coefficients in Equation 1 using the ordinary least squares (OLS) estimator. We computed the focal firm's abnormal return (AR) to the announcement as the actual return minus the expected return.

On the one hand, shareholders may become aware of the firm's decision a few days before the official announcement (McWilliams and Siegel 1997). Such leakage means that shareholder reaction may manifest a few days before the announcement. On the other hand, shareholders may continue to update their expectations a few days after the announcement. We account for this plausibility by computing the AR for each day in the 11 days centered on the date of announcement/event—that is, 5 days before the event (to account for leakage), the day of the event (day 0), and 5 days after the event (to account for the delay in shareholders updating their expectations).

Lastly, we compute the cumulative abnormal return (CAR) for 15 event windows in the above 11-day period: [−5,0], [−3,0], [−4,0], [−2,0], [−1,0], [0,1], [0,2], [0,3], [0,4], [0,5], [−1,1], [−2,2], [−3,3], [−4,4], and [−5,5]. For each window, CAR is the sum of ARs for all days in the window. As discussed subsequently, we chose the window that produces a strongly statistically significant CAR while mitigating the odds of confounding. This window happens to be [−3,3].

Regression for Explaining the Variation in CAR

The event study allows us to determine whether the mean CAR to firm announcements of disengagement from Russia is statistically different from zero. Next, we explain the variation in the CAR by the market emphasis (H₁), social emphasis (H₂), the interaction of market emphasis and social emphasis (H₃), and the moderation of the two main effects by delay (H₄ and H₅). Our specification includes other covariates that may impact the shareholder reaction.

CAR i , t ( − 3 , 3 ) = β 0 + β 1 Market orientation i + β 2 Social orientation i + β 3 ( Social orientation i × Market orientation i ) + β 4 ( Delay i × Market orientation i ) + β 5 ( Delay i × Social orientation i ) + β 6 Delay i + Γ × Z i ( t − 1 ) + Θ × Y c ( t − 1 ) + IMR i , t + ε i , t .

(2)

Subscripts i, t, and c index firm, year (2022 or 2023), and country of headquarters of firm i or country of the stock exchange where i's common stock is traded, respectively. CAR_i,t(−3,3) is the cumulative abnormal return for the [−3,3] window of the stock of firm i that made a disengagement announcement in year t. Z_i,t−1 is the vector of announcement-specific and announcing firm-specific control variables. Y_c,t−1 is the vector of control variables specific to the announcing firm's country (of headquarters or stock exchange). IMR_it is the inverse Mills ratio that is estimated from Equation 3, and ɛ_it is the normally distributed error term. We correct the standard errors to account for the estimated (and not observed) nature of the IMR. β₁–β₅ are regression parameters for the hypothesized variables, β₆ is the regression parameter for the moderator, and Γ and Θ are the vectors of regression parameters for the control variables.

Descriptive Statistics

Table WB3 provides descriptive statistics for variables in our specification. Table WB4 provides Pearson bivariate correlation coefficients. We checked the variance inflation factor (VIF) for each explanatory variable. The highest VIF is 4.13, and the average VIF is 1.95. Thus, multicollinearity is likely not a concern (Hair et al. 2010).

Model-Free Evidence

We use bar graphs and correlation coefficients to obtain model-free evidence for associations between our two explanatory variables—market emphasis and social emphasis—and CAR. The bar graph in Figure WB1 illustrates the model-free relation between market emphasis and CAR. The insight is that as we move from observations on the lowest quartile of market emphasis to the highest (that is, left to right on the x-axis), the mean CAR[−3,3] becomes less negative. Further, Table WB-2 reports that the Pearson correlation coefficient between market emphasis and CAR is .103, p < .10. So, the model-free evidence is consistent with our H₁. Figure WB2 shows that social emphasis does not vary in a specific direction as we move from left to right on the x-axis. Further, the correlation coefficient between social emphasis and CAR is .013, p > .10, suggesting that the data do not support our H₂.

Results from Cross-Sectional Analysis: Controlling for Self-Selection Bias

Firms that announce their disengagement may systematically differ from those that choose to stay silent. Because we sample firms that announced their disengagements, our sample may be beset with self-selection, potentially biasing our estimates. We use the two-stage Heckman (1979) method to control for the potential bias.

We identified the relevant set of firms from three sources: (1) Yale Institute's list, (2) KSE Institute’s list, and (3) Mergent Online database. Combining the names from these sources led us to 598 distinct firms that had exposure to Russia but did not disengage. After checking the availability of information on all our control variables, we ended up with 224 public firms that chose not to withdraw. Therefore, our first-stage regression uses 711 (487 + 224) observations as the sample.

The first-stage regression includes 11 variables (listed in Table WD1) that could impact the probability of a firm's disengagement from Russia. The first-stage specification also needs an additional variable (called the excluded variable) that influences the disengagement decision but not the shareholders’ reaction. Consider firm i headquartered in country c. This firm had reason to disengage with Russia but chose not to. Our excluded variable is the level of political disagreement between country c and Russia. Web Appendix D describes the data source and measurement of political disagreement. We used the probit model to estimate the following equation, with the DV coded as 1 if a firm disengaged and 0 otherwise.

Pr it ( Disengagement = 1 | X ) = Φ ( X ) ;

X = α 0 + α 1 Political disagreement i ( t − 1 ) + Γ × Z i ( t − 1 ) + Θ × Y c ( t − 1 ) + ω it .

(3)

Φ(·) is the cumulative standard normal distribution function, Z_i(t−1) and Y_i(t−1) are subsets of control variables (excluding announcement-specific controls) from Equation (2), and ω_it is the error term. The sample includes 487 firms that disengaged and 224 firms that did not disengage. We obtained the estimated IMR from Equation 3 to control for any systematic differences between the firms that announced disengagement and those that did not. We include the IMR in Equation 2 to test H₁–H₅. Estimates from the first-stage regression (Table WD1) show that political disagreement increases the firm's likelihood of disengaging with Russia—a finding consistent with intuition. The following variables are associated with higher odds of disengagement: riskiness of the firm's environment, dependence on the Russian market (for revenue), and sales growth. In contrast, financial leverage lowers the odds of disengagement. In addition, manufacturing firms are less likely to disengage than service (nonmanufacturing) firms. Lastly, the volume of news about the invasion/war is negatively associated with the firm's odds of disengaging. The chi-square for the likelihood ratio is 418.51, which is significant at p < .001, suggesting that the model strongly fits the data.

Cross-Sectional Regression

The CAR in the window [−3, 3] is our dependent variable (DV). We estimate Equation 2 to test our hypotheses. We mean-center the continuous explanatory and moderator variables for ease of interpretation of interaction terms (Wies, Bleier, and Edeling 2023). Table 2 reports estimates from five regressions.

OPEN IN VIEWER

Table 2.

Estimates from Cross-Sectional Regression (Heckman Second Stage).

DV	Model I	Model II	Model III	Model IV	Model V
	CAR[−3,+3]	CAR[−3,+3]	CAR[−3,+3]	CAR[−3,+3]	CAR[−3,+3]
Market emphasis	.131 [.012] (.052)	.145 [.005] (.052)	.137 [.009] (.052)	.134 [.010] (.052)	.152 [.004] (.052)
Social emphasis	−.038 [.341] (.039)	−.027 [.487] (.039)	−.039 [.322] (.039)	−.041 [.298] (.039)	−.030 [.449] (.039)
Social emphasis × Market emphasis		−.940 [.027](.426)			−.919 [.031](.425)
Delay	.009 [.002](.003)	.010 [.001](.003)	.009 [.003](.003)	.008 [.005](.003)	.009 [.003](.003)
Delay × Market emphasis			−.065 [.048](.033)		−.060 [.090](.035)
Delay × Social emphasis				−.034 [.232](.028)	−.010 [.753](.031)
Scaling back	−.018 [.125](.012)	−.017 [.150](.012)	−.017 [.135](.012)	−.018 [.128](.012)	−.016 [.162](.012)
Suspension	−.009 [.340](.010)	−.009 [.338](.010)	−.010 [.322](.010)	−.009 [.340](.010)	−.010 [.323](.010)
Withdrawal	−.018 [.064](.010)	−.018 [.064](.009)	−.017 [.080](.010)	−.016 [.085](.010)	−.016 [.086](.009)
CSP rating	−.013 [.442](.017)	−.012 [.480](.017)	−.015 [.366](.017)	−.013 [.421](.017)	−.014 [.399](.017)
Announcement sentiment	−.073 [.276](.067)	−.101 [.138](.068)	−.067 [.322](.067)	−.068 [.316](.068)	−.093 [.174](.068)
Upstream	−.001 [.588](.001)	−.001 [.344](.001)	−.001 [.080](.001)	−.001 [.587](.001)	−.001 [.397](.001)
Downstream	−.001 [.298](.001)	−.001 [.589](.001)	−.001 [.393](.001)	−.001 [.296](.001)	−.000 [.703](.001)
No. of public statements on war	−.005 [.392](.006)	−.006 [.371](.006)	−.005 [.390](.006)	−.005 [.411](.006)	−.006 [.375](.006)
Firm-specific risk environment	−.001 [.718](.003)	−.001 [.690](.003)	−.001 [.793](.003)	−.001 [.828](.003)	−.001 [.792](.003)
Dependence on Russia	.007 [.504](.010)	.006 [.571](.010)	.007 [.447](.010)	.007 [.485](.010)	.006 [.508](.010)
Profitability	.002 [.960](.035)	.005 [.896](.035)	.001 [.969](.035)	.002 [.959](.035)	.004 [.905](.035)
Firm size	.003 [.259](.003)	.003 [.220](.003)	.003 [.234](.003)	.003 [.278](.003)	.003 [.205](.003)
Financial leverage	−.004 [.421](.005)	−.003 [.446](.004)	−.003 [.461](.004)	−.004 [.411](.005)	−.003 [.481](.004)
Sales growth	−.002 [.889](.015)	−.002 [.896](.015)	−.005 [.754](.015)	−.003 [.836](.015)	−.005 [.756](.015)
Manufacturing	−.012 [.068](.007)	−.012 [.070](.007)	−.012 [.074](.007)	−.012 [.071](.007)	−.012 [.076](.007)
USA	−.010 [.349](.010)	−.009 [.370](.010)	−.008 [.452](.010)	−.010 [.315](.010)	−.008 [.459](.010)
B2B	.019 [.001](.006)	.020 [.001](.006)	.019 [.001](.006)	.020 [.001](.006)	.020 [.001](.006)
CAR to invasion[−10,0]	−.036 [.341](.037)	−.033 [.382](.037)	−.036 [.328](.037)	−.035 [.349](.037)	−.033 [.371](.037)
Precedence	−.000 [.893](.000)	−.000 [.755](.000)	.000 [.998](.000)	−.000 [.879](.000)	−.000 [.852](.000)
Sanction	.020 [.105](.012)	.021 [.086](.012)	.023 [.060](.012)	.021 [.092](.012)	.024 [.049](.012)
Country-specific CSR sentiment	−1.775 [.265](1.591)	−1.833 [.247](1.584)	−1.733 [.274](1.585)	−1.653 [.299](1.592)	−1.758 [.266](1.581)
Country-specific news volume	.000 [.919](.002)	.000 [.984](.002)	−.000 [.944](.002)	.000 [.971](.002)	−.000 [.876](.002)
Intercept	.451 [.283](.420)	.465 [.266](.418)	.437 [.295](.418)	.418 [.319](.420)	.443 [.288](.417)
IMR	.009 [.526](.015)	.010 [.484](.015)	.010 [.480](.015)	.010 [.481](.015)	.012 [.433](.015)
Wald chi-square	68.49 [<.0001]	73.98 [<.0001]	72.95 [<.0001]	70.08 [<.0001]	78.51 [<.0001]
Adj. R-squared	.074	.081	.079	.075	.085
No. of observations	487	487	487	487	487

Notes: p-values in brackets; two-tailed tests of significance. Standard errors in parentheses.

Model I is the main-effects-only model. Other than the control variables and the IMR from the first-stage regression, Model I includes the explanatory variables for the two linguistic frames of the disengagement announcement: that is, market emphasis and social emphasis, corresponding to our H₁ and H₂, respectively. Model II is Model I plus social emphasis × market emphasis (for our H₃). Model III is Model I plus delay × market emphasis (for our H₄). Model IV is Model I plus delay × social emphasis (for our H₅). Last, Model V—our full specification—is Model I plus all three interaction terms.

We use estimates from Model V to report the associations of market emphasis (i.e., H₁, which expects a positive β₁), social emphasis (i.e., H₂, negative β₂), market emphasis × social emphasis (H₃, negative β₃), delay × market emphasis (i.e., H₄, negative β₄), and delay × social emphasis (i.e., H₅, negative β₅).

Market emphasis’s estimated coefficient is positive and significant (Model V: β₁ = .152, p = .004). The coefficient means that the higher the announcement's focus on the product market, the less punitive the shareholders’ reaction, thus supporting our H₁. For an average firm in our sample, an announcement with a 23.4% market emphasis ⁹ nullifies the shareholders’ punitive reaction. Social emphasis's coefficient is negative but statistically insignificant (Model V: β₂ = −.030, p = .449). Thus, we do not find statistical support for our H₂.

Social emphasis × market emphasis’ coefficient is negative and significant (Model V: β₃= −.919, p = .031), supporting our H₃ that the firm's social emphasis in its disengagement announcement weakens the positive association between the announcement's market emphasis on shareholder reactions. Figure 1, Panel A, depicts the interaction effect graphically. The red dashed line reports the slope of market emphasis on shareholder penalty for announcements with the median value of social emphasis. The slope of market emphasis on CARs is more positive for announcements with lower social emphasis scores (the blue line with the diamond symbol) than for announcements with higher social emphasis scores (the green line with the triangle symbol).

OPEN IN VIEWER

Figure 1.

Association Between Market Emphasis and Shareholder Reactions at Different Percentiles of Social Emphasis and Delay.

Delay × market emphasis’s coefficient is negative and marginally significant (Model V: β₄ = −.060, p = .090), supporting our H₄ that delay in announcing the disengagement weakens the positive association between market emphasis and shareholder reactions. Figure 1, Panel B, depicts the moderation effect. It shows that high market emphasis in the disengagement announcement mostly helps firms with a delay at or below the median value compared to firms with a high delay.

The delay × social emphasis’ coefficient is negative but statistically insignificant (β₅= −.010, p = .753). Thus, we do not find statistical support for our H₅ that the firm's delay in announcing the disengagement strengthens the negative association between social emphasis and shareholder penalty.

Robustness to the Inclusion of Confounded Announcements and Announcements with Missing Data on Covariates

The results of our event study are robust to the inclusion of (1) 141 confounded announcements (announcement sample = 487 + 141 = 628), (2) 185 announcements for which Compustat did not provide data for the regressors (sample = 487 + 185 = 672), and (3) both (sample = 487 + 141 + 672 = 813). Tables WE1, WE2, and WE3 in Web Appendix E present the results. Because our focus is on the cross-sectional variation in CAR, we next report the robustness of our regression results.

Robustness to Alternate Event Windows

We use CARs for three alternate windows to ensure that our results are robust to the choice of window (e.g., Homburg, Vollmayr, and Hahn 2014). Specifically, we use the windows of [−2,2], [−4,4], and [−5,5] to ensure that using the windows shorter or longer than our main window of [−3,3] yields similar results. The results (in Table WE4) with these alternate measures of the DV are largely consistent with those reported in Model V of Table 2.

Robustness to an Alternate Measure of Shareholders’ Short-Term Reaction

A firm's stock price reflects the sum of the value of its operating business and the value of its nonoperating assets, minus the value of its debt. Skiera, Bayer, and Schöler (2017) reasoned that because marketing events impact the value of the firm's operating business only, a more valid measure of shareholders’ reaction is the CAR divided by the effect on the firm's leverage. The new measure reflects shareholders’ valuation of the impact of the firm's announcement on its operating business. We examine the robustness of our results to Skiera, Bayer, and Schöler’s measure. Our findings with this alternate DV (Column I of Table WE6) are qualitatively similar to those reported in Model V of Table 2.

Robustness to Two Alternate Methods of Correcting for Potential Self-Selection Bias

Our main analysis uses the Heckman two-stage method to correct for potential bias caused by the self-selection of observations in our sample. We next test whether our results are robust to two alternate methods of correcting for this bias.

The first method uses a control function and considers four different levels of decision (as opposed to a binary decision): (1) buying time, (2) scaling back, (3) suspension, and (4) withdrawal. We estimate a multinomial logistic regression in the first stage (with the same regressors we use in the Heckman first-stage regression). Table WE5 reports the estimates from the multinomial logit regression in the first stage (Column I to Column IV for four levels of disengagement, with the “digging in” category as the base). The second-stage specification includes the residuals, the indicator variables for the levels of disengagement, and covariates. The estimates (Column II of Table WE6) are largely consistent with those reported in Model V of Table 2.

The second approach follows Homburg, Vollmayr, and Hahn (2014) and uses the multinomial logistic selection model proposed by Dubin and McFadden (1984). This method suggests that because the levels of disengagement are mutually exclusive, we do not need to allow for correlated choices, and a multinomial logistic regression suffices. The first stage of this method uses the same set of regressors as the Heckman regression. We use the predicted conditional probabilities for each level of disengagement given the conditional probabilities of all other choices (from the first stage) to calculate the selection correction terms (i.e., the multinomial logit equivalents to the Heckman's IMR) λ₁ for scaling back, λ₂ for suspension, and λ₃ for withdrawal (Bourguignon, Fournier, and Gurgand 2007; Dubin and McFadden 1984). Specifically, for each firm i, λ ij = J ⋅ ln ( p ij ) + ∑ k ≠ j p ik . ln ( p ik ) / ( 1 − p ik ) , where p_ij is the predicted probability that the firm chooses the level of disengagement j, and J is the total number of disengagement levels. Next, our second-stage regression includes the three lambdas to control for the selection bias (Li and Prabhala 2007). The estimates from this method (Column III of Table WE6) are largely consistent with those reported in Model V of Table 2.

Implications for Theory

Our research contributes to the literature streams on impression management and firm disengagement from a geopolitically uncertain market.

First, management academics have theorized that managers can use linguistic frames in their disclosures and announcements, thus managing stakeholders’ impressions of a firm and receiving favorable outcomes (Falchetti, Cattani, and Ferriani 2022; Fiss and Zajac 2006; Guo, Sengul, and Yu 2021; Rhee and Fiss 2014). In parallel, marketing academics have theorized the performance benefits of a firm's market emphasis (e.g., Narver, Slater, and MacLachlan 2004; Saboo and Grewal 2013; Srivastava, Kashmiri, and Mahajan 2023) and its social counterpart (Fritz 1996; Gabler, Landers, and Richey 2020). We intersect management and marketing research to propose that a firm can use market and social emphases in its announcements of disengagement from a geopolitically uncertain market. Empirical evidence suggests the effectiveness of the market emphasis frame and the ineffectiveness of the social emphasis frame in influencing the firm's shareholders in the short term—but not the long term. This demonstration suggests that linguistic framing is an impression management tactic. Therefore, we also contribute to the literature on how a firm can manage its impressions on shareholders (Elsbach, Sutton, and Principe 1998; Graffin, Haleblian, and Kiley 2016).

Second, we contribute to the nascent literature on firm disengagement from a geopolitically uncertain market (Luo 2024; Sethuram and Gaur 2024). While the evidence on a firm's investment and engagement in a new market is voluminous, research on disengagement, disinvestment, and divestment is marginal (Sethuram and Gaur 2024). As more firms decide whether—and if yes, how strongly—to disengage from an uncertain market (Luo 2024), their decisions may signal economic prudence or social pressure. Prior research in this domain has considered the disengagement announcing firm's shareholders’ reactions to the announcement. Interestingly, the evidence is mixed, with some studies reporting a penalty (e.g., Wright and Ferris 1997), others documenting a reward (e.g., Posnikoff 1997), and the remaining finding an insignificant reaction (e.g., McWilliams and Siegel 1997). We contribute by reasoning that what matters is not the disengagement announcement per se but how the firm linguistically frames the announcement (Rhee and Fiss 2014). Managers could announce disengagement to prevent future loss of operating in an uncertain market (i.e., market emphasis) and/or to demonstrate their social imperative (i.e., social emphasis). Shareholders cannot observe the managerial motives and thus screen the announcement text to discern the latent motives. We contribute to the literature by theorizing that an announcement's linguistic framing can influence shareholder reaction to the announcement.

Implications for Practice

Managers often decide whether to exit a geopolitically sensitive market. Exiting the market will cause an immediate loss of sales revenue and, thus, a decline in the firm's stock price in the short term. However, such an exit may remove the firm from public denunciations and boycotts, thus saving the firm the resources it would expend to manage the issue. Further, a timely exit could help the firm win customers’ and employees’ support and avert a public relations crisis (Ganesan and Mallapragada 2025). Thus, managers face a dilemma. Prior academic evidence is mixed on whether a disengagement announcement elicits a shareholder penalty, reward, or no reaction from the firm's shareholders in the short term. Further, there is no evidence of the impact of disengagement announcements on shareholder value in the longer term. Thus, managers want to know whether—and, if yes, in what direction and to what magnitude—their decisions would impact the firm's shareholder value.

First, we contribute by informing managers about the average impact on stock returns in the short and longer term. Specifically, we report that, on average and in the short term, shareholders punish a firm that announces disengagement from a geopolitically uncertain market and that the stronger the disengagement, the greater the punishment. Importantly, however, the reaction reverses in the longer term, suggesting that shareholders may realize that the disengagement's benefits (accrued via stronger customer and employee support) dominate its costs (incurred via a drop in sales). Managers would benefit from knowing the asymmetry in the shareholders’ short-term and longer-term reactions.

Second, managers may anticipate a penalty in the short term because an exit from a market means an immediate loss of sales revenue and thus hurts the firm's profit and cash flow. Therefore, they would next like to know how to limit the short-term penalty. Simply put, we advise managers to mention product-market activities and stakeholders in their announcement's linguistic framing. Such framing would signal to shareholders that the managers are disengaging because conducting business in an uncertain market (e.g., Russia) is less viable and thus lower their punitive reactions. Further, managers should temper social emphasis in the announcement because shareholders may interpret dual emphasis announcements as obfuscatory rather than informative (Bushee and Huang 2024). Last, we advise managers to be proactive in announcing disengagement. While the delay in announcing disengagement positively impacts shareholder penalty on average, it backfires when combined with an announcement with a high market emphasis or high social emphasis. The intuition is that an announcement with market emphasis but delayed may send ambivalent signals to shareholders on the managers’ underlying motives and may thus weaken the market emphasis’ positive effect on shareholder penalty.

Limitations and Future Research

Our research can be extended in six ways. First, managers may disengage because they believe that disengagement aligns with nonshareholding stakeholders’ interests, and that shareholder penalty would dissipate in the longer term. While we offer evidence supporting the former, future research may consider testing whether other stakeholders—such as customers, suppliers, employees, and politicians—reward the firm for disengaging from a geopolitically uncertain market and meeting their expectations. This reward is particularly plausible for private firms or firms that target a niche customer segment, characterized by unique political values.

Second, some firms may have earned the reputation of adopting the “cancel culture.” How might such firms’ past behaviors of cancel culture impact shareholder response? The logic of expectancy suggests that shareholders may account for such compulsive disengagement and react weakly as a firm moves from one issue to another. An alternative view holds that shareholders may strengthen their penalty to discipline managers. Empiricists may consider testing which view holds.

Third, we assume that disengagement drives shareholder reactions. However, the reverse might hold as well. That is, how shareholders react to a reference firm's disengagement could shape the focal firm's future disengagement. Academics may consider theorizing and measuring shareholder reaction as a governance mechanism (Connelly et al. 2010).

Fourth, our dependent variable measures reactions from all classes of shareholders. However, shareholders are heterogeneous in their temporal horizons (DesJardine, Shi, and Sun 2022; Guo and Yu 2024) and, consequently, how much they emphasize a firm's social responsibility in their investment decisions (Cheah, Chan, and Chieng 2007). Future research can consider examining this variation.

Fifth, we report that a firm's linguistic framing of an announcement sways shareholders. Future research can examine whether such framing can be employed as “cheap talk” or is actually “costly.” This direction is particularly relevant because we measure reactions to announcements, not actual disengagement. Future research can investigate shareholder reactions to firms that announce leaving war-torn countries.

Sixth, we focus on firms’ announcements of (and not actual) disengagements, shareholders’ reactions to the announcements, and managers’ linguistic framing of announcements to manage impressions on shareholders. Future research could focus on actual disengagements and their performance implications.

In summary, we believe we offer theoretically novel and managerially actionable insights into the emerging phenomenon of corporate disengagement from a geopolitically uncertain market, as manifested in the context of Russia.