Exit or stay: A critical incident analysis of decision-making in conflict-torn countries

Deploying small world representation logic, we examine the context-specific factors that inform managerial decision-making in conflict-torn countries. Drawing on insights from thirty-one managers, we spotlight nine higher-order heuristics that commonly inform MNEs ’ mental representations and their managers ’ decision to exit or stay. These heuristics were identified by categorising the commonalities arising from our respondents ’ ac-counts on what information they search for ( discovery heuristics ) and how it was evaluated ( evaluation heuristics ). We discovered that information accessibility, conditioned by firms ’ in-country experiences, is paramount. Furthermore, since employees are strategy shapers, they can undermine the resilience-enhancing benefits of operational flexibility in conflict-torn countries.


Introduction
Events in Afghanistan, Libya, Nigeria and Ukraine are indicative of the hostile environments MNEs can face when confronted with violent conflict. Defined as "organised and sustained use of physical force that results in injury or death to persons and/or damage or destruction to property" (Oetzel & Getz, 2012, p.168), violent conflict can distort the business landscape and push MNEs and their managers to rapidly interpret the situation and respond (Cavusgil et al., 2020;Chen, 2017). Although infrequent and sporadic, episodes of violent conflict are becoming a common occurrence in the international business environment (Dai, Eden & Beamish, 2022;Gonchar & Greve, 2022). When faced with such unpredictable events, MNEs and their managers make different choices: some stay, while others exit. Heterogenous behaviours have been documented in, for example, Colombia, the Democratic Republic of Congo, Iraq, Libya, Myanmar, Nigeria, Papua New Guinea and Sudan (Darendeli & Hill, 2016, Guáqueta, 2007Kolk & Lenfant, 2012;Meyer & Thein, 2014;Patey, 2006;Rosenau et al., 2009).
Based on this backdrop, the relationship between violent conflict and business strategies has gained traction in academic discourse. Although this work contributes to opening the black box of heterogenous responses (Dai, Eden & Beamish, 2013;Lee & Chung, 2022), we argue that disconnections are emerging between theory-led conceptualisations and the decision-making realities faced by affected managers. With some exceptions (e.g., Al Khattab, Anchor & Davies, 2007;Belhoste & Nivet, 2021;Darendeli & Hill, 2016;Jamali & Mirshak, 2010;Oetzel & Getz, 2012), researchers have investigated this complex phenomenon from an organisational level. By sidestepping managers, a one-sided narrative is taking hold that focuses on higher-level actions and not the lower-level vehicles in which these actions are nested (Buckley et al., 2016). Consequently, research can overstretch the importance of theorised decision-influencing factors that may not accurately capture the factors given prominence by managers. We thus argue that a spotlight should be placed on the managers that make the strategic decisions. This is an important pivot, with contemporary work acknowledging that exit decisions are influenced by managers' perceptions, rather than rational, deterministic economic calculations (Belhoste & Nivet, 2021;Lee & Chung, 2022;Liu et al., 2022;. In response to this detachment, we pose the following research question: what context-specific factors are commonly factored into the decision-making calculi of managers when confronted with violent conflict? To explore this question, we use insights from behavioural strategy research, notably work on managerial mental representations. Growing in popularity, mental representation frameworks, such as small world representations, indicate that in complex and imperfect situations managers will implicitly or explicitly simplify the dimensionality of the environment (Maitland & Sammartino, 2015;Menon, 2018). Put simply, afflicted managers will develop a representation of the situation that considers only their perceived salient factors to inform their decisions to exit or stay. With violent conflict being a noteworthy example of imperfect situations, extending small world representation logic into our context will allow us to sift through theorised decision-influencing factors. Therefore, bridging the gap between theory-led conceptualisations and decision-making realities.
In practical terms, we first map out the current state of play within relevant research. To facilitate this review, we use Patey's (2006) typology to categorise the theorised factors. This typology has received attention in earlier work (Kolk & Lenfant, 2013) and has an alignment with how scholars have recently categorised decision-influencing factors, notably Belhoste and Nivet (2021). This mapping exercise provides an overview of the multitude of decision-influencing factors that may enter a manager's simplified representation. Second, we use the critical decision method to collect and map primary data from thirty-one managers about their decision-making experiences. Each manager was responsible for defining their conflict events, location and scope. Most managers recounted their experiences with subnational conflicts (58% of incidences) while the remaining 42% were nationwide events. Third, we apply the discovery and evaluation heuristics steps in small world representations to evaluate the information searching, collection and analysis stages in our managers' construction of their mental representations.
Our study contributes to the extant literature in two ways. First, we spotlight the main decision-influencing factors that determine firm-level strategic responses in conflict-torn countries by specifically responding to calls for research on how managers frame violent conflict events (Jamali & Mirshak, 2010;Oh & Oetzel, 2011, 2017. Accordingly, our study acknowledges the criticisms (e.g., Buckley et al., 2016;Powell et al., 2011) that research has side-lined the role of managerial cognition when trying to explain firm-level outcomes. Drawing on comments from senior decision-makers in conflict-torn countries, we spotlight this underexamined group. Focusing on what context-specific factors were commonly entered into their decision-making calculi, we pinpoint influential context-specific factors. Hence, we consolidate the current literature and begin to resolve the tension between theory-led conceptualisations and the decision-making realities of affected managers. Second, our research adds to the growing literature linking managerial mental representations to MNE higher-level strategies. By drawing on Maitland and Sammartino's (2015) small world representations, we display how focusing on lower-level vehicles can broaden our understanding of firm-level strategies. By emphasising the stages of information searching, collection, and analysis in the development of mental representations, our analytical approach offers insight into how managers understand and mitigate the challenges arising from doing business in environments characterised by violent conflict.

Violent conflict, MNEs and behavioural strategy insights
Given that violent conflict may include "war, revolution, rebellion, insurgency, and sustained campaigns of violence or terrorism" (Oetzel & Getz, 2012, p.168), the presence of violent conflict can significantly alter the international business environment (Bader, Schuster & Dickmann, 2015). In response, researchers have sought to determine how violent conflict affects the firm, and which context-specific factors determine the management teams' chosen response. Notable factors, for instance, are related to spatiotemporal dynamics of combat frontiers (Dai, Eden & Beamish, 2013Witte et al., 2017, a firm's resilience endowments and at-risk resources (Dai, Eden & Beamish, 2017;Darendeli et al., 2020;Gonchar & Greve, 2022;Lee & Chung, 2022), proximate and distant stakeholders (Meyer & Thein, 2014), and the broader temporal dynamics in conflict-torn countries Xiaopeng & Pheng, 2013). This growing body of work tells us that there is an abundance of context-specific factors that managers may take into consideration when deciding to exit or stay.
In response, recent arguments have centred on the premise that in non-routine, high-stake situations, managers may not consider all factors associated with the situation. Instead, they will simplify the situation (Yasuda & Kotabe, 2021) based on their "subjective perceptions with some factors given prominence and others filtered out" (Czinkota et al., 2010, p.835). This may be voluntary or down to short-sightedness (Belhoste & Nivet, 2021) whereby dangers are misinterpreted, or not fully understood, due to the inherent wrestle between uncertainty, complexity, time pressures and accountability (Shortland et al., 2020). Based on this narrative, the emergence of behavioural strategy logic in the research domain is not surprising. Contemporary work, for example, is turning to behavioural strategy theories to explore how CEO self-interests can direct MNE strategic choices when exposed to political risk and uncertainty (Benischke et al., 2022). Additionally, Liu et al (2022) draw on construal-level theory to denote that executives' social distance from terrorist attacks will affect whether a subsidiary will divest operations in the afflicted country. Finally, Maitland and Sammartino (2015) employ heuristics and small world representations to explain how decision-makers ponder, interpret and respond to politically hazardous environments. Collectively, there is a consensus that managerial cognitive processes and actions will determine the responses at the higher level (Buckley et al., 2016) -a clear indication that individuals matter when it comes to deciding between exiting or staying.
Confronted with explaining such complex and ill-structured phenomena, behavioural strategists offer the notion of heuristics, which have been applied in numerous ways in international business research (see Guercini & Milanesi, 2020, for a systematic review). Decision heuristics have been implemented in some contexts to explore the "procedures or rules of thumb that decision makers rely on to simplify complex problems and/or information processing tasks" (Guercini & Milanesi, 2022, p.2). Unlike less complex and defined situations where managers can consider all context-specific information and take time to reach an optimum decision, situations that threaten business continuity force managers to quickly scan the environment, process information and respond (Sarkar & Osiyevskyy, 2018). In these situations, with managers often being cognitively limited (Andrews et al., 2022;Menon, 2018;Shortland et al., 2020), heuristic logic suggests that managers will build simplified mental representations. These mental representations act to reduce the dimensionality of 'grand world' problems (Csaszar & Ostler, 2020) into more manageable 'small world' problems (Levinthal, 2011) that often entail a tug-of-war between optimising and satisficing (Dai, Eden & Beamish, 2022). This reductionist process begins with environmental scanning, information gathering and processing (Csaszar & Ostler, 2020;Nadkarni & Barr, 2008). At this stage, managerial preferences and decision heuristics influence what information is gathered, how the information is interpreted, and what informational cues enter the decision-making calculi (Artinger et al., 2015;Buckley et al., 2016Buckley et al., , 2018. Following this process, managers reduce the scale and scope of their mental representations, with only the most salient factors being considered (Miller & Lin, 2015). With a more limited configuration of factors to consider, managers will then evaluate the situation they face and decide how to respond (Csaszar & Levinthal, 2016). Given the structured lens provided by heuristics and mental representations, we turn to small world representation models to understand the filtering processes managers undertake when confronted with violent conflict.

Small world representations and violent conflict environments
To capture how managers construct simplified mental representations, our study draws on the work of Maitland and Sammartino (2015). This model breaks the heuristics process of small world representations into two phases: discovery and evaluation. We apply both to managers' interpretation, understanding and response to violent conflict events. Managers will first use their discovery heuristics to decide on what information to look for and from where. Representing a decision rule framed by cognitive architectures, learned procedures and rules of thumb (Gavetti, Levinthal & Rivkin, 2005;Levinthal, 2011), discovery heuristics guide managers in their attempts to quickly understand the unfolding events. From the outset, managers will want to understand how the violent conflict events will spatiotemporally evolve, how actors' possible defensive and offensive reactions may prolong the conflict, how these events could impact the viability of safeguarding at-risk resources and assets, and how this may determine their capacity to continue business operations.
Guided by these questions, the collection of information acts as individual building blocks to develop higher-order heuristics that capture (context) specific elements of the problem environment (Maitland & Sammartino, 2015). These higher-order heuristics provide afflicted managers with an initial sketch of the problem environment, which is then evaluated to make informed decisions. Under analogical reasoning (Gavetti & Levinthal, 2000;Miller & Lin, 2015), managers are not expected to address the discovery process uniformly. Instead, experienced managers who possess decision-making experience with similar problems in the past (Csaszar & Levinthal, 2016) will call upon their learned wisdom to make quick inferences about the current problem. These inferences will assist experienced managers in making timely decisions about search, stopping and decision rules. Such learned capabilities will better position affected managers in their command and control of the information searching and gathering task. Equally, experience can also shape a manager's ordering of information and possible decisions, which can both produce or preclude strategic choices (Dai, Eden & Beamish, 2022).
Given that the eruption of violent conflict can be sudden and unexpected, the timeframe of discovery heuristics is often limited with a high degree of urgency (Sarkar & Osiyevskky, 2018). Consequently, managers may be less vigilant (Schoemaker & Day, 2021), suffer from panic (Dai, Eden & Beamish, 2017), and decide to search for less information or stop searching prematurely. Equally, as the situation is fast-paced and rapidly changing, information can be highly incomplete, non-definitive and ambiguous (Feduzi et al., 2022). Thereby placing restrictions on the quantity and quality of available information. Further contributing to the fuzziness of discovery heuristics is the fact that managers may possess informational disadvantages (Buckley et al., 2016;McDougal, 2010) due to a lack of in-country relationships and experience (Henisz & Delios, 2004). Based on this backdrop, the discovery stage may be characterised by fast-and-frugal heuristics (Artinger et al., 2015;Luan, Reb & Gigerenzer, 2019), whereby managers are forced to frame the situation with relatively little information and within a short timeframe. These conditions will impact the level of detail in a manager's initial problem sketch and the subsequent richness of their small world representation.
After a manager has collected their chosen, or available, information, they will enter the second phase that involves evaluation heuristics. At this stage, managers will interpret the information and build a small world representation. This involves managers processing and ordering the information, estimating the probability that violent conflict events will materially impact the firm, predicting the magnitude of the impact, and trying to determine whether conflict events are likely to infringe on business continuation prospects. It is expected during this stage that collected information will be filtered, with some higher-order heuristics (i.e., context-specific factors) given prominence and others filtered out (Czinkota et al., 2010;Yasuda & Kotabe, 2021). This filtering process will spotlight certain information points, which will then shape the development of managerial small world representations (Maitland & Sammartino, 2015). Inexperienced managers facing the decision-making problem for the first time may consider too much or inadequate information, which can lead to inaccurate impressions of the unfolding events (Ramos & Ashby, 2017). Despite potential complications in the evaluation stage, managers will still reach a 'make do' (Menon, 2018) scenario with their small world representation, which, in turn, will inform their assessment of strategic options, and the final decision made by the firm.
Collectively, the phases of discovery and evaluation heuristics can offer insights into how managers tackle incidences of violent conflict. Therefore, the small world representation perspective (Maitland and Sammartino, 2015) can help highlight what context-specific factors are commonly entered into the decision-making calculi of affected managers. By comparing the factors considered in these small world representations, we can highlight the instrumental decision-influencing factors from existing research.

Violent conflict and mnes: unbundling the multiplicity of decision-influencing factors
The vibrant work surrounding MNEs and their strategic behaviours in conflict-torn countries indicates that strategic options and choices can be influenced by a multitude of context-specific factors. To categorise these factors, we use Patey's (2006) three-level typology to provide an overview of the current state of inquiry (see Table 1). As indicated earlier, each of these factors may enter the small world representations of afflicted managers, dependent on the discovery and evaluation heuristics that these decision-makers adopt.

Firm-specific considerations
Scholars have long argued that MNEs' attributes can shape their organisational preparedness (Czinkota et al., 2010), resilience and coping capacities (Dai, Eden & Beamish, 2017) when confronted with violent conflict. Possessing advantageous attributes can sway strategic responses, with such attributes typically being a product of country and context-specific experiences (Oetzel & Oh, 2014;. Research, for example, has found that prior experience with high-intensity, discontinuous risks such as violent conflict, can provide indispensable 'know-how', non-market effectiveness (Oetzel, Getz & Ladek, 2007;Schnyder & Sallai, 2020), and superior firm capabilities (Buckley et al., 2020;Jiménez & Lupton, 2021). Experience can also alter the dominant logic for risk avoidance (Oh, Shin & Oetzel, 2017. The main argument here is grounded on the premise that experienced firms/managers gain 'informational advantages' (Buckley et al., 2016) that can enable them to better grasp and interpret the context (Delios & Henisz, 2003;Henisz & Delios, 2004;Hodgkinson & Clarke, 2007). This, in turn, facilitates quick evaluations of the unfolding situation. Contrariwise, inexperienced managers may lack access to information or collect inaccurate information, which can lead to overestimating or underestimating conflict events. Such obstacles can lead to premature or unwarranted risky decisions. As such, experience is a major factor in informing MNEs' decision-making in conflict-torn countries (Dai, Eden & Beamish, 2022) via information access and accuracy.
Another important MNE attribute relates to resource configurations and structural capabilities (McDougal, 2010;Oetzel, 2005). When conflict unfolds, it is expected that managers take stock of their firm's activities to assess vulnerabilities and resilience endowments. The resource bundling perspective and real options theory have previously been employed to ascertain the role that resources (Dai, Eden & Beamish, 2017) and investment compositions (Burger, Ianchovichina & Rijkers, 2016) can play in the dilemma of exiting or staying. A prominent chain of thought pertains to the downside risk of placing valuable, hard-to-replace resources in jeopardy, especially when at-risk resources play a dominant role in broader network value creation (e.g., Yeo & Lee, 2021). By possessing such resources, MNEs and their management teams will weigh the benefits of withdrawing and protecting critical resources, rather than exposing them to the troubles of war. Conversely, these valuable, hard-to-replace resources may force an MNE to remain in the country to capitalise on their endowments and avoid erosions to their competitive advantages (Chen, 2017;Dai, Eden & Beamish, 2017). Thereby, reframing a manager's mental representation.
In addition to experience, mental representations are also dependent on structural capabilities, resource immobility and the operational slack Table 1 Factors relevant to decision-making in conflict-torn countries.

Supply Chain Adaptability
Extent to which an MNE can disperse their production functions spatially and temporally following the changing combat frontier. Conflicts may be relatively short-lived while others may persist for decades. The latter tend to be associated with failed states. Getz and Oetzel (2009)

Conflict Rationale and Issue
The issue of a conflict refers to what the fighting is about. Intrinsic conflict issues, such as ethnicity or religion, appear complex and insoluble. However, if economic factors, such as greed or grievance appear to be very prominent in the conflict, the prospect of engagement is greater. Wolf, Deitelhoff and Engert (2007) Disruptiveness Degree to which a conflict disrupts basic service provision or the integrity of critical infrastructure. Berman (2002); Brück, Naudé and Verwimp (2013); Getz and Oetzel (2009). Proximity to the Combat Frontier Geographical proximity to firm (close or far away), and the changing spatial contours over time. Beamish (2013, 2017); Oetzel (2009), Witte, Burger, Ianchovichina andPennings (2017).

Social-sector Stakeholders
Various social-sector stakeholder groups may bring issues to the attention of the firm. The groups can also create coercive, utilitarian and normative pressures. Social-sector stakeholders include non-governmental organisations, local consumers, local communities, and local media outlets. Bennett (2002); Boele, Fabig and Wheeler (2001); Darendeli and Hill (2016); Darendeli, Hill, Rajwani and Cheng (2020); Oetzel and Getz (2012).

Political-sector Stakeholders
Political-sector stakeholder groups can restrict or constrain the activities of foreign firms through utilitarian and normative pressures. Political-sector stakeholders include national government and local community leaders. Calvano (2008); Darendeli and Hill (2016); Darendeli, Hill, Rajwani and Cheng (2020); Oetzel and Getz (2012).

International Arena
Formal institutions (home and foreign governments) Cognitive, normative and regulative pressures applied to foreign firms to exert influence over or constrain the activities of MNEs abroad. Actions include condemnation, economic sanctions, calls for repatriation, and involvement in military intervention.

Financial Market Performance
Economic consequences of conflict can impact stock market indices, which can lead to a devaluation in company stock prices and cause financial distress and undermine the rewards of staying. Abadie and Gardeazabal (2003); Fielding (2003); La Ferrara (2007, 2010); Patey (2006Patey ( , 2007. (continued on next page) of the firm's activities . If an MNE is highly dependent on the resources and economic returns from the conflict-torn country, they are likely to tolerate the turbulence of war (Guidolin & La Ferrara, 2007;Li, Murshed & Tanna, 2017). This tolerance is dependent on network structure and supply chain adaptability, which has become a staple in recent research (Abadie & Gardeazabal, 2003;Amodio & Di Maio, 2018). For affected firms to remain in warring countries, they must be able to disperse their functions spatially and temporally under the changing combat frontier (McDougal, 2010). Without such agility, managers are likely to view the possible impact of conflict events as more severe, thereby altering their mental representation and swaying them toward exiting.

Host country environment
Focused on the hazardousness of place, the host country's environment is undoubtedly prominent in mental representations. A review of this literature has identified that the contours of combat frontiers (Dai, Eden & Beamish, 2013;Getz & Oetzel, 2009;McDougal, 2010), the duration, disruptiveness and obstructability of conflict zones (Jamali & Mirshak, 2010, Naudé, Verwimp & Brück, 2013, conflict rationale and issue (Wolf, Deitelhoff & Engert, 2007), and actor constellations (Ballentine & Nitzschke, 2004;Berman, 2000) are prominent spatiotemporal dynamics that informs an MNE's on-the-ground interactions. The fundamental proposition here is that localised, small conflict events pose less risk and obstructability to critical infrastructure, whereas dispersed, nationwide fighting can make basic infrastructure redundant and halt the normal functioning of an economy (Kadry, Osman & Georgy, 2017;Patey, 2006, Toft, 2011. Hence, dispersed conflicts can make it futile for firms to sustain operations and remain (Brück, Naudé & Verwimp, 2011, thus swaying a manager's interpretation of the event. However, there is the caveat of firm location relative to conflict zones, i.e., are they located inside or outside zones of fighting? Empirical work has confirmed that MNEs can be equally affected by small-scale, localised conflict events when they are located within the parameters of fighting (Dai, Eden & Beamish, 2013). When located close to, or inside, conflict zones, MNEs can find their 'business interests' located in de facto areas of sovereignty controlled by rebel groups (Le Billon, 2004), which can make them strategic targets (McDougal, 2010). This predatory behaviour, even when the zone is small, can bring greater intensity and exposure to the doorstep of firms (Witte et al., 2017). Thereby bringing conflict dynamics to the forefront of a manager's evaluation of the situation.
Although conflict dynamics will determine a manager's perception of the severity of the event, these events do not occur within a vacuum. Instead, their trajectory and impacts are determined by domestic, formal institutions and their governance capabilities (Kolk & Lenfant, 2015;Voors & Bulte, 2014). A core proposal is that the stability of domestic governance can determine the host government's capacity, resources, and resolve to control the situation and its repercussions (Banfield, Haufler & Lilly, 2005;Getz & Oetzel, 2009). If domestic governance mechanisms are stable, they can send a signal of reassurance to MNEs  as firms would assume that the host government will act or have the capacity to curb the violence. With such signals, managers will likely view the unfolding conflict-torn country as less severe, thereby placing governance stability at the heart of their mental representations.
Another group of factors that could enter mental representations is local stakeholders, i.e., proximate mechanisms (Lee, 2011). It is argued that ethical pressures stemming from stakeholders can determine whether and how a firm decides to respond (Oetzel & Getz, 2012). These pressures present constraints to business operations and place expectations on the conduct of the firm. By internalising stakeholder pressures, an organisation's employees can predict organisational commitment in times of war. Based on management and psychology theory, Reade and Lee (2012) found that operating in environments beset with violent conflict may exact an indirect cost on the firm. This is partly through disconnectedness and lowered employee commitment. Disconnectedness from employees, although damning for business continuation, stems from a firm's interpretation of the conflict landscape and the provisions that they put in place. When deemed inappropriate, or not enough, by employees, they are likely to view the decisions of the MNE in a negative light. Given employees' prominent role in organisational resilience (Suder et al., 2017), a loss of commitment can weaken firms' tolerance and ability to remain. Hence, internal and external local stakeholders have the potential to feature prominently in a manager's mental representation given that they can enact crushing constraints on strategic possibilities when caught between the rock and a hard place.

International arena
Researchers have framed responses to violent conflict as a function of extra-territorial pressures which can introduce new dynamics into the decision-making calculi. Although not an immediate effect, factors at the macro-level can factor into mental representations through constraints and contingencies being placed on managers via salient stakeholders or fear of wider ramifications for the MNE.
Research has shown that home and international institutions can influence and/or constrain activities (Abdalla & Ak, 2015; Kolk & Lenfant, 2013;Meyer & Thein, 2014). Capturing the mechanisms through which international institutions shape outcomes, Meyer and Thein (2014) place emphasis on cognitive, normative and regulative pressures. Even when operating in remote locations, MNEs need to adhere to the norms, principles and legal regulations of their home country. Such pressures undoubtedly frame how a manager will view the unfolding situation as failure to comply with expectations will raise attention and result in institutional stakeholders applying pressures. A common mechanism is the application of regulative pressures such as sanctions (Sherman, 2001). Not only do sanctions place financial pressures on MNEs, but they also place them under the spotlight. Creating this spotlight are the activities from stakeholder groups applying cognitive and normative pressures. Civil society, NGOs and media institutions are known to influence MNE behaviour by drawing attention to ongoing operations, and questioning whether their practices are ethical (Oetzel & Getz, 2012;Oetzel et al., 2009;Boele, Fabig & Wheeler, 2011). Faced with this spotlight, which can not only amplify the threat but also introduces new consequences to the mix, MNEs and decision makers often opt for low-profile strategies with exiting a common outcome to appease actual or perceived expectations. Therefore, it has become standardised that afflicted managers will consider the pressures, or risk of pressures, emanating from outside the conflict-torn country when framing the problem environment. Meyer and Thein (2014); Oetzel and Getz (2012); Patey (2006Patey ( , 2007. Another issue that MNEs need to contend with is market value volatility and shareholder demands. Research has acknowledged the economic consequences of operating in risky locations as warring countries can negatively impact stock market indices (Arin, Coferri & Spagnolo, 2008;Guidolin & La Ferrara, 2010;Karolyi & Martell, 2010), devalue company stocks (Patey, 2006), and cause financial distress through abnormal returns. Under such circumstances, previously favourable conditions that created a strong incentive to remain (Chen, 2017;Rosenau et al., 2009) can be removed from decision-making calculi. A prime rationale is investor sentiment. For example, the Sudanese war led to stock market volatility and devaluations for remaining MNEs, which increased shareholder opposition and forced affected MNEs to take steps to improve their image (Patey, 2006). Given these pressures, it is common for managers to emphasize satisfying shareholder demands in their decision-making via exiting.

Critical decision method
To construct the small world representations developed by afflicted managers, we adopt the critical decision method. This method permits researchers to elicit knowledge from experienced personnel who are confronted with non-routine, challenging decision-making incidents that are characterised by ill-structured problems and uncertain dynamic environments (Harenčárová, 2015;Klein, Calderwood & Macgregor, 1989). Through cognitive probes, the critical decision method allows us to understand our managers' judgements, while uncovering what context-specific factors and critical information are sourced, collected and evaluated to construct their small world representations.

Sampling and study setting
The population of this study consists of MNEs from the oil and gas industries. Several reasons accounted for this selection. The first is the paradoxical 'resource curse' (Basedau & Lay, 2009). This is a well-known phenomenon that proclaims that resource-rich countries are more prone to experience episodes of violent conflict. The second, which is interconnected, is the prevalence of oil and gas MNEs in conflict-torn countries given the sector's high asset specificity, asset immobility and limited locational ambidexterity. This is evidenced within political risk literature (e.g., Obi, 2010;Patey, 2007). Despite limitations to locational ambidexterity, it is important to mention that oil and gas MNEs do not uniformly remain in conflict-torn countries. The third reason is the fact that extractive infrastructures are prime targets when violent conflict erupts, commonly being sabotaged, seized by rival groups, and the first year, which allowed us to develop a structured timeframe for our CDM interviews and small world representations. Participants were responsible for defining their own conflict events, its location and scope based on their lived experiences.
ransacked to finance war activities (Ross, 2004;Toft, 2011). Potential informants from the oil and gas industries were identified through a rigorous and iterative screening process that involved: i Using the Armed Conflict Location & Event Data Project (ACLED) to identify countries that have experienced a form of violent conflict.
ii Matching oil and gas MNEs to the respective conflict locations. iii Identifying relevant top decision-makers with the pre-requisite that they were in post before and during conflict events.
This purposeful sampling approach resulted in a target population of 214 managers with thirty-one agreeing to participate in the study.

Fig. 1.
Example decision chart (Respondent #24, Situation 1), Step 1 Note: 'I' denotes information was that collected, or sensed, 'D' shows the decisions that were made, and 'A' reveals the actions that were taken. The DAT shows Informant #24's experience in the Kurdistan Region of Iraq.
Collectively, respondents had first-hand experiences in twenty conflicttorn countries. Aligning to ACLED's definitions, our respondents' violent conflict experience captured subnational events, which were centred around one location in the country, but also nationwide conflicts that had multiple zones of fighting dispersed across many localities. The sample offers ample variations in MNE ownership (i.e., state, public, private), managerial experiences, and conflict dynamics (see Table 2).

Data collection and analysis
Semi-structured interviews were conducted online, recorded and transcribed. Structural recommendations from Mendonca (2007) assisted in the development of our interview guide, probing questions and procedure.
Following these structural recommendations, before the interview, informants were sent a pre-interview Critical Incident Form (Keatinge, 2002) to encourage self-reflection and identify the set of conflict incidents for each informant. Using the returned Critical Incident Forms, each interview then began with the investigator encouraging informants to provide an unstructured incident account with probes and prompts used to encourage elaboration. After the incident was outlined, a timeline of events was established using direct questioning. The product of this stage was the identification of points where information was received or sensed, decisions were made, and actions were taken. Following timeline verification, probe questions were used in the 'deepening' phase to focus on specific aspects of decision points and their role in shaping their strategic responses. After developing a detailed picture, 'what-if' queries were used to pose hypothetical changes to the incident to explore how variations in conditions or situations could impact response decisions. This process was iterated for each critical incident noted by the informant. Each interview ended with a contact summary sheet to summarise salient points. The thirty-one interviews ranged from 90 to 120 minutes long and were conducted in the English language.
Following the transcription of the interviews, organisation and analysis of the data followed a five-step structured approach outlined by Harenčárová (2015).
Step 1 involved the creation of a decision chart to outline the sequence in which information was retrieved, decisions made, and actions taken (see Fig. 1).
Step 2 (Fig. 2) used the decision chart to organise the relevant details from the transcripts into chronological order.
Step 3 created decision analysis tables focused on the causal links between situational assessments, incident-related factors, goals and rationale with the decision outcomes (see Table 3).
Step 4 used the data analysis tables to identify items of interest, based on the study's frame of reference (what context-specific factors are commonly added into mental representations of the problem environment). The last step   included comparing and collating common items of interest across all reported incidents. Procedures to audit content for credibility were undertaken with a focus on reliability and validity. Two tests were conducted, which were independent extraction and inter-rater reliability checks (see Butterfield et al., 2005 for an explanation). These tests returned an auditor-investigator agreement of 89% and an inter-rater reliability score of 86%, indicating no issues.

Findings: heuristics and small world representations
Our critical decision method analysis returned 117 decision outcomes (Table 4), with exiting being the most common response choice. Given our research question, Table 5 summarises the heuristics that informed the decisions of our respondents. Despite some respondents' variations, our approach allowed us to identify common discovery and evaluation heuristics.
Our discovery heuristics (in Table 5) capture what information managers were looking for (framed as questions) to develop the initial   sketch of their respective problem environments. Guiding their information search and collection, these questions allowed our managers to build higher-order heuristics that captured specific elements of the violent conflict environment. As shown in our categorisation (bold headers), our managers routinely focused on nine higher-order heuristics. After information was collected, and higher-order heuristics developed, managers then used evaluation heuristics to assess the data and understand the unfolding situation. Such evaluation phases facilitated the creation of small world representations, which were then used to review strategic options and select the optimal response between exiting or staying. Fig. 3 presents a common small world representation of the thirtyone managers in our sample. The grey box represents the firm in the conflict-torn country. As depicted, the firm box includes the manager, the employees and the internal higher-order heuristic of 'flexibility of operations'. Rectangles denote actors (e.g., conflict actors, national government, suppliers). Circles represent conflict events and properties within the country (e.g., conflict events, infrastructure and information), with arrowed lines indicating direct relationships between actors, actors and events, and text summarising its nature/context. Nonarrowed lines represent the impact of actors on events and properties within the country. The conflict-torn country is contained within dashed lines, with outside factors (e.g., headquarters, shareholders, home governments) moving past these dashed lines to have an impact on our managers' representations and decision-making.

Access to information
A major element of analysing the respondents' small world representation was their appraisal of information accessibility. When conflict erupted, discovery heuristics began with attempts to understand what information they had access to, and from where. Interwoven with relational connections to local networks and public officials, managers repeatedly acknowledged that inaccessibility impeded their understanding of unfolding events and clouded judgements about how best to respond. Supporting the concept of 'informational advantages' (Buckley et al., 2016), the prevalence of this higher-order heuristic confirms the significant role that information plays in permitting managers to develop richer sketches of the environment. If managers did not have access to information, fast-and-frugal heuristics would take hold in which they had to understand the situation with little information. Logically, this factor will have a direct impact on the reliability and Table 5 Common discovery and evaluation heuristics of managers in conflict-torn countries.

Discovery heuristics Evaluation heuristics (How information is evaluated) (Questions asked to search for information, and grasp the situation) Access to Information, 52% Participant
Rate Validity a Where is our information coming from?
Can we get timely updates on the situation? Who can we rely on in the country? Will our connections help us?
Trusted and timely updates make it possible to grasp the unfolding situation. Understanding of the situation allows timely adaptations, which can help protect equipment and personnel.

Conflict location, 97% Participant Rate Validity
Where are the conflict events located? Are these events contained in one location, or spread across many places? How close are the conflict zones to our location?
Assess probability that conflict events will threaten business operations and employee safety. Localised conflicts can be avoided with adaptations to normal activities. Dispersed conflicts will make it difficult to transport machinery, workers and assets between owned sites in the country.  Capital restrictions will affect our operational cash-flows. We will face challenges in meeting dayto-day expenses. Shareholders will not accept the investment risk and can force us to stop operations and leave. Sanctions will raise scrutiny on our operations and create unease for staying.
a To capture content validity and construct relevance, levels of agreement were calculated for each category and subcategory (Butterfield et al., 2005). Based on thresholds for participant rate validity, rates greater than, or equal to, 25% speak for enough of the data and are valid.
richness of the other higher-order heuristics in the small world representation as indicated in Fig. 3. When informants did not have access to timely and reliable information, they remarked that they simply would not "know what was going on and where" (#9) and would not be able to "avoid the worst of the fighting" (#8). This made managers feel uneasy about the situation, with decisions more likely to be made prematurely. Explaining information inaccessibility, seven managers mentioned that they were relatively new to the country and had not developed meaningful connections with local and national officials. For example, informant #11 recalled that their company only had one year's worth of experience in Pakistan and did not have access to informers to gain intelligence on Al Qaeda's whereabouts and intentions. Inabilities to plug the information gaps raised the probability that managers would not be able to effectively adjust operations to reduce the impact on the firm. Thereby tilting management decisions more towards the exit.

Conflict location
Anchored on information accessibility, the most common discovery heuristic was to ascertain where fighting was happening within the country. Based on these geo-location information cues, managers could then estimate the probability that conflict events would pose a threat to resources, personnel and business activities. If information revealed that fighting was located far-away from company sites, managers would evaluate these cues to determine a lower probability of risk and obstruction. As noted by respondent #25, for example, "risk is only risk when it is at the door. When it is not close to you, it is a mere story to tell the grandkids." Respondent #1 further supported this position: …if the bombing is 50 miles away from you and there was a target that was 50 miles away from you, that has nothing to do with you, it might as well be 5,000 miles away. You were not the target; your risk profile had not necessarily changed. For example, I was with [Name] in Damascus, and the bombings were being carried out away from the capital. From where I was, it may have been the other side of the world.
Conversely, if geo-location cues revealed that company sites were located close to, or within, a zone of fighting, this intensified the perceived threat for managers and dealt a "final blow" (#6) to operations. Introducing interrelationships into the small world representation, when fighting was in the proximity of their compounds, managers recalled that it increased employee anxieties, placed valuable resources on the radar for looting or destruction, and tested the flexibility capabilities of the company. Faced with such reality, this higher-order heuristic would hold more weight in small world representations and sway decision-making toward exiting. For example, respondent #11 recollected that when they identified that Shia'a Islamists were advancing on their location in Pakistan, they had no option but to withdraw. The respondent noted their lack of strategic options was due to the extreme tactics used by the Shia's Islamists including attacks on compounds, bombings and hostage-taking.

Non-state actors' involvement
Interconnected to the higher-order heuristics on conflict location, as highlighted in Fig. 3, is managers' search for information cues on actors involved in conflict events. Our respondents recounted that their discovery heuristics uncovered attempts to understand which non-state actors were participating in the violence and what their motivations are. Although there were variations in non-state actors (e.g., Al-Qaeda- linked terrorists in Algeria, tribal clans in Nigeria, and civilians in Egypt and Tunisia), when these cues were collected, they were used to anticipate the duration and intensity of the fighting. This contributed to mental representations of the expected magnitude of the impact on the firm.
If non-state actors were motivated by economic gains or causing maximum obstruction, managers estimated a greater probability of interruptions and direct attacks. This swayed our respondents to place more emphasis on these higher-order heuristics, which increased the likelihood of exiting. Alternatively, respondents described multiple situations where fighting was intensifying, but non-state actors avoided inflicting damage on their operations and targeting their personnel. This permitted our respondents to evaluate these cues and reach the timespecific decision that they could remain unscathed, thereby tolerating the conflict events. Based on experiences in Nigeria's Niger Delta, respondent #7 summarised this phenomenon by stating that: The clans did not want to inflict any damage on us and did not want to scare us, as we were the only sources of employment. Because of this, we went untouched. So, there was no real risk to us, and we did not need to leave the area. We felt safe! Unfortunately, information on motivations and expected actions can be incomplete, non-definitive and ambiguous. In these situations, managers may be pushed towards fast-and-frugal heuristics with an emphasis on rapid decision-making on limited information. Alternatively, managers may opt to continue their attempts to collect information on non-state actors, which can result in them being a 'sitting duck' as events continue to unfold around them. For example, respondent #15 mentioned that, despite attempts to collect cues, there was no prior intelligence to indicate that terrorists would attack their compound, declaring that: …when Islamist terrorists attacked our In Amenas compound, we knew the situation was dire, and we had to get out as soon as possible.
Accordingly, as an important higher-order heuristic, when managers can collect information on participating non-state actors, they can better grasp how events will unfold. This results in a more accurate small world representation. However, when this information is unattainable or incomplete, managerial decision-making will enter an uncertain phase, which can undermine tolerance and shift the preference of strategic options toward exiting.

Government stability
Contributing to our respondents' initial sketch of their environments, discovery and evaluation heuristics commonly featured assessments of the host governments' perceived strength and projected stability. When faced with outbreaks of violent conflict, our respondents tried to collect cues on their host government's ability to maintain law-and-order, remain in power and avoid power struggles. Reflecting on the interrelationships in Fig. 3, respondents perceived that government stability would have a direct impact on the intensity, duration and geographical scope of conflict events with the probability that interruptions and destruction will affect firms' infrastructure.
From the outset, a sizeable portion of our respondents reported that they searched for signals into how host governments could, or would, manage the situation. If collected cues suggested that the government would not be able to combat and control the situation, our managers feared that there could be anarchy on the streets with "crimes being committed in broad daylight" (#18). If such tribulations of war started to enter the fabric of day-to-day society, perceptions would start to shift with managers beginning to feel less comfortable with continuing operations. Moreover, if managers gauged that governments were politically weakened, or could fall, they anticipated that Pandora's box could open, with a multitude of possibilities, including conflict escalation, asset repatriation, and long-term resource shortages. Reflecting on their experiences with the 2011 uprising in Libya, respondent #23 said that "when Gaddafi had control, we knew that we were guaranteed support and our contractual rights would be honoured", but when Gaddafi's regime was toppled "big question marks" were raised that made the managers quickly reassess the situation with updated information. Facing such circumstances, respondents recollected that they viewed staying as too risky, too costly and untenable. Therefore, framing their small world representations in a negative light, and pushing the firm toward exiting. Capturing this predicament, respondent #18 noted: It was absolute chaos and anarchy. The government had absolutely no control with their policing. There were riots on the streets, crimes being committed in broad day light, vandalism, and the worst of all, looting against our premises. You could be driving through the area on the way to work and you would be fearful to stop at traffic lights in case you were carjacked. It was a 'free for all.' It made it too costly to remain in the country as the rewards were unknown.

Transport Restrictions
We observed that our respondents' evaluation heuristics tried to estimate the probability that transportation restrictions would occur, where they would occur, and if, how and where these obstructions would impact the activities of their company. When collected information signalled that transportation obstructions and restrictions would occur, our respondents noted concerns around blocked roads from demonstrations, bombing and fighter presence at critical infrastructure points, and 'holes' emerging in the inventory at production sites. When these concerns came to fruition, our managers experienced increasing inefficiencies, which impeded their ability to function and meet production quotas. Correspondingly, management teams began to question the logic of remaining in the country if they are "sitting ducks with no work to do" (#11). Reflecting on their time in Syria, respondent #14 acknowledged that: Failures to be able to plug the holes in [Name] activities and supply chain left us with only one choice, and that was to exit Syria. After all, why would you stay if operations are essentially halted? You would essentially be sitting there in the middle of a battlefield, with no purpose, while still taking on all the risks associated with being in a dangerous environment.
Therefore, when interrelationships between the conflict-specific higher-order heuristics indicated that the firm would face difficulties in their transportation activities, the evaluation of small world representations swayed decision-making toward exiting.

Flexibility of operations
To approximate whether they would be 'sitting ducks', our managers engaged in internal calculations to see whether they had flexibilities embedded in their country operations. These discovery heuristics asked questions about whether it was possible to move company activities to safer locations in the country, whether the company had inventory reserves, "assets in different locations" (#11) and whether their geographical dispersion would allow them to avoid the worst of the fighting. If the discovery heuristics returned positive signals about operational flexibilities, our respondents mentioned that they positively evaluated their circumstances. Such flexibilities were noted to allow our respondents to buffer against resource shortages, move employees to safer locations, and continue production activities. Rather than rely on intermittent resource imports from unstable ports, which were often under siege and looted, respondents recalled how they were, to some degree, self-reliant and unfazed. As such, through evaluation heuristics, managers were able to picture how this operational advantage would allow the company to weaken the impact of violent conflict events, as shown in the arrowed lines in Fig. 3. This direct relationship in the small world representation made staying an optimal choice. This is captured by respondent #16 who explained that:

With our equipment located in multiple locations, it was easier for us to continue operations at certain facilities, as we always had back-up resources close-by. It would have been a different picture if this was not the case, as bringing resources from outside the country to the Port of Aden was a definite no-no.
There were, however, incidences in which the higher-order heuristic of operational flexibility could be undermined. Numerous respondents reflected on conversations that they had with their headquarters in which they were informed that the parent company had to avoid "being associated with the regimes atrocities" (#21), that the short-term economic returns were not worthwhile, and that "they were stuck between a rock and a hard place" (#19). These signals immediately shifted the managers' small world representation with a realisation that business continuity was no longer an option, even if other higher-order heuristics were indicative of continuity and staying in the country.

Scrutiny
As conflict events erupted and evolved, managerial heuristics attempt to identify and interpret any scrutiny that they were facing. Often through the channel of media outlets, scrutiny was observed to have an impact on other higher-order heuristics in the small world representation. As depicted in Fig. 3, media scrutiny can unease employees and make them reluctant to remain in a conflict-torn country. This can then undermine the strategic advantages provided by operational flexibilities. Operational flexibilities can also be undermined if media scrutiny creates reputational concerns at the headquarters level. When media scrutiny moves beyond the conflict-torn country, it can create pressures for the headquarters to stop operating in the focal country and instantly shift managerial mental representations, making exiting the only option.
During their discovery heuristics, several respondents recollected that local media and newspapers were printing stories that questioned why they were continuing "unethical operations" (#5) and "exploiting the resources of their homeland" (#11). Under such negative limelight, respondents predicted that their firm could face new risks such as local communities opposing the continuation of their operations, and their locations becoming possible targets for attacks by non-state fighters. Respondent #1, for example, recalled that: The local media was portraying us in a very negative light. I remember that El Khahar, El Djazair, and Echorouck El Yawmi ran stories about our presence in the country and how we were taking the peoples resources away. This created a lot of tension with the local communities, and we had incidences of riots. Of course, this posed some risks to our employees and our facilities. When the stories kept running during the conflict, we decided to get out of there.
Such circumstances caused scrutiny to take a more prominent role in evaluation heuristics, with managerial assessments being more negative in their outcomes. However, firms were not just placed under the magnifying glass in their respective host countries, there were several instances where media scrutiny occurred in the firms' home countries. With the caveat that these pressures only meaningfully entered the small world representations for managers of publicly listed firms, respondents noted that media attention at home created a stark image of the conflict events that were often exaggerated and shone a negative light on the company's continuing operations. This scrutiny created a ripple effect that influenced shareholders' sentiment. Accordingly, evaluations of strategic options were often inundated with shareholders venting their "concerns with the situation" (#17) declaring that "they couldn't care less if operations could continue" (#2). These pressures made respondents reassess their small world representation as they wanted to avoid the spiralling situation of detrimental attention. Accordingly, managerial sketches of the problem environment were reframed, with this pressure taking prominence and pushing managers toward the exit. Capturing this phenomenon, respondent #24 informed us that: …the Dallas Morning News wrote about all the fighters that were closing in on Erbil and [Name] operations in that area, and people would ask "doesn't that cause you concern?" I would respond no. However, our shareholders were a lot more worried. They did not want to take the heat, so they applied pressure to our executive management in Texas, and we had to withdraw from Iraq.
Accordingly, these lived experiences portray the strong influence that media attention and scrutiny can have on mental representations through the interrelationships with other higher-order heuristics.

Employee concerns
Possibly negating the positive signals provided by other higher-order heuristics, respondents noted that they were constantly asking questions about their employees' willingness to remain in the conflict-torn country. This was an important discovery heuristic to update promptly as it could quickly shift capacities to remain in the country. Noting issues with employee retention, inability to preserve a "skeleton staff" (#4) and employees' "unwillingness to stay in the country and travel to the office" (#13), respondents reported that despite "risk assessments not raising a red flag" (#15), they were dealt a check-mate card by their employees who restricted their abilities to continue operations. Such concerns arose irrespective of the company's employee base, i.e., the ratio of expatriate versus local employees.
Reflecting on experiences in Iraq, respondent #6 acutely summarised that "employees are the backbone of what we do. When they do not want to work, you cannot simply replace them at a click of a finger. They were the ultimate decision-makers for us." Consequently, this significantly complicates the small world representation process for afflicted managers as they will need to acknowledge that employees will develop their mental representations. Affected by other higher-order heuristics (shown in Fig. 3), employees will consider their information points and decide how they want to respond. If employees become unwilling to remain in the conflict-torn country, this will undermine the company's ability to continue operations, make operational flexibilities redundant, and sway managers toward the exit.

Capital restrictions
With the risk that capital restrictions can be a severe higher-order heuristic when they occur, managers repeatedly asked questions about the probability that governments would apply economic sanctions on the conflict-torn country, or even directly on their companies' operations. If signals were received that indicated sanctions were being considered, respondents noted concerns about capital inflows and outflows within the country, which could undermine their abilities to pay employees, contractors and suppliers. This could then lead to constrained resources, holes in supply chains and discouraged employees. Faced with the issues of currency inconvertibility and restrictions on financial transactions, it was recollected that headquarters became nervous about keeping sizeable capital reserves in the country, which also reduced their willingness to commit more capital. Parent companies unanimously ordered the removal of excess capital, thus reducing the timescale in which country managers could continue to fund operations. Capturing this predicament, respondent #17 declared that: It became impossible to freely convert money out of Syria. I think because the government wanted to restrict resources leaving the country. This unfortunately had an impact on the money coming into the country. [Name] in London was only able to send small amounts of money.
In a similar vein, three respondents also recalled how the American Government notified them that if they remained in Yemen and Syria then sanctions would be applied. These intense actions were evaluated to constrain options, which left no wiggle room for affected managers. For example, respondent #11 shared that: We were able to continue operations, but the Americans threatened sanctions, which pushed us to leave Yemen. We were accused of financing violence by staying. We did not immediately leave as we thought we were calling their bluff. However, they did apply the sanctions, and our money soon dried up; our stocks were very poor. We had no other choice but to leave.
The emergence of these issues in small world representation can have a direct knock-on effect on other higher-order heuristics. As captured in Fig. 3, capital restrictions can weaken the positive role of operational flexibilities while increasing the scrutiny that companies may receive in the media. These knock-on effects can subsequently make staying a suboptimal choice for the manager. Overall, our small world representation indicates that afflicted managers will be impacted by nine higherorder heuristics, with their interrelationships determining which contextspecific factors will feature prominently in decision-making calculi, and whether any would be filtered out. Consequently, these interrelationships will determine whether managers exit or stay.

Discussion
Previous work has outlined the multiplicity of decision-influencing factors that routinely affect MNE responses to violent conflict. However, spotlighting the decision-making realities of affected managers, coupled with our intention to contribute to calls for further research (Jamali & Mirshak, 2010;Oh & Oetzel, 2011, 2017, this study demystifies the configuration of context-specific factors that are commonly filtered into a manager's decision-making calculi (Czinkota et al., 2010). Specifically, through our use of small world representations, we identified nine higher-order heuristics (refer back to Table 5) that populate a common small world representation of managers confronted with violent conflict. Each of these higher-order heuristics was found to directly influence decision-making, whilst also intensifying or subduing the impacts of other heuristics through the interrelationships within the small world representation as depicted in Fig. 3. To contextualise these findings, we relate the higher-order heuristics back to the literature using Patey's typology (Table 1) to pinpoint the specific contributions of our findings.

Higher-order heuristics and the host country environment
Given that the host country environment contains the physical threat, our findings underline that managers focus their discovery and evaluation heuristics on trying to estimate how violent conflict events will spatiotemporally evolve -whether the conflict will cause operational disruptions and restrictions, whether the host government can control the situation, and whether the firm will receive unwanted attention and scrutiny. Using the collected information on these contextspecific factors, managers constructed their small world representation to determine the level of physical risk to their company, the prospects of circumnavigating the risks and continuing business activities in a manner that helps navigate the choice between exiting and staying.
Conflict locations -whether located inside or outside a conflict zone, and the spatiotemporal morphing of the size and location of conflict events (Dai, Eden & Beamish, 2013) -were centrefold in the small world representation. This also moderated whether a manager viewed conflict events as something that warranted an exit response. We observed many instances where conflict zones were located far-away from the firm, with managers concluding that the fighting would not impact them. Conversely, when fighting was located close to the firm, the higher-order heuristic had knock-on effects on other higher-order heuristics, notably transport restrictions, flexibility of operations, and employee concerns. Such observations add weight to Beamish's (2017, p.1493) declaration that the greatest concern for decision-makers is the "non-uniform distributions of warfare" and their spatiotemporal interactions with MNEs.
Furthermore, when fighting is intense and 'at the door', rather than instantly deciding to exit, we discovered that managers would ask questions about proximate stakeholders such as media scrutiny, local communities and the non-state actors involved in the fighting. Often deemed the most salient and influential context-specific factor (Oetzel & Getz, 2012), proximate stakeholders can undermine any strategic advantages that the firm possesses via the interrelationships in the presented small world representation. This prominence stems from their capacity to determine the extent to which an MNE will be targeted and face scrutiny, thereby defining the magnitude of additive exposures and sensitivities. Faced with favourable stakeholder interactions such as government-backed securitisation or non-state actors' avoidant of inflicting damage on operations, managers often view conflict events as mere obstacles that do not involve them. Contrariwise, when interactions are negative, these stakeholders added a new risk dimension to the managers' calculi and undermined the advantages that the firm possessed in being able to adapt and continue operations. With proximate stakeholders featuring prominently, we argue that it is paramount for research to build on the wealth of existing knowledge (e.g., Oetzel & Getz's, 2012) to further delineate the role of proximate stakeholders in shaping resilience. Specific focus should also be placed on isolating the most influential proximate stakeholders and their impacts on tipping points.
Further complicating decision-making, conflicts' disruptiveness, obstructability and corresponding restrictions determine whether conflict events are impactful. MNEs can be located inside zones of fighting and face immense stakeholder pressures but buffer the threat and function through operational flexibilities. With this favourable higherorder heuristic, managers take a dogmatic approach and accept that they are operating in a risky location and place less emphasis on the problem environment. Yet, even the most dogmatic manager can face untenable hurdles. When combat frontiers morph and become deleterious, a country's infrastructure can be destroyed, thereby crippling supply chains and undermining the feasibility of day-to-day operations (Getz & Oetzel, 2009). Unable to continue operations and avoid becoming a 'sitting duck,' managers in countries suffering more destructive fighting are prone to exiting. We thus argue that strategic responses are more interlinked with the extent to which day-to-day society continues, and functionality remains, rather than the inside-outside logic offered by location studies.
Further strengthening this argument is the finding that managers actively asked questions about the stability of the host government in their discovery heuristics. Managers anticipated that if the government were stable and capable of controlling the conflict events, then the magnitude of impact on the firm would be reduced. This was most notable in the interrelationships between government stability and transport restrictions in the small world representation (Fig. 3). This is an important insight with managerial relevance. Understanding the fragility of national governments and capacities of containment and securitisation can provide an early red flag that can trigger swift reconfiguration or timely exiting decisions. Therefore, in a practical sense, managers must develop capabilities of 'reading' governance warning signals to decipher the trajectory of conflict situations and identify possible early warning signals.
Drawing from Table 1, our study confirms the role of the host country environment in shaping managerial decision-making. However, we underscore those impacts are not equal. As such, some interesting nuances are raised. First, conflict zone locations are important, but only when they are deleterious and accompanied by weak country-wide governance. Second, proximate stakeholders can overtake the prominence of conflict location, more so when fighting is not severe. Third, only conflict events that directly disrupt business infrastructures will sway strategic outcomes, as most managers take dogmatic approaches and accept fighting as a facet of operations.

Higher-Order Heuristics and Firm-Specific Considerations
Our small world representation indicated three higher-order heuristics that are categorised in firm-specific considerations. When violent conflict erupts, we observed that managerial discovery and evaluation heuristics involved internal assessments of their ability to access timely information, the extent to which their operations could adapt to conflict events, as well as gauging the state-of-mind of their employees.
As an instrumental higher-order heuristic, we find that a manager's access to information will characterise the scale and scope of their discovery and evaluation heuristic processes. From the onset, we observed that managers are quick to interpret their ability to quickly accumulate information to improve their ability to forecast and react proportionately when conflict erupts. When managers had access to information, largely through their experience-related 'informational advantages' (Buckley et al., 2016), they were able to better grasp the spatiotemporal dynamics of conflict events and build a richer small world representation. This provided managers with a degree of confidence when assessing how best to respond. On the flipside, when access to information was limited, we witnessed a shift toward fast-and-frugal heuristics in which discovery and evaluation heuristics were conducted in a shorter timeframe. This resulted in a smaller world representation that made it difficult to grasp the situation and clouded judgments about likely trajectories and impacts, leading to premature exits or riskier stay decisions. Regarding access points to information, our study underscores a fine balancing act for managers as most relational capital that managers develop is with public officials in the ruling party. This can produce reputational liabilities when there are power struggles or non-state actors are politically motivated. The prominence of the higher-order heuristic on 'access to information' provides support to work on experiential learning and its strategic importance (e.g., Buckley et al., 2016;. We also confirm that the dispersal of resource locations will provide managers with a strategic advantage when trying to maintain functionality. The small world representation highlighted several interrelationships in which 'flexibility of operations' subdued the impact of conflict location and transport restrictions. This was not surprising given the wealth of empirical studies that reveal that dispersal capabilities dictate the extent managers can avoid the worst of the fighting (e.g., Dai, Eden & Beamish, 2013;McDougal, 2010). That said, notable insights from our study include that dispersal capabilities are primarily a strategic advantage afforded to managers that can rapidly gather information. Second, MNEs that avoid short-term resource depletion and premature exits are those that do not heavily rely on frequent resource imports as they have 'back-up' stocks dispersed across their country network. Third, MNEs need to have backup protocols for when their command-and-control offices cannot function. Adopting such measures can provide temporary refuge and take pressure off the necessity to immediately react.
Another critical finding that merits further discussion concerns the role of employees in managerial decision-making. Our higher-order heuristic of 'employee concerns' in the small world representation highlights the interrelationships and direct impacts that employees can have on the decision-making calculi. Through the development of their interpretations of the unfolding events, we witnessed that employees undermined flexibility of operations, and magnified other higher-order heuristics such as scrutiny, which shifted the mental representations of affected managers. Therefore, when in conflict-torn countries, employees will exact indirect costs on managers through lowered commitment and even resistance to a manager's modicum of control over strategic decision-making. With existing research on terrorism acknowledging the invisible impacts stemming from employees (e.g., Bader & Berg, 2013), we add a new dimension to HRM perspectives. Essentially, when employees are unwilling to remain, they present a final blow to strategic options. Employees, in such scenarios, do not simply lose commitment and adversely impact performance, as previously argued, but they can remove decision-making powers from management teams. This is the case for both expatriates and local employees.
This warrants further investigation as previous research tends to assume that employees are conduits of strategy implementation, whereas we discover that they are strategy shapers, and arguably the key decision-makers in conflict-torn countries. Thus, we echo the demand of Reade and Lee (2012) for more research on employee attachment and behaviours but take it a step further and call for work that examines internal power structures and shifting nuclei of decision-making. Such insights are strategically relevant for managers as they can highlight the critical points that need to be tackled earlier to avoid employee-induced restrictions. One way to achieve this would be to explore the discovery and evaluation heuristics of employees to pinpoint what context-specific factors feature in their small world representations, and how these small worlds interact with the representations developed by affected managers.
Our firm-specific considerations findings highlighted in Fig. 1 confirm the role of resources with further interesting insights. First, employees are paramount -an MNE can be endowed with information access, reconfiguration capabilities and stock flexibility, but if employees are unwilling to remain, such endowments become redundant. Second, experience is important, but not all experience is significant. Unlike prior studies, we only identify that country-specific experience is important with information accessibility featuring as a prominent rationale. Third, despite popular wisdom that the value and quantity of at-risk resources are integral, we find this may not be the case. Instead, if there is some form of 'back-up', even the most valuable resources may feature weakly in decision-making calculi.

Higher-order Heuristics and the International Arena
Given that managers simplify the problem environment, and tend to focus on proximate, lower-level context-specific factors, only two higher-order heuristics touched upon the international arena. These higher-order heuristics were scrutiny and capital restrictions. When these two higher-order heuristics are factored into a manager's small world representation, they had deleterious impacts on the flexibility of operations and employee concerns, which made it more difficult to maintain business functionality.
As captured in the small world representation, a predominant impact that managers can face from outside the boundaries of the conflict-torn country are sanctions. When sanctions are threatened or applied, managers can find their 'on the ground' strategies being halted, corroborating anecdotal studies (e.g., Kolk & Lenfant, 2012;Oetzel & Getz, 2012). Through interrelationships in the small world representation, we observed that decisions to respond to the threat of sanctions came from shareholders rather than afflicted managers. In essence, when governments raise the prospect of regulatory actions, it sent worrisome signals to shareholders and the headquarters. These pressures redirected evaluation heuristics away from higher-order heuristics that provided strategic advantages such as flexibility of operations, which make managers more prone to exit.
Without regulatory threats, actors beyond the conflict-torn country are unlikely to be given prominence in mental representations. This contradicts previous studies in that we find that when affected managers are contending with bombing next door, they are less preoccupied with the "hearsay that is occurring thousands of miles away" (#24). As such, even though these factors can sway decision-making, they only come to the fray in the most serious circumstances and are not regular features in conflict-torn countries. Our findings related to the international arena do raise some promising avenues for future research. Given that we focused solely on the individual level and did not place our participants into their organisational structures, we were not able to tease out where the mental representations were being made, and whether broader forces within headquarters-subsidiary dyadic relationships were accountable for the filtering process. As such, this offers a logical next step for further research.

Theoretical implications
Our study makes important contributions to existing research. First, by focusing on the lower-level vehicles in which decisions are nested (Buckley et al., 2016), our study differentiates itself by sifting through the multitude of theorised decision-influencing factors (Table 1) to highlight what factors, in practice, drive exit or stay decisions (Fig. 3). Rather than concentrate on the organisational level, or focus on singular factors, such as experience , conflict location (Dai, Eden & Beamish, 2013), or at-risk resources (Dai, Eden & Beamish, 2017), our microfoundational approach took an initial step in bundling together possible decision-influencing factors and letting experienced managers guide the sifting process. This approach returned nine higher-order heuristics that spotlight what matters and will hopefully direct future research. Second, our study builds on work that acknowledges that experience is paramount . However, rather than just confirming that experience guides decision-making, we also delineate the pathways in which experience can shape outcomes. From Fig. 3, we identify that experience matters because it shapes information accessibility that subsequently determines a manager's understanding of the unfolding situation. Finally, our study offers insights into the limited body of work on organisational resilience and vulnerability (e.g., Dai, Eden & Beamish, 2013, 2017. Our findings acknowledge that location (inside or outside) is important when delineating whether an MNE can function in the face of adversity, but we also highlight that proximate stakeholders such as employees are instrumental in determining organisational resilience. These stakeholders can undermine coping capacities (e.g., operational flexibility) and make exit the only choice.

Managerial relevance
We offer important implications for future decision-makers. Our findings suggest that afflicted managers should prioritise reflecting on the geographical dispersal of their operations. If not dispersed, managers should consider early exit as an optimum response given that a slight change in conflict-related events can quickly alter the situation and make them sitting ducks. Alternatively, when endowed with dispersed operations, our respondents' experiences indicate that managers can utilise this flexibility to function in the short-term (at least), thus allowing them to adopt a wait-and-see approach. However, we must add a caveat; before adopting a wait-and-see approach, our findings stipulate that managers should first engage with their employees. If employees view the situation, or managerial actions, as being untenable this can create complications and erode the buffers a manager holds regarding resilience-enhancing benefits (e.g., operational flexibility). Accordingly, by keeping on top of their employees' concerns, affected managers can better pinpoint when the pendulum swings from support for business continuity to exit being the only option. Our small world representation (Fig. 3) further implies that managers should initially reflect on their incountry experiences as this will determine their relational capital with formal and informal institutions, and thus information accessibility. If managers, or their firms, are relatively new to the conflict-torn country, they should consider an early exit and monitor the situation from afar as logic suggests they will face informational disadvantages (Buckley et al., 2016), which can give them an inaccurate picture of the unfolding events.

Limitations and further research
Regardless of our contributions, our study is not without its limitations. First, as our focus was on constructing a common small world representation, we did not consider the variations in conflict event type, duration and severity. Our respondents' recounted experiences were varied and included violent conflict events surrounding the Arab Spring, Niger Delta militants in Nigeria, the Syrian civil war and the Russian annexation of Crimea. Based on such variations, it is logical to expect that small world representations will differ following the conflict event type. Therefore, future research could examine the differences in higher-order heuristics given the specific event type.
Second, although we used Maitland and Sammartino's (2015) theoretical model on small world representations, we did not conceptualise the time dimension of the model. Even though our approach organised information in chronological order, timeline insights were beyond the scope of this study. Hence, future work could focus on adding a time dimension to small world representations. By delineating timelines, we could have more understanding of when certain factors become influential, and which configuration of factors represents the tipping point for MNEs to exit conflict-torn countries. To support this next step, researchers could consider Feduzi et al.'s (2022) recent updates to small world representations. An appropriate way to operationalise this would be to use longitudinal case studies that map out a manager's heuristic steps from the inception of violent conflict to the day the decision between exit or stay was made.
Finally, we acknowledge that analogical reasoning (Gavetti & Levinthal, 2000;Miller & Lin, 2015) can play a role in the construction of managerial mental representations, but we do not activate this reasoning in the analysis of our respondents' small world representations. Recognising that analogical reasoning and experience can play a dominant role in discovery and evaluation heuristics with impacts on 'informational advantages' (Buckley et al., 2016) and decision rules (e. g., searching rules, stopping rules, and decision rules), we recommend that future research could focus on categorising managers according to experience-based metrics, and then constructing category-orientated small world representations. This would facilitate a comparison analysis to identify whether there are differences, what the differences are, and whether experienced managers will construct richer small world representations. This could also open the door to making stronger interconnections between small world representations and logic from fast-and-frugal heuristics.