INTRODUCTION

Decisions on when and how to exit from conflict zones have become increasingly salient for MNEs due to the worldwide surge in violent conflict. In 2016, more countries experienced political conflict – including interstate and civil wars, as well as terrorism – than at any time in the previous 30 years (United Nations & World Bank, 2018). By 2030, more than half of the world’s poor are predicted to live in conflict-afflicted countries (pp. xvii). The costs of violent conflict are enormous: for example, worldwide economic costs were estimated to be 10.5% of global GDP in 2019, rising to 36.4% of GDP for the 10 most politically violent countries (Institute for Economics and Peace, 2021). How MNEs extricate business operations from conflict zones is therefore an important topic of study.

International business (IB) scholars are now paying significant attention to this research area, chiefly examining the question of whether to exit from host countries suffering from violent conflict, with recent contributions focusing on war (Dai, Eden, & Beamish, 2013, 2017; Eden, 2022), civil unrest (Hiatt & Sine, 2014; Soule, Swaminathan, & Tihanyi, 2014), armed conflict (Oetzel & Getz, 2012; Oh & Oetzel, 2011), and terrorism (Liu & Li, 2020; Liu, Li, Eden, & Lyles, 2022; Oh & Oetzel, 2011). However, without exception, these studies treat foreign exit as the outcome of interest, even as the antecedents of foreign exit from conflict zones may affect the timing and mode of exit in different ways.

We argue that foreign exit deserves scrutiny for its strategic components, especially in the context of violent conflicts, when firm deliberations on when and how to exit may not be “business as usual.” The purpose of this note is thus to explore strategic interdependencies in exit timing and mode, using a sample of MNEs that have chosen to exit from conflict zones. We address a two-part research question: How are the exit timing and exit mode of a foreign firm operating in a conflict-afflicted country affected by its own and peers’ experiences, and how does the ordering of the firm’s two decisions affect its choices?

To address this research question, we draw on behavioral theory, which accounts for adjustments of goals in response to relevant changes in the environment (Cyert & March, 1963). The tenets of behavioral theory make it suitable for examining firm exit from conflict zones, where firms may scarcely exhibit even “near-optimal reactions to missing information or limited information processing capabilities” (Surdu, Greve, & Benito, 2021: 1048). In behavioral theory, firms are said to satisfice instead of optimize (March & Simon, 1958) – an apt assumption given the disruptive nature of conflicts – and carry out problemistic search for solutions (e.g., market exit) when performance falls below an aspiration level, i.e., “the smallest outcome deemed satisfactory by a decision-maker” (Schneider, 1992: 1053). Problemistic search, triggered when a problem occurs and concluded when a solution is identified, is further oriented towards matching strategies to contexts, e.g., conflict zones. Finally, an important logic in behavioral theory is that firms will imitate peers, given uncertainty to economize on search costs.

Our study makes several contributions to the literature. First, we extend research on foreign exit by not only examining antecedents of timing and mode but also showing the importance of the ordering of these decisions. While exit timing and mode have received separate study (e.g., Mata & Portugal, 2000; Rangan, 1998), our work is, to the best of our knowledge, the first to treat these as interdependent decisions. In doing so, our analyses further reveal boundary conditions for satisficing as a behavioral mechanism, and reconcile optimizing and satisficing behaviors. Second, we advance behavioral theory by building on the problemistic search and vicarious learning literatures to examine how a firm’s own and peers’ experiences affect, respectively, its exit timing and mode. Third, our work furthers research on MNE exit in response to conflict by providing a fine-grained analysis of when and how such exits occur.

LITERATURE REVIEW

Research on foreign exit typically views exit and survival as “two sides of the same coin” (Coudounaris, Orero-Blat, & Rodríguez-García, 2020). However, when modeled as a simple “stay versus go” proposition, the timing and mode subdecisions intrinsic to the exit decision are obscured. Moreover, there is “not yet sufficient work on how firms choose their goals” (Surdu et al., 2021: 1056); that is, we know little about why firms would consider exit timing before mode, or vice versa, even though exit timing and mode are known to affect each other (Moschieri & Mair, 2017), and have antecedents distinct from those of exit per se (Balcaen, Manigart, & Ooghe, 2011). However, research has been conducted on the interrelatedness of foreign entry timing and mode, dating back to Buckley and Casson (1981) and Hirsch (1976). More recent studies on the timing and mode of foreign entry (Fisch, 2008; Gaba, Pan, & Ungson, 2002; Isobe, Makino, & Montgomery, 2000; Pennings & Sleuwaegen, 2004; Ursacki & Vertinsky, 1992) suggest that these decisions may also be related in the case of foreign exit.

In terms of exit timing, scholars have pointed to time-sensitive opportunities in the home country as a determinant (Iurkov & Benito, 2020), citing the timely divestment of assets abroad as a precondition for pursuing opportunities back home. In addition to “pull” pressures, “push” factors due to competition can induce firms to divest foreign operations earlier rather than later (Hutzschenreuter & Gröne, 2009). Whereas investment (entry) timing depends on certain asset attributes (Rivoli & Salorio, 1996), research on divestment (exit) timing is less consistent, e.g., host countries with similar (Belderbos & Zou, 2009) and dissimilar macroeconomic conditions (Rangan, 1998) have both been found to engender earlier exits.

Exit modes, in contrast, have “seldom been examined” (Arte & Larimo, 2019: 14), although there is “general awareness… of the multiplicity of exit routes” (Cefis, Bettinelli, Coad, & Marsili, 2022: 440). To date, studies of exit mode have primarily been theoretical (Benito & Welch, 1997), or consist of cases (Vissak & Francioni, 2013; Vissak, Francioni, & Freeman, 2020) which typically focus on re-entry as the unit of analysis (Javalgi, Deligonul, Dixit, & Cavusgil, 2011), with mentions of partial exit (Belderbos & Zou, 2007) and whole exit (Benito & Welch, 1997; Surdu, Mellahi, Glaister, & Nardella, 2018; Welch & Welch, 2009). An exception is Mata and Portugal (2000), who found that greenfield subsidiaries are more likely to be shut down than acquired ones, because firms are less inclined to sell assets built from scratch.

THEORY AND HYPOTHESES

In this note, we apply a strategic lens to foreign exit, where the choice to exit earlier versus later must be made somewhat in tandem with the choice to withdraw fully or to retain some operations in the host market. Central to our theory is the interdependent consideration of subdecisions to a decision, where subdecisions are “a set of actions and dynamic factors that begins with the identification of a stimulus for action and ends with the specific commitment to action” (Mintzberg, Raisinghani, & Theoret, 1976: 246).

As illustrated in Figure 1, we examine how two behavioral factors – the firm’s own and its peers’ experience, respectively – affect its decisions regarding exit mode and timing. We take the premise that one decision supersedes the other in criticality, given variations in “the amount of importance allocated to specific goals at a given point in time” (Surdu et al., 2021: 1050). We argue that, when a firm considers (1) its own experience (i.e., its performance), the mode decision supersedes that on timing, while, when observing (2), its peer firms’ experience (i.e., their exit from the conflict zone), the timing decision supersedes that on mode. Our findings shed light on the strategic consequences of decision ordering, where prioritizing one decision over another may preclude outcomes for the strategy under consideration.

Figure 1
figure 1

Behavioral antecedents of foreign exit timing and mode.

Full size image

Own Experience

The role of firm performance in the foreign exit literature is a topic of ongoing debate. In IB research, foreign exit has historically been viewed as a failure (Benito, 1997; McDermott, 2010), which is somewhat surprising given its prevalence, i.e., on average, established firms exit from two countries for every country they enter (Chung, Lee, Beamish, Southam, & Nam, 2013). While MNEs do divest poorly performing subsidiaries (Benito & Welch, 1997; Boddewyn, 1979; Duhaime & Grant, 1984; Iurkov & Benito, 2020), performance may also be unrelated to exit (Soule et al., 2014). Going abroad is so costly and time-consuming that, even if a subsidiary performs poorly, MNEs may not resort to exit (Belderbos & Zou, 2009). This logic is arguably more relevant and has been found to be the case for firms in conflict zones (Dai et al., 2017), given location-bound firm-specific advantages that merit challenges associated with entering in the first place (Rugman & Verbeke, 1992). For a poorly performing firm, however, problemistic search raises questions on host-country presence, leading to either increased commitment (Hui, Gong, Cui, & Jiang, 2021) or market exit (Greve, 1998). While such “modes” of conduct may be assessed in light of possible re-entry (Hadjikhani & Johanson, 1996), we expect that the motivation for assessing how to serve the market (e.g., not at all, by exiting) depends on the extent of poor performance.

We argue that a firm’s own experience (i.e., its performance shortfall) is more likely to affect the mode than the timing of its exit. Problemistic search in response to unsatisfactory performance centers on “playing catch-up,” given resource, knowledge, and network constraints that dictate “what to do” more than “when to do so.” For instance, Kuusela, Keil, and Maula (2017) noted that poorly performing firms carry out resource-freeing actions, such as divestment, and avoid resource-consuming acquisitions. By way of problemistic search, poorly performing firms are prompted to choose an alternate mode to serve – if at all – a host market (Surdu et al., 2021). Upon reaching a decision to exit, the extent of poor performance should be decisive in whether only partial exit occurs. Performance that falls too far below an aspiration level may warrant a more drastic shift in strategy: given the outlays required to correct larger performance gaps (Kuusela et al., 2017), whole exit may be chosen as a means for containing losses (Hui et al., 2021).

These behavioral responses should be exacerbated in a conflict zone, which creates additional hazards for firms that include but are not limited to physical harm and economic loss (Dai et al., 2013; Hiatt & Sine, 2014). A focal firm is thus likely confronted with a crisis situation, where the conflict poses high threats to its goals, allows little time for decision-making, and hinders means for anticipation or planning (Eden, Hermann, & Miller, 2021). If firm performance drops precipitously in such a scenario, a tipping point may be reached, given the daily onslaught of potentially fatal setbacks rendering operations and employees “at risk” (Dai et al., 2017). We therefore expect that firms will choose whole rather than partial exit, in the absence of any economic upside to merit such risk exposure, and posit that:

Hypothesis 1a:

The greater its performance shortfall, the more likely that a focal firm’s mode of exit from a conflict-afflicted country will be whole rather than partial.

When a focal firm prioritizes exit mode over exit timing, and chooses to exit in whole in response to its own poor performance, the costs of completely forfeiting host-country stakes are likely to warrant a wait-and-see approach (Rivoli & Salorio, 1996). With losses to firm reputation (Surdu et al., 2018) and local ties (Welch & Welch, 2009), as well as assets and customers (Surdu, Mellahi, & Glaister, 2019), whole exits represent a last resort, even for poorly performing firms (Belderbos & Zou, 2009), and are therefore likely to occur later rather than earlier. Partial exits, in contrast, leave a greater capacity for re-entry. As less information is needed to make a less irreversible move, e.g., partial exit (Damaraju, Barney, & Makhija, 2015), there is reduced value in waiting, so that partial exits are more likely to occur earlier than later. Since partial exits tend to be less consequential than whole exits (Konara & Ganotakis, 2020), firms may incur greater losses by delaying a partial exit during a conflict, where the costs of delaying exit can quickly outweigh those of exiting (and re-entry should the conflict subside). We therefore posit that:

Hypothesis 1b:

When exit mode is prioritized over timing, a focal firm’s whole exit from a conflict-afflicted country will more likely occur later and its partial exit earlier.

Peer Firms’ Experience

Behavioral theory emphasizes vicarious learning as a mechanism underlying firm strategy (Levitt & March, 1988). In conflict zones, where firms face similar predicaments, vicarious learning may emerge as a particularly salient mechanism for shaping decisions. When unsure about the right course of action, firms tend to resort to observing the actions of other firms (Rao, Greve, & Davis, 2001), and adopt actions adopted by peers in the same situation. Especially in the case of extreme uncertainty, as in conflicts, imitation is trusted to impart a better outcome than acting alone (Surdu et al., 2021). Yet, conflicts can render firms equally uninformed, thereby reducing their reliance on each other (Liu & Li, 2020) for decisions that warrant more alertness in the search process (Verbeke & Greidanus, 2009), e.g., for a suitable buyer in the event of exit. Thus, even as peer exits may incite designs on exiting, the act of exit – via partial divestiture or shutting down an entire outfit – should depend more on a firm’s own experience.

We argue that, while a firm’s own experience affects how exit occurs, its peers’ experience (i.e., peer exits) affects when, or the urgency with which, it chooses to exit. Theories of foreign entry timing emphasize the need to react swiftly to peers’ moves (Flowers, 1976; Yu & Ito, 1988). Like foreign entry, foreign exit is highly visible and critical for a firm’s strategic position (Delios, Gaur, & Makino, 2008), making imitation tenable and (boundedly, at least) reliable. While vicarious learning may not occur in all firms – given variations in local experience (Kim, Delios, & Xu, 2010) and the perceived credibility of peers (Liu & Li, 2020) – the speed with which it occurs in firms should increase as the number of peer exits increases. Ignoring the actions of peer firms in conflict zones can elicit losses that can strategically (Dai et al., 2017) and reputationally (Oetzel & Getz, 2012) cripple operations in the host market and beyond.

Especially in violent conflicts, the behavior of peer firms represents an influential source of information (Henisz & Delios, 2004). As radical disruptions, violent conflicts are known to create crisis situations that challenge the limits of incremental reasoning (Hadjikhani, 2000), and limit the time available for decision-making (Eden et al., 2021). Like other entities, firms may be unclear about the trajectories of a conflict, given the bounded reliability (Verbeke & Greidanus, 2009) of media sources and the time-consuming nature of learning under uncertainty (Johanson & Vahlne, 1990). For firms that operate “without any perception about what to react to” (Figueira-de-Lemos & Hadjikhani, 2014: 336), vicarious learning may be a less cumbersome means of comprehending the environment (Kim & Miner, 2007). In such settings, i.e., volatile, uncertain, complex, and ambiguous (van Tulder, Jankowska, & Verbeke, 2019), we expect firms to be reactive and to only consider exiting when peer firms do so. Since unfavorable changes can trigger decreases in foreign commitment with the same urgency as the speed of knowledge loss about one’s environs (Benito & Welch, 1997), we conjecture that:

Hypothesis 2a:

The greater the number of peer firms’ exits from a conflict-afflicted country, the more likely that a focal firm will exit earlier rather than later.

When a firm prioritizes exit timing over exit mode as a response to peer firms’ exits, and chooses to follow their peers by exiting earlier, we argue that its earlier exit is likely a whole exit. As the literature on competitive behavior shows, it is firms’ visible actions that resolve the intrinsic uncertainty underlying any strategic choice (Ethiraj & Zhu, 2008), and prompt fast responses by other firms (Giachetti, Lampel, & Pira, 2017). Unlike partial exits, whole exits entail job losses, and are thus more likely to be highly visible (Richbell & Watts, 2000). Friebel and Heinz (2014), for example, found that plant closures by foreign firms attracted significant media attention. Partial exits, in comparison, may be better positioned to “fly under the radar” and so avoid public scrutiny (Puck, Rogers, & Mohr, 2013). Since time and attention are scarce resources (March & Olsen, 1976), fast behavioral responses have long been predicted to occur only when moves are visible (MacMillan, McCaffery, & Van Wijk, 1985) and draw external attention (Weick, 1976). Since a firm would have little reason to take note of – much less follow – peer firms’ exits, unless such exits were detectable en masse, and whole exits are more detectable than partial exits, we argue that:

Hypothesis 2b:

When exit timing is prioritized over mode, a focal firm’s earlier exit from a conflict-afflicted country will more likely be a whole exit and its later exit a partial exit.

METHODS

Data and Sample

Our hypotheses are tested on a sample of 101 Japanese MNE exits in 11 conflict-afflicted countries between 1991 and 2005. We examine hitherto unexplored relationships, where “a basic lack of knowledge about which variables matter, how they are causally related, etc., often warrants small-N samples” (Jonsson & Foss, 2011: 1083). Our sample contains the exit events of all Japanese MNE subsidiaries in conflict-afflicted countries worldwide during the sampling period, and is thus small but highly representative. Firm data drawn from the Nikkei Economic Electronic Databank and Toyo Keizai Japanese Overseas Investments Dataset were matched with data on conflicts from the Armed Conflict and Battle-Related Deaths datasets of the Uppsala Conflict Data Program. Only conflicts involving government forces and generating at least 1,000 battle-related deaths in any given year were included (Sambanis, 2001). Country-level controls were compiled from the World Development Indicators.

Measures

Exit mode

We treat exit mode as it relates to two types of de-commitment that can vary in scale and finality: (1) closure – whole exit – where a subsidiary ceases operation in the host country, and (2) divestiture – partial exit – where a remnant of the subsidiary’s operations remains in the host country, since MNEs are known to maintain a toehold of functions to facilitate re-entry into a country (Hadjikhani & Johanson, 1996). In line with prior work (Mata & Portugal, 2000; Soule et al., 2014), we identify whole exits by verifying the year that an MNE ceases to report data and partial exits by the first year that foreign capital participation falls below 10% from a prior year (percentages are used to control for size bias). Exit mode is an indicator variable with a value of 1 for whole exit and 0 for partial exit.

Exit timing

Drawing on foreign re-entry research (Welch & Welch, 2009), we treat timing as continuous and delimited in years, rather than in bin1ary terms that impose arbitrary “early” versus “late” designations. Exit timing is defined as the year of either partial or whole exit – whichever comes first – minus the year that the conflict started, as specified in the UCDP database. For example, if conflict breaks out in a country in 1997 and an MNE exits in 1999, the variable assumes a value of 2. Smaller (larger) values reflect earlier (later) exit. Where partial exit is followed by whole exit, only the first instance of exit, i.e., the partial exit, is considered, to avoid any confounding effects (Damaraju et al., 2015).

Performance shortfall

We follow prior research on firm performance and foreign exit by operationalizing performance shortfall relative to a firm-specific threshold rather than financial outcomes, using an ordinal measure of managerial satisfaction with performance containing three categories: 1 for gain, 2 for break-even (baseline), and 3 for loss (performance shortfall) (Tan & Sousa, 2019).

Peer exits

This variable is measured as the number of home-country firms in the same industry as a focal firm that exit from the host country in a prior year (Greve, 1995), since firms learn best from firms from the same home country and the same industry (Kim et al., 2010).

We control for several variables that could affect foreign exit timing and mode. Controls are included for subsidiary age, a key determinant of foreign exit (Mata & Freitas, 2012), and subsidiary size, measured as the natural logarithm of the number of employees (Mata & Portugal, 2000). We use two variables to control for the prospect that firms with strong ties to a country will be less likely to exit (Johanson & Vahlne, 2009). The first, host-country commitment, is measured as a firm’s number of subsidiaries in the host country as a percentage of its subsidiaries worldwide (Mohr, Batsakis, & Stone, 2018). The second is a count of local partners, for the decision-making power they hold in foreign exits (Dhanaraj & Beamish, 2004). Since entry mode is considered a factor in foreign exit (Benito, 1997), we control for entry mode, equal to 1 for greenfield entry and 0 for entry via acquisition. The timing and mode of exit from conflict-afflicted countries may also be influenced by a firm’s experience with conflict zones elsewhere (Delios & Henisz, 2003); we therefore control for conflict experience using an entropy measure of the dispersion of the MNE’s operations in conflict-afflicted countries, calculated as:

$$D_{jt} = \mathop \sum \limits_{j = 1}^{j} \left[ {S_{j} { } \times {\text{ ln}}\left( {\frac{1}{{S_{j} }}} \right)} \right]$$
(1)

where Sj is the ratio of an MNE’s number of subsidiaries in conflict-afflicted country j to its total number of foreign subsidiaries, and ln(1/Sj) is the weight given to each conflict-afflicted country j (Kim, 1989).

Lastly, we control for macrofactors relevant to our context of study. Since conflicts in recent history increase the risk of new conflicts, we account for prior conflicts by including a count of host-country conflicts for the last 40 years (Sambanis, 2001). Per capita GDP is used to control for market size and dummy variables to control for investment purpose, namely extraction, manufacturing, and exporting.

Methodology

To see the influence of exit mode on exit timing and vice versa, we estimate two instrumental variable (IV) simultaneous equations in which exit mode is an independent variable in an equation for exit timing, and vice versa. Since unobserved variables that affect both timing and mode may cause the mode variable to correlate with the error term in the timing equation, and vice versa, we depict the reciprocal relationship between these endogenous variables with non-recursive IV models (Guo & Fraser, 2009).

First, since conventional 2SLS models cannot account for binary instrumented variables, i.e., exit mode, we estimate exit timing with a treatment effects model (Cerulli, 2012). Exit mode, a binary variable for the treatment condition in which 1 denotes whole exit and 0 partial exit, is entered into a regression; an outcome variable, exit timing, is observed for both conditions. Second, to estimate exit mode as a binary dependent variable with exit timing as a continuous instrumented variable, we employ an IV probit model (Guo & Fraser, 2009). To identify the equations and to correct for endogeneity, we use explanatory variables – performance shortfall and peer exits – to instrument for exit mode and timing (Greene, 2003).

Our criteria for selecting “strong” and “relevant” instruments are that they would explain the independent variable, but also be “independent” and “exogenous”, i.e., theoretically unrelated to the dependent variable and the error term (Stock, Wright, & Yogo, 2002). Since we argue that performance shortfalls would affect exit mode (Hypothesis 1a), and that peer exits would affect exit timing (Hypothesis 2a), these two instruments also represent independent variables in first-stage equations, where exit mode and exit timing are dependent variables, respectively. In second-stage IV equations, exit mode and exit timing are in turn instrumented variables that estimate the effect of exit mode on exit timing (Hypothesis 1b) and vice versa (Hypothesis 2b).

Instrumental variables must correlate with the second-stage dependent variable only via their correlation with the first-stage variable (Bascle, 2008). Prior research suggests that performance shortfalls are correlated with exit mode (Hui et al., 2021; Mohr et al., 2018) and peer exits with exit timing (Dai et al., 2013; Soule et al., 2014), whereas there are no a priori reasons to expect their correlation with exit timing and exit mode, respectively. A correlation analysis reveals that performance shortfalls and peer exits correlate, respectively, with exit mode and exit timing (r = 0.09 and r = − 0.45; p = 0.001), but not with exit timing and exit mode (r = − 0.08 and r = 0.04; n.s.). Additionally, instruments should not be related to unobserved variables, e.g., firm outlook on the conflict, that potentially determine both the independent and dependent variables (Bascle, 2008). There is no reason to expect a firm’s performance or its peers’ exits be predicated on its attitude towards the conflict per se. The absence of any significant correlations between the instruments and the error term suggests that our instruments satisfy this requirement.

Finally, we test for both under- and over-identification and for weak instruments using post-estimation tests (Baum, Schaffer, & Stillman, 2007). The Anderson canonical correlation LR test rejects the null hypothesis that the model is under-identified and the instruments are not relevant. The Cragg–Donald statistic (21.56, p = 0.000) is higher than the critical value of 7.77 suggested by Stock and Yogo (2005). We perform the Sargan test for over-identifying restrictions (Wooldridge, 2002). The insignificant results (χ2 = 0.720, p = 0.396) confirm the validity of the instruments. To test the exogeneity of the IVs, we conduct the Hansen over-identification test. Since the J statistic (0.623, p = 0.430) is not significant, we do not reject the null hypothesis that the IVs are exogenous. We further use the Anderson-Rubin and Stock-Wright tests to check the IVs’ explanatory power. The Anderson-Rubin (χ2 = 7.85, p = 0.005) and Stock-Wright (χ2 = 7.37, p = 0.007) statistics are both significant. We thus reject the null hypothesis for weak IVs.

Results

Descriptive statistics and correlations are shown in Table 1. Correlations and VIFs are well below standard cut-off thresholds of 0.7 and 10, with VIFs ranging between 1.06 and 1.62. Log-likelihood tests show that the exit timing and mode coefficients are statistically distinct. As shown in Table 2, the instrumental variables exhibit statistical significance with the instrumented variables (β = − 0.602, p = 0.000; β = − 0.388, p = 0.000). Turning to the main results, we find that Hypothesis 1a, in which performance shortfalls are argued to induce whole rather than partial exits, is not supported (β = − 1.606, p = 0.015). Our predicted effect for exit mode on exit timing in Hypothesis 1b is supported, with whole exits taking place later and partial exits earlier (β = 0.599, p = 0.005). Hypothesis 2a is also supported, as we find that an increase in the number of peer exits promotes earlier exit (β = − 0.417, p = 0.000). Lastly, we find support for our prediction in Hypothesis 2b that earlier exits tend to be whole exits, while later exits likely occur as partial exits (β = − 0.058, p = 0.010).

Table 1. Descriptive statistics and correlations
Full size table
Table 2 Instrumental variable-two stage least squares (IV-2SLS) models
Full size table

We consider the practical significance of our results by examining effect sizes, i.e., treatment effects coefficients and IV probit marginal effects, the latter conditional on the distribution of variables being what they are in the sample (Bowman & Wiersema, 2004). For firms experiencing a performance shortfall, the probability of partial exit increases by 24.2%. When exit mode is prioritized over exit timing, whole exits take 0.6 years – roughly 7 months – longer to carry out than partial exits. For every exit by a peer firm, exit timing is speeded up by 0.417 years, or approximately 5 months, and for every year that a firm exits earlier, the probability of whole (partial) exit is increased (reduced) by 1.27%.

Discussion

Foreign exit has been pronounced a highly complicated strategic action that requires substantial managerial attention and time to implement (Boddewyn, 1983; McDermott, 2010). We apply a behavioral lens in the context of violent conflict to unveil hitherto unexamined relationships that both defend and refute this notion. Namely, while foreign exit does undergo strategizing when the decision is prompted by a firm’s own experience, the act of leaving a country may be satisficed in a less-than-strategic manner when motivated by the experience of peer firms. Our treatment of the foreign exit decision thus reconciles Simon’s (1955) satisficing with “calculative/optimizing behavior” (Verbeke & Greidanus, 2009: 1479).

According to the satisficing principle, problemistic search stops once a satisfactory solution is found (Cyert & March, 1963). By disentangling the exit timing and mode decisions, our analysis of a multidecision problem expands boundary conditions for the satisficing mechanism, which “kicks in” once one subdecision is resolved (Moliterno & Wiersema, 2007), in effect “pre-determining” the other. In their study of subdecisions, Witte, Joost, and Thimm (1972: 180) noted that “human beings cannot gather information without in some way simultaneously developing alternatives. They cannot avoid evaluating these alternatives immediately, and in doing this they are forced to a decision.” Our findings suggest that, especially in turbulent contexts where decisions are often made in the absence of thorough deliberation, the importance ordering of subdecisions may be critical for shaping strategic outcomes, and therefore its mastery is the basis for a core firm-specific advantage in the MNE (Verbeke, 2013).

This insight points to a seldom recognized distinction between permutations in decision-making order and combinations of strategic outcomes. Evidently, permutations in how decisions are importance-ordered have the capacity to produce different combinations of strategic outcomes. As such, our study offers insights for managers by highlighting the need to pay attention to what one pays attention to, given the possibility that certain strategic outcomes may be closed off. We find that, if a firm prioritizes exit timing over mode, a whole (partial) exit is more likely to occur later (earlier). Prioritizing timing over mode, however, renders an earlier (later) exit more likely to be whole (partial). Thus, a change in decision ordering can reverse a firm’s preferred choices. The particular grouping of exit timing and mode depends on whether the mode decision is driving the timing decision, or vice versa. Prioritization of the when versus the how questions, in turn, depends on the salience of a firm’s own experience versus that of its peers.

In terms of a firm’s own experience, poor performance leads to partial, not whole, exit (Hypothesis 1a). This result, contrary to our prediction in Hypothesis 1a, may be explained by Konara and Ganotakis’s (2020) theory that poor performance is only related to certain types of foreign exit. Especially for Japanese firms that have long invested abroad for resource-seeking purposes (Buckley, 2009), reductions in profitability may scarcely warrant a complete reversal of strategy, i.e., whole exit. Since Japanese firms tend to internationalize via joint ventures more than firms from other countries (Beamish, Delios, & Lecraw, 1997), partial exits may be more feasible with local partners as ready buyers. That a preference for partial exits may be specific to Japanese MNEs also finds support in Hennart, Kim, and Zeng’s (1998) study of Japanese MNEs, in which they showed that larger performance shortfalls may result in partial as opposed to whole exit, because a very unprofitable operation may only be sold piecemeal rather than as an entity.

Our result may, nonetheless, be generalizable, as firms can experience performance shortfalls due to a host of reasons (Surdu et al., 2019), many of which may not warrant closure. Indeed, not all firm closures are due to poor performance, with almost half even being profitable (Duhaime & Grant, 1984). While it is widely held that foreign exit is affected by firm performance (Boddewyn, 1979, 1983), our result points to the mode of such exit being shaped by a comparison of firm performance with opportunity costs (Iurkov & Benito, 2020), and corroborates Surdu et al. (2018) that poor performance prior to foreign exit induces re-entry, a goal of firms that exit partially (Hadjikhani & Johanson, 1996). Even in conflicts, firms “hang on” to the extent possible, bar any physical harm (Dai et al., 2013) or explicit sanctions (Soule et al., 2014). Firms with large performance shortfalls may also face tight supervision and scrutiny from stakeholders, making it harder for closures to be approved (Desai, 2016). While performance shortfalls are often associated with radical change (Labianca, Fairbank, Andrevski, & Parzen, 2009), our fine-grained analyses in the context of foreign exit instead uphold Surdu et al.’s (2021: 1051) assertion that “problemistic search usually uncovers modifications of current activities as solutions.”

With respect to others’ experience (Hypothesis 2a), we find that an increase in peer exits is positively associated with earlier exit. In a departure from extant research (Henisz & Delios, 2004; Liu & Li, 2020), we disentangle subdecisions to foreign exit and show that peer exits do not, in fact, affect the mode of exit. Our study thus contributes to a growing body of literature on interfirm effects in foreign exit amid war (Dai et al., 2013; Soule et al., 2014). While the literature deems it a low-risk strategy to follow peers (Abrahamson & Rosenkopf, 1993; Lieberman & Asaba, 2006), our work points to second-order effects of imitation on decision-making: specifically, prioritizing the timing aspect of a strategy can limit its mode of execution (Hypothesis 2b). In the context of foreign exit, for instance, exiting earlier on (thus necessitating whole exit) may mean ceding market share in the host country to competitors. Studying how interrelated decisions diffuse can thus reveal new and interesting insights into the dark side of vicarious learning.

Relative to firms from other countries, Japanese firms may be more likely to exit later rather than earlier, making our result for Hypothesis 2a a conservative estimate. Japanese firms tend to refrain from exiting in the first few years of foreign operations (Delios & Beamish, 2001), given the long-term orientation and willingness of MNE parents to allow foreign subsidiaries a 5- to 10-year window to “get set up” abroad (Tachiki, 1999). Japan itself is moreover highly susceptible to natural disasters of various types and severities, such that a break-out of violent conflict elsewhere would perhaps pose less impact on Japanese MNEs than on counterparts from other countries that do not routinely undergo such experiential learning.

Given that “a central part of [behavioral] research is to correctly understand the firm’s goals” (Surdu et al., 2021: 1059), we surmise that the result that earlier exits tend to be in whole (Hypothesis 2b) may stem from an impetus to cut losses than to salvage investments. The urgency prompted by peer exits may lead firms to exit in full by abandoning assets early on, to prevent “throwing good money after bad.” For example, to escape conflict in 1984, Chevron abandoned assets in Sudan that included oil rigs, airplanes, buildings, and trucks (pers. comm.). The higher write-off value of such assets makes whole exits more apt than partial exits, as the latter can incur delays due to the need to decide which assets to divest (Balcaen et al., 2011). Our finding is especially appropriate for Japanese MNEs, as they typically do not invest abroad without obtaining political risk insurance from Nippon Export and Investment Insurance, which in turn does not indemnify losses if operations are continued even partially, requiring a “full inability to operate” in order to receive coverage for “political risk materialization”, e.g., conflict (Papanastasiou, 2021: 171).

Our results for the effect of exit mode on exit timing and vice versa signal the role of being proactive versus reactive in decision-making. Evidently, being proactive about one renders the other reactive. During a conflict, when the search costs of finding a buyer for assets are greater than usual (Collier & Goderis, 2009), firms that choose to exit in whole (not reactively, e.g., by writing off assets) may be constrained to deferring exit. With acquisition activity made scarce by uncertainty, even firms with buyers on hand can face roadblocks in making timely asset transfers. The findings suggest that, when less able to salvage host-country assets, firms may not act quickly to reverse investments, i.e., exit early.

Conversely, being proactive on exit timing means having less leeway for exit mode: partial exits are made possible only by postponing exit, which may be less tenable during a conflict. As a case in point, in response to Indonesia’s 1997 conflict between government militia and separatist Free Aceh Movement forces, the Japanese liquid natural gas (LNG) firm JILCO exited partially from the Arun Natural Gas Liquefaction Company (its joint venture with Mobil and Pertamina, the Indonesian state-owned oil firm) in March 2001. From its location in Aceh province, which accounted for a third of Indonesia’s gas production, JILCO had shipped 80% of its LNG to Japan (Iyer & Mitchell, 2007). To gauge Japan’s dependence on Indonesia (Japan’s largest supplier), Tokyo Electric Power Company, which had a monopoly over regional electricity markets, lost all access to LNG following JILCO’s exit, leading to JILCO re-entering Indonesia in July 2001 (Harrison, 2002). Our finding is in line with extant theory that postponing exit may lead to partial exit, as firms tend to increase their commitment to the host country over time (Surdu et al., 2019). Especially for Japanese MNEs that are reputed to act in alignment with Japan’s national interests, the alternative of whole exit in terminating commitments (Benito, 2005; Benito & Welch, 1997) may not be practical in industries, e.g., energy, considered critical to national security.

Our study has some limitations. First, we do not study decision-making using qualitative interview or survey data. Nonetheless, our instrumental variables analyses of a sample of realized exits allow us to test novel theory, as “intentions that do not materialize later on cannot be considered as decisions” (Damaraju et al., 2015: 733). In addition, our use of archival data removes risks of respondent and common method biases. We further choose instrumental variables on the basis of theory – exit timing and mode are theorized for concurrent resolution (Payne, 2006) – as opposed to simply lagging variables, a method often adopted due to data limitations. Second, our focus on own and others’ experience neglects behavioral logics from elsewhere within the MNE, e.g., head office (Verbeke & Greidanus, 2009). As such, further study is needed on multifaceted decision inputs to foreign exit. Furthermore, since MNEs from different countries may have varying loss tolerance levels, and reliance on peer firms that shape when and how they exit, future research could use multiple home countries to gauge the generalizability of our findings. Lastly, because not all foreign exits may necessarily map neatly to early partial exits or to late whole exits, our work should be viewed as a first step toward developing a typology of foreign exit.

To the best of our knowledge, our study represents the first attempt in the IB literature to examine how these strategic subdecisions are reciprocally resolved. While we focus on foreign exit timing and mode decisions, future research could study other strategic decisions that involve subdecisions, examining their antecedents and the importance of decision ordering. Researchers could also examine foreign exit timing and mode interdependencies in different contexts, e.g., comparing “business as usual” with crisis situations. In extending behavioral theory to the study of exit, we show that interrelated subdecisions cannot be considered separately from one another without rendering a biased analysis of an overarching strategy. In particular, the ordering of these decisions as a function of behavioral factors constitutes a strategic element that can in itself produce as well as exclude certain strategic outcomes.

Note

  1. 1

    We conducted robustness tests using a 20% cutoff in delineating partial exits for the exit mode variable, and the main results remained substantively the same.