Sending Expats or Hiring Locals ? The Impact of Communication Barriers on Foreign Subsidiary CEO Staffing

Communication between headquarters (HQs) and foreign subsidiaries of multinational corporations (MNCs) is crucial for coordination, control, and knowledge transfer, but language barriers and geographic distance impede this exchange. Hypothesizing that MNCs react to these hurdles by appointing subsidiary top managers with adequate communication skills, we investigate how the native language barrier, foreign language barrier, and geographic distance between HQs and a foreign subsidiary influence the choice between parent and host/third country nationals as subsidiary CEOs. Testing our hypotheses on a sample of 101 staffing decisions made by German firms in 33 countries, we find a negligible effect of the native language barrier, but establish that a foreign language barrier enhances and higher geographic distance lowers firms’ propensity to staff subsidiary CEO positions with parent country nationals. An MNC’s international experience was found to moderate these relationships.


INTRODUCTION
International business scholars agree that effective communication is crucial for MNC success (Tran & Tran, 2016;Louhiala-Salminen, 2010) and particularly for productive HQ-subsidiary relationships (Peltokorpi, 2015a).HQs need to communicate effectively with their globally dispersed subunits to coordinate, control, and align them with the corporate strategy, to transfer and to exchange knowledge (Slangen, 2011).If subsidiaries are dispersed across the globe, however, managers face severe communication barriers.Language differences arise because their units are located in countries with different official languages and because knowledge of English as the lingua franca of business is distributed unevenly across the world (ETS, 2016).
Geographic distance additionally limits communication by restricting the use of certain communication channels (Nijkamp et al., 1990).
An important strategic measure to overcome these communication barriers is global staffing.MNCs frequently appoint expatriatesparent country nationals (PCNs) assigned to spend a significant period of time in the host country (Suutari & Brewster, 2001)as the CEOs of their foreign subsidiaries.These individuals span boundaries between HQs and subsidiary by leveraging their ties to HQs (Gupta & Govindarajan, 2000), their socialization with HQs' values (Colakoglu & Caligiuri, 2008), and their familiarity with HQs' processes and practices (Fang et al., 2010).Being skilled in the parent country language, they also act as "language nodes" (Marschan- Piekkari et al., 1999a) or "linking pins" (Harzing et al., 2011) between the HQs and their local subunits (Torbiörn, 1994).
Recent studies have suggested that communication barriers influence staffing decisions (Peltokorpi & Vaara, 2014;Peltokorpi, 2017) and expatriate assignments in foreign subsidiaries (Selmer & Lauring, 2015;Zhang & Harzing, 2016).Whereas these publications have emphasized the general importance of communication for the selection of candidates, they have not distinguished between different types of communication barriers.
Aiming at a fine-grained, communication-based theory of global staffing, we investigate the impact of different communication hurdles on MNCs' propensity to staff the CEO position of newly established foreign subsidiaries with PCN expatriates as opposed to host country nationals (HCNs) or third country nationals (TCNs).Based on Slangen (2011), we distinguish two types of linguistic communication barriers.First, when home and host country languages differ and employees do not master each other's native tongue, the resulting native language barrier hampers the decoding of messages and renders the exchange of information less efficient (Dow & Karunaratna, 2006;Slangen, 2011).Second, when MNCs designate a lingua franca for inter-unit communication, i.e. "a common language different from the parties' native language, very often English" (Cuypers et al., 2015: 430), and when employees do not master this language sufficiently, a foreign language barrier ensues (Slangen, 2011).In addition, we consider geographic distance, which impedes communication due to long travel times and potential time zone differences (Harzing & Noorderhaven, 2006a;Slangen, 2011;Zaheer, 2000).
The magnitude to which these barriers influence staffing decisions may be influenced by the MNC's general international experience, i.e., the knowledge the corporation has gained by managing foreign operations (Padmanabhan & Cho, 1999).Through international expansion, firms accumulate know-how that serves as a bridging factor between distant home and host countries (Child et al., 2002), develop routines for easing the transfer of knowledge between HQs and subsidiary (Dow & Larimo, 2011), and generally refine their international human resource management strategies (Taylor et al., 1996).We therefore investigate how MNCs' use of expatriates in the presence of communication barriers changes with different levels of international experience.We test our hypotheses based on a sample of 101 staffing decisions made by German firms in 33 countries.
Our study contributes to global staffing research by disentangling the influence of different distance measures on subsidiary CEO staffing and by expanding the contingency perspective on expatriation with international experience as a new moderator.We furthermore advance the active debate on language diversity in MNCs by highlighting the paramount role of English as "the language of global success" (Neeley, 2017) and by showing its importance for subsidiary CEO staffing.Our study informs the ongoing debate about whether distance still matters in today's interconnected world in a dialectical way.On the one hand, we show that distance does matter for communication, and, hence, for global staffing.On the other hand, we also demonstrate that its importance declines as MNCs gain increasing international experience.

THEORETICAL DEVELOPMENT OF THE RESEARCH MODEL
Effective HQ-subsidiary communication is a prerequisite for reciprocal knowledge transfer (Peltokorpi, 2017;Reiche et al., 2015) and for the coordination and control of foreign affiliates (Björkman & Piekkari, 2009).Knowledge-related HQ-subsidiary communication has generally risen in importance over time (Harzing & Noorderhaven, 2006b).Knowledge outflows from the subsidiary are particularly important for so-called "global innovator" subsidiaries, which are established to develop innovations for the whole MNC, whereas knowledge inflows to the subsidiary are crucial for "implementors", which are expected to put HQs directives into practice (Gupta & Govindarajan, 1991).Knowledge needs to flow in both directions and HQsubsidiary communication thus has to be particularly intense for "integrated players", which are both creators and receivers of innovations (Gupta & Govindarajan, 1991)."Global innovators" and "integrated players" are additionally subject to strong behaviour control, i.e. surveillance of subsidiary decisions and actions (Gupta & Govindarajan, 1991), which requires intense control-related communication.
However, an active stream of research has established that communication barriers hamper these communication processes (Harzing & Feely, 2008;Piekkari et al., 2014).We investigate the staffing implications of communication barriers, which we conceptualize as linguistic and geographic constraints that complicate, impede, or block the transmission of messages (Slangen, 2011).Considering two different language barriers along with geographic distance, our study provides a nuanced communication perspective on global staffing.
Consistent with an understanding of language as "the basic means of communication in organizations [and] the basis for knowledge creation" (Vaara et al., 2005: 595), language has come to be considered "a social phenomenon, which affects every aspect of international management" (Piekkari et al., 2014;Lecomte et al., 2018: 2).Consequently, language diversity in international business has gained fast growing attention in recent years (for overviews see Piekkari et al., 2014;Tenzer et al., 2017).Specifically regarding HQ-subsidiary communication, the lack of a shared language was associated with slower decision making (Harzing et al., 2011), uncertainty, anxiety, mistrust (Harzing & Feely, 2008), misunderstandings, conflict, and parallel information networks (Harzing & Pudelko, 2014).It also affects subsidiaries' tacit knowledge inflows from HQs (Reiche et al., 2015) and leads MNCs to control their overseas units through higher centralization and formalization (Björkman & Piekkari, 2009).Slangen (2011) conceptualizes two kinds of linguistic communication barriers.First, the native language barrier encompasses the dissimilarities between the home and the host country language depending on their closeness in the classification of language families (Grimes & Grimes, 1996) along with the proportion of home and host country nationals being able to speak each other's languages.This barrier is relevant in the context of our study, as similarity of the local languages spoken at HQs and in a foreign subsidiary and employees' proficiency in the respective other tongue facilitates communication to exchange knowledge (Schomaker & Zaheer, 2014).Second, a foreign language barrier exists when HQs rely on a lingua franca like English as a shared medium of communication with foreign subsidiaries (Komori-Glatz, 2018), but managers' proficiency in that language is insufficient (Cuypers et al., 2015).This barrier is crucial for HQ-subsidiary communication given the high prevalence of business English as a lingua franca in the inter-unit communication of today's MNCs (Kankaanranta & Planken, 2010).
In addition, Slangen (2011) considers the geographic distance between HQs and foreign subsidiary as a barrier to communication.This physical distance reduces the opportunity for face-to-face communication due to long travel times and even limits synchronous computermediated exchange due to time differences, rendering coordination (Zaheer, 2000) and knowledge exchange (Harzing & Noorderhaven, 2006a) more difficult.Therefore, extant literature also highlights physical distance between HQs and subsidiary as a source of communication problems (Harzing & Pudelko, 2014).
When selecting candidates for top management positions in a foreign subsidiary, an MNC can choose between PCNs, HCNs and TCNs (Gaur et al., 2007).If HQs want to improve interunit communication, they often decide in favour of PCNs, usually assigned to the subsidiary on an expatriate contract (Harzing, 2001b).Expatriates are familiar with corporate control systems (Scullion, 1994) and share similar social and cultural backgrounds with HQ managers (Ando et al., 2008), which makes them effective boundary spanners between employees in the foreign subsidiary and HQ managers.They are expected to act as communication conduits (Boyacigiller, 1990), who facilitate coordination, control, and knowledge transfer (Belderbos & Heijltjes, 2005;Delios & Björkman, 2000) and increase the richness of formal and informal exchange between HQs and subsidiary (Fang et al., 2010).Creating a common frame of reference for HQs and the foreign subsidiary, expatriates can also ease the flow of knowledge between the two parties (Belderbos & Heijltjes, 2005).
Whereas much of the extant literature focuses on expatriates bridging cultural or institutional barriers (e.g., Gaur et al., 2007;Gong, 2003), only few authors have explicitly studied expatriation from a communication-based perspective (e.g., Selmer & Lauring, 2015;Zhang & Harzing, 2016;Zhang & Peltokorpi, 2016).These studies mostly remain on the individual level, dealing with assignees' adjustment challenges imposed by native language barriers.We take the linguistic perspective on expatriation to the firm level of analysis by developing and testing a communication-based framework explaining MNCs' staffing decisions for subsidiary top management.More specifically, we offer a nuanced perspective on HQ-subsidiary communication by distinguishing the impact of native language barriers, foreign language barriers, and geographic distance on MNCs' propensity to staff a subsidiary's CEO position with a PCN expatriate as opposed to a HCN or TCN.
Consistent with prior literature emphasizing the importance of contextual factors for global staffing (e.g., Harzing, 2001a), we consider international experience as a boundary condition of the focal relationships between the three communication barriers and the MNC's staffing decision.International experience has attracted the attention of global staffing researchers (e.g., Ando et al., 2008), as it develops an MNC's capabilities to cope with dissimilar foreign contexts and institutional environments (Ando, 2011;Ando & Paik, 2013).Consistent with Shen's (2006: 310) understanding of international experience as a "catalytic agent of change", we investigate how an MNC's general international experience moderates the relationship between communication barriers and foreign subsidiary expatriate staffing.

HYPOTHESES Communication Barriers and Staffing
Linguistic communication barriers are likely to influence MNCs' decisions whether to appoint a PCN expatriate or local employee as CEO of a foreign subsidiary.As indicated above, HQsubsidiary relations are affected by a native language barrier if the local languages of HQs and subsidiary differ and if managers find it difficult to understand and learn their counterpart's tongue.This barrier rises with higher linguistic distance between parent and host country language (Chiswick & Miller, 2005), but decreases if a substantial proportion of subsidiary employees speak the HQ's language, and if a number of HQ managers speak the subsidiary's tongue.A high native language barrier raises the costs of communication, as HQs and subsidiaries have to invest more effort until their respective employees understand oral or written messages (Barner-Rasmussen & Björkman, 2005;Luo & Shenkar, 2006).The additional work needed to transfer knowledge, e.g. through translation of documents, delays exchange processes and increases transfer costs (Buckley et al., 2005).Control-related information sharing or cross-functional coordination also require extensive communication in a shared language (Björkman & Piekkari, 2009;Luo & Shenkar, 2006).This communication is hampered if HQs and subsidiary use different languages and particularly if those tongues are very different (Marschan-Piekkari et al., 1999b).
Several studies have shown that PCN expatriates can facilitate this exchange especially in the initial stage after establishing a subsidiary (e.g., Edström & Galbraith, 1977;Harzing, 2001b), possibly due to them sharing a language with HQ managers (Harzing et al., 2011).
Hence, MNCs may appoint a PCN as subsidiary head in order to mitigate the native language barrier between HQs and subsidiary.In sum, we argue: Hypothesis 1.The higher the native language barrier, the higher an MNC's propensity to staff the CEO position in a foreign subsidiary with a PCN.
Besides the multiple local languages spoken in MNCs' globally dispersed units, English is now widely accepted as the lingua franca of business (Komori-Glatz, 2018;Neeley, 2017).If host country managers' or HQ representatives' proficiency in this lingua franca is limited, MNCs are facing a foreign language barrier, which substantially increases the costs of communication (Slangen, 2011).Proficiency in a common working language is important for knowledge transfer (Buckley et al., 2005), as the shared language increases employees' sense of commonality and strengthens informal connections between them (Cuypers et al., 2015;Zhang & Peltokorpi, 2016).Fluency in a common corporate language also supports formal reporting and improves access to documents (Marschan- Piekkari et al., 1999a), thus facilitating communication, coordination, control and cohesion (Aichhorn & Puck, 2017).In contrast, if HQ managers or subsidiary employees are unable to properly speak a common corporate language, a subsidiary is isolated and disconnected (Barner-Rasmussen & Björkman, 2007;Björkman & Piekkari, 2009).
To overcome this barrier, MNCs may again appoint PCNs as subsidiary top managers.
Whereas these managers may themselves face language barriers when communicating with their local colleagues and subordinates (Tenzer & Schuster, 2017), they provide an access point for HQ managers into the subsidiary.Based on this argumentation, we hypothesize: Hypothesis 2. The higher the foreign language barrier, the higher an MNC's propensity to staff the CEO position in a foreign subsidiary with a PCN.HQ-subsidiary communication for knowledge exchange, coordination, and control also becomes more cumbersome with higher geographic distance.The further a subsidiary is located from HQs, the longer travel times between home and host country managers have to face in order to establish face-to-face communication (Slangen, 2011).Computer-mediated communication technologies offer various virtual alternatives to face-to-face communication, which aim to bridge geographic distance (Gilson et al., 2014;Koutsabasis et al., 2012).
However, opportunities for synchronous communication through phone or video conferences are often restricted, since greater geographic distance between HQs and subsidiary often implies time zone differences (Zaheer & Hernandez, 2011).According to media synchronicity theory (Dennis et al., 2008), asynchronous communication through email is useful for conveying large amounts of uncontroversial information, but less apt for negotiating diverging viewpoints in an interactive manner.Therefore, the media choices dictated by large geographic distance are often suboptimal for today's knowledge-intensive and innovation-driven work (Tenzer & Pudelko, 2016).
Whereas geographic distance clearly constitutes communication barriers between an MNC's HQs and foreign subsidiaries, we would argue that the limitations of asynchronous media affect the communication with PCN, TCN, and HCN subsidiary CEOs in similar ways.
Considering this, HQs should have little motivation to assign PCN expatriates, the most expensive employees within an MNC (Scullion & Brewster, 2002;Harzing & Noorderhaven, 2006a).Moreover, with increasing distance between home and host country it becomes more difficult to find PCNs who are willing to relocate to such distant locations (Abdellatif et al., 2010).Instead, HQs are more likely to use indirect measures (e.g., budgets) to coordinate and control subsidiaries, which they can also apply to subsidiaries headed by HCNs or TCNs (Harzing, 2001a).Thus, we hypothesize: Hypothesis 3. The higher the geographic distance barrier, the higher an MNC's propensity to staff the CEO position in a foreign subsidiary with a HCN or TCN.

The Moderating Effect of International Experience
We argued that MNCs are more likely to employ PCN expatriates as subsidiary CEOs in the presence of high native language barriers.However, corporations with prior international experience may be able to reduce such communicative challenges (Lu & Beamish, 2004).As MNCs gather international experience in a diverse set of country markets, they gain extensive practice in establishing links between HQs and subsidiaries across borders (Schuler et al., 1993) and develop specific skills and processes for international knowledge exchange (Gupta & Govindarajan, 2000), which eases the exchange of knowledge even if both parties' proficiency in each other's language is limited.
More importantly, as MNCs gain international experience, they implement planning systems and formalize strategies (Morgan & Katsikeas, 1997).Whereas corporations with limited international experience might follow a more reactive approach by implementing language policies on an ad-hoc basis during international expansion (Welch & Welch, 2015), experienced MNCs usually deal with their multilingualism more systematically.They determine the language(s) used for internal communication and documentation as well as for communication with local stakeholders and establish measures such as language-sensitive recruitment or language training to ensure that employees have the required linguistic competencies (Feely & Harzing, 2003;Peltokorpi, 2015b).Thereby, internationally experienced MNCs alleviate complications resulting from native language barriers, which in turn eases HQs' communication with HCNs or TCNs as subsidiary CEOs.In sum, we argue: Hypothesis 4a.The propensity to staff the subsidiary CEO position with a PCN due to a high native language barrier decreases with rising international experience of the MNC.
If HCNs and TCNs have a low proficiency in English as the lingua franca of business, thereby creating a high foreign language barrier, MNCs are prone to relying on expatriates as communication brokers.However, international experience may reduce the foreign languageinduced propensity to appoint PCN expatriates as subsidiary CEOs.With increasing international experience, MNCs learn how to deal with unfamiliar environments (Kostova & Zaheer, 1999), perceive less uncertainty (Kim et al., 2012), and develop a "cosmopolitan attitude", which grants subsidiaries more autonomy (Myloni et al., 2007(Myloni et al., : 2060) ) and decreases their need for tight control in newly entered markets (Erramilli, 1991).If less coordination-and control-related communication between HQs and subsidiary is required, the challenges of a foreign language will be less pronounced and the costs of uncertainty-induced monitoring decrease even if relying on HCNs or TCNs.Among internationally experienced MNCs, it is therefore less likely that high foreign language barriers lead to an increased use of PCNs in overseas subsidiaries.In sum, we argue: Hypothesis 4b.The propensity to staff the subsidiary CEO position with a PCN due to a high foreign language barrier decreases with rising international experience of the MNC.
When the geographic distance between home and host country is high, we argued that MNCs are more likely to draw on HCNs or TCNs as foreign subsidiary CEOs.Parallel to the case of language barriers, we expect this influence of geographic distance on staffing decisions to decrease as MNCs gain in international experience.As corporations establish an increasing number of foreign subsidiaries in many countries, their foreign operations rise in importance compared to domestic activities.Consequently, MNCs increase foreign direct investment into their overseas subsidiaries (Yu, 1990) and refine their international human resource management strategies (Taylor et al., 1996).Whereas less experienced MNCs may still prefer HCNs as CEOs of geographically distant subsidiaries to save cost, more experienced MNCs will rather select subsidiary CEOs to match their strategic goals (Ando, 2011).
Furthermore, we argued that MNCs appoint many HCN CEOs to geographically distant subsidiaries, as increasing distance makes it harder to find PCNs willing to relocate to far away subsidiaries.Internationally experienced MNCs are less likely to face this restriction, since they typically have developed a pool of internationally minded PCNs capable of dealing with the challenges of distant assignment locations (Ando & Paik, 2013).Among MNCs with extensive overseas experience, it is therefore less likely that high geographic distance will lead to an increased use of HCNs in overseas subsidiaries.Overall, we hypothesize: Hypothesis 4c.The propensity to staff the subsidiary CEO position with a HCN or TCN due to high geographic distance decreases with rising international experience of the MNC.

Data
We tested our hypotheses by surveying a sample of German MNCs.Using a survey is useful to approach our research question, because information on the subsidiary-staffing decisions of companies is not accessible with enough detail in secondary data.Furthermore, the surveydesign enables us to consider several determinants of staffing decisions which are not publicly available, such as the initial motive to set up a subsidiary and its degree of autonomy from the German MNC.We obtained contact details of 2,313 MNCs that were internationally active with at least one majority owned FDI (i.e., acquisition or greenfield investment) from the AMADEUS database.
To investigate the interplay between communication and staffing, we developed a questionnaire based on established scales.We administered our instrument in German, as our respondents in German HQs were almost exclusively native German speakers.To translate the source scales from their English original, we followed the back-translation method (Brislin, 1970;Chidlow et al., 2014).We pretested the questionnaire with several academics knowledgeable in international management and through personal interviews with Human Resource executives.
In the second half of 2014, we sent a paper-based version of the questionnaire to the MNCs' CEOs or leading Human Resource managers, as these individuals were relatively easy to identify and most likely to oversee their companies' international activities.To increase the response rate, we reminded all MNCs by e-mail and attached our questionnaire.We also called each MNC that had not responded at that time to remind it once more and, again, e-mailed the questionnaire to those MNCs that generally agreed to participate in the survey.This procedure yielded 127 responses, which amounts to a 6 percent response rate.This is consistent with Harzing (1997), who reports that international mail surveys typically reach response rates between 6 and 16 percent.We do not consider our response rate unusually low, given that we exclusively directed our questionnaire at high-level executives and that the pace of business and the use of mail surveys has intensified since Harzing's study (Harzing & Noorderhaven, 2006a).To avoid problems with memory bias, we excluded foreign market entries that took place more than 10 years ago.Due to missing data in our dependent and independent variables, our final sample included 101 MNCs.1 These firms were operating mostly in the manufacturing sector, covering a wide variety of branches such as automotive, engineering, and food production.As depicted in Table 1, the subsidiaries were located all over the world, with China, India and the United States representing the three most frequently targeted countries.To check for nonresponse bias, we compared early and late respondents (first and last 20 percent of the sample).Following Miller and Smith (1983), we performed a t-test on key MNC characteristics such as size, age, or type of business, but did not obtain significant differences.

Dependent Variable
Our dependent variable CEO staffing refers to whether the first CEO after the focal subsidiary's establishment was a PCN expatriate (or not) (e.g., Belderbos & Heijltjes, 2005).To obtain data on CEO staffing, we asked the respondents whether the CEO presiding over the company's most recently established foreign subsidiary immediately after its inception was German (PCN expatriate) or a host country citizen.In case none of this was true, we asked which other nationality the CEO had.We constructed a dichotomous variable with PCNs coded as '1' and other nationalities (i.e., HCN or TCN) coded as '0'.

Independent Variables
Our independent variables are the three communication barriers covered by Slangen (2011), which we measured by drawing on the sources he specified.That is, to measure the native language barrier between Germany and the respective host country, we drew on the composite factor consisting of three 5-point Likert-scale items Dow and Karunaratna (2006) developed to depict differences in language.As the composite index and its underlying items are described in detail in Appendices A-C in Dow and Karunaratna (2006), in the following we only shortly present the general idea of the index and refer interested readers to the detailed descriptions in Dow and Karunaratna (2006).The first item measures the difference between German and the closest major language in the respective host country.A major language is defined as any language which more than 20 percent of a host country's population are able to speak or which has a special status in a country (e.g., English in India).Based on the Language Family Index by Grimes and Grimes (1996), the difference to German is assessed on a scale ranging from 1 (same language) to 5 (different language families).The second item reflects the proportion of a host country's population being able to speak German, whereas the third measures the proportion of Germans skilled in the host country's major language(s).These latter two items are assessed on a scale ranging from 1 (90 percent or more) to 5 (less than 1 percent).Following Slangen (2011) and Dow and Karnaratna (2006), our measure of native language barriers then consists of the mean value of these three items.
Following Slangen (2011), we operationalize the foreign language barrier as the average English language deficiency in the host country, given that English is the most frequently used foreign language in MNCs (Harzing & Pudelko, 2013).We drew on the scores achieved by examinees from a focal host country on the Test of English as a Foreign Language (TOEFL), as the TOEFL represents the most widely accepted test for English proficiency worldwide (Slangen, 2011).Using scores from language exams is particularly useful for our analysis, as candidates for CEOs of newly established subsidiaries in a given country are likely to emerge from the pool of individuals who have demonstrated international ability through language exams.For the years 1992 to 2006, we used the average scores of examinees, who took the paper-based test from July of the year preceding the MNC's market entry into the respective host country until June of the market entry's year.As the paper-based test was no longer available for the years after 2006, we drew on the internet-based test results achieved by examinees in the full year of the entry in a focal host country.As the paper-based and the internet-based test results are presented on different point scales, we calculated the percentage of maximum points reached to standardize the English language proficiency of examinees between both procedures.We then inverted the percentage to yield the deficiency in English language with higher values indicating a higher foreign language barrier.We collected the test results from the website of the Educational Testing Services (ETS, 2017), which designed the TOEFL and from other internet sources (cgsnet.org). 2inally, geographic distance encompasses the great-circle distance in kilometres between the midpoints of the respective cities where the HQs and the foreign subsidiary are located.We utilized the Google Maps Distance Calculator developed by Daft Logic to determine this distance (Daft Logic, 2017).To correct for skewness in the distance distribution, we used the logarithm of the distances for our calculations.
We measure our moderator variable general international experience as the number of countries an MNC was doing business in at the time of the survey.We subtracted 1 from this value to receive the number of countries the MNC operated in before the most recent subsidiary's inception.This corresponds to our dependent variable, given that our questionnaire asked for the subsidiary CEO's nationality at the time the latest subsidiary was established.Our approach follows extant research (e.g., Slangen & Van Tulder, 2009) and uses an established measure to assess an MNC's international experience.

Control Variables
Our analysis also included several control variables.We controlled for MNC size by including the logarithmic transformation of the number of employees, because larger firms may possess a larger pool of PCNs that can be deployed to foreign subsidiaries (Ando & Paik, 2013;Delios & Björkman, 2000) or can even draw on a larger amount of HCNs employed by the MNC.
Moreover, firms showing better performance might find it easier to afford the costly deployment of expatriates (Bonache et al., 2001).We therefore included MNC performance as a control variable, asking respondents to assess on a 5-point Likert scale how satisfied they were with the company's overall performance over the last three years relative to their competitors (Hult et al., 2003).Using a subjective performance measure is the most suitable approach in our survey design.As many firms in Germany are privately held, they do not disclose objective performance measures, which ultimately limits our abilities to access such data and at the same time makes respondents reluctant to report their objective performance measures in surveys (Deutscher et al., 2016).Furthermore, we controlled for the share of family ownership as prior research found family firms to differ in their internationalization strategies from non-family firms (Fernández & Nieto, 2006;Graves & Thomas, 2006).In a similar vein and to control for firm-specific ties that shape internationalization decisions, we considered the degree to which foreign firms owned shares of the company by retrieving information on the share of foreign ownership in the AMADEUS database.To control for industry-specific effects, we also added a binary variable manufacturing that indicated whether a firm was operating in manufacturing industries (or service industries).
As cultural distance has been shown to affect a firm's readiness to post expatriates in foreign subsidiaries (Gong, 2003), we also controlled for cultural distance, measured with the index developed by Kogut and Singh (1988) based on Hofstede's (1980) cultural dimensions.We further assessed an MNC's experience with prior foreign market entries of the same type, distinguishing between greenfield experience, partial company takeovers, and full company takeovers (Agarwal & Ramaswami, 1992;Brouthers & Nakos, 2004).Past mode experience was thus assessed by asking the respondents to depict on a 5-point Likert scale to what extent they agreed or disagreed with the statement of having broad experience with the respective foreign operation mode.Furthermore, we measured on a five-point Likert scale how important MNCs had considered access to knowledge, access to raw materials, access to markets and gaining cost-advantages as motives for setting up the respective subsidiary.These were potentially important controls, as an MNC's strategic motives for the market entry may influence knowledge transfer and communication behaviour (Bresman et al., 1999).Finally, we controlled for the planned level of marketing autonomy (two item-scale, Cronbach's alpha = 0.70) as well as technological autonomy (three-item scale, Cronbach's alpha = 0.75) that was granted to the subsidiary at the time of its inception, as these indicators determine the extent of knowledge exchange, coordination and control efforts (Slangen & Hennart, 2008).To fully leverage the information of our dependent and independent variables, we mean-imputed missing information of control variables.3

Assessing Common Method Variance
Common method variance (CMV) presents a potential problem for survey-based research when the variables included in an analysis come from a single source (Podsakoff et al., 2003).In our study, however, we believe CMV not to be a major problem for at least four reasons.First, our dependent variable is more objective in nature and therefore less likely to produce a bias compared to a perceived measure.Second, we obtained our independent variables (i.e., native language barrier, foreign language barrier, and geographic distance) from secondary sources.Third, we included interaction terms in our analysis, which present complex structures that are less likely to be part of respondents' considerations and thus are likely to reduce CMV (Chang et al., 2010).Fourth, we conducted Harman's single factor test as suggested by Podsakoff and Organ (1986) revealing six factors with the largest one accounting for 18 percent of the variance.

RESULTS
Table 2 depicts the mean values, standard deviations, variance inflation factor (VIF) values, and bivariate correlations of our variables.The correlation coefficients show no serious risk of multicollinearity, as they are below the critical value of 0.70 (Alm & Mason, 2008).Likewise, the highest VIF amounts to 2.24, which is well below the critical value of 2.50 (Allison, 1999).
Hence, we do not consider multicollinearity to be a major problem in our data.

[INSERT TABLE 2 ABOUT HERE]
As our dependent variable is dichotomous, we perform a binary logistic regression analysis to test our hypotheses.Table 3 gives the results.Following Aiken and West (1991), we present different models that allow highlighting changes in the model fit and explanatory power between alternative models.Model 1 includes the control variables of which only MNC performance and family ownership show a constantly significant effect on CEO staffing.In model 2, we add the three direct effects of the communication barriers, which increases the explanatory power of the model compared to model 1from 13 to 19 percent (Cox and Snell) and from 17 to 26 percent (Nagelkerke), respectively.We find that the native language barrier has no statistically significant effect on the staffing decision (p > 0.10), leading us to reject hypothesis H1.However, the foreign language barrier has a statistically significant and positive effect (r = 0.96; p ≤ 0.05) on the staffing of PCNs in foreign subsidiaries, whereas the geographic distance barrier shows a significant effect in the opposite direction (r = -0.62;p ≤ 0.10), thus decreasing the likelihood of choosing an expatriate to manage a subsidiary abroad.
These results thus confirm both hypothesis H2 and H3.

[INSERT TABLE 3 ABOUT HERE]
In model 3, we included the direct effect of our moderator variable, which shows a significant and negative effect (r = -0.93;p ≤ 0.05) on the staffing decision of PCNs as subsidiary CEOs.
We added the interaction effects in models 4-6.These models show a significant increase in model fit (∆χ2) compared to model 3, indicating that the interaction terms are not trivial (Jaccard, 2001).Whereas model 4 illustrates a weakly significant moderating effect of international experience on the native language barrierexpatriate staffing relation (r = 0.86, p ≤ 0.10), model 5 shows a significant moderating effect (r = 0.85; p ≤ 0.05) of general international experience on the relationship between the foreign language barrier and the MNC's expatriate staffing decision.Model 6 indicates a moderating effect for geographic distance (r = 0.81; p ≤ 0.05).
Analysing interactions in non-linear models is more complex than in linear models, as they cannot be interpreted simply on the basis of the direction, magnitude, and statistical significance of the coefficients (Hoetker, 2007;Norton et al., 2004;Zelner, 2009).Therefore, we use graphs to analyse the interaction effects.Following Zelner's (2009) recommendations, we report two sets of plots in Figures 1-3.In the first set of plots, we compare the projected probabilities of PCN staffing (y-axis) at different levels of native language barrier, foreign language barrier, and geographic distance (x-axis), respectively, for high (dashed line) and low (solid line) levels of international experience.To further assess the significance of the interaction effects, we include a second set of plots showing the difference between high and low levels of international experience for different levels of native language barrier, foreign language barrier, and geographic distance, respectively (i.e., the vertical distance between the dashed and the solid line in the first set of plots).The bars depict the 95 percent confidence intervals for these differences.The interaction effect is significant in the areas where the confidence interval lies either completely above or below zero.In contrast, where the confidence interval includes zero, the MNC's general international experience has no significant moderating effect.

[INSERT FIGURES 1-3 ABOUT HERE]
Although model 2 shows no significant direct effect of the native language barrier, model 4 points to a significant moderating effect of international experience on the relationship between the native language barrier and subsidiary CEO staffing.The graphical analysis provides an even more detailed perspective, as Figure 1a suggests that at high levels of international experience, the probability of choosing an expatriate as subsidiary CEO is generally lower than at low levels of international experience.Figure 1b additionally demonstrates a significant negative moderating effect from low to high values of the native language barrier as indicated by the confidence intervals that do not include zero in this range.Hence, we find an indication for the theoretical rationale developed in hypothesis H4a arguing for a weaker effect of the native language barrier on PCN staffing in the presence of high international experience.
Figures 2a and 2b indicate that international experience is a significant negative moderator only from low to medium values of the foreign language barrier, providing only weak support for our hypothesis H4b, in which we suggest that international experience weakens the positive effect of foreign language barrier on staffing subsidiaries with PCNs. Figure 3a demonstrates that at high levels of international experience, the effect of geographic distance on subsidiary CEO staffing is weak as indicated by the almost horizontal dashed line.It also shows that the MNC's likelihood to employ expatriates is generally lower when its international experience is high.In Figure 3b, we see that international experience is a significant positive moderator from low to medium-high values of geographic distance.These results support our hypothesis H4c predicting a weaker negative effect of geographic distance in the presence of high international experience, leading us to accept this hypothesis.
We conducted several robustness checks to corroborate our empirical analysis. 4First, we addressed the concern that while an individual's nationality in most cases determines his or her mother tongue, it does not cover the individual's entire communication skills.Looking only at the subsidiary CEOs' nationalities may lead us to underestimate the hypothesized impact of native language barriers on expatriate staffing, as we would underestimate HCN subsidiary CEOs' German skills and PCN subsidiary CEOs' host country language skills.To test in how far the possible heterogeneity in communication skills indeed affects our empirical analysis, we conducted a robustness check by taking advantage of two specific questions on the mastering of languages by subsidiary CEOs, which were included in our survey: First, we asked each respondent to assess the German language ability of the CEO in the subsidiary on a scale from 1 (does not speak German) to 5 (speaks fluent German).With this information, we reran our regressions on a sample including only those CEOs who are either of German nationality or non-Germans who do not master German (rated 1 or 2 on the scale of language ability).Also in this smaller sample of 61 observations that mitigates heterogeneity between an individual's nationality and communications skills, the impact of native language barriers on CEO staffing remains statistically insignificant.Next, we drew on a question asking each respondent to assess the subsidiary CEO's host country language proficiency on a scale from 1 (does not speak the host country language) to 5 (speaks the host country language fluently).We then again reran our regressions only including non-Germans speaking the host country language on the one hand and Germans without host country language skills (rated 1 or 2 on the scale of language ability) on the other hand. 5Again, and confirming our main results, the impact of native language barriers on CEO staffing is statistically insignificant.The results of these robustness checks therefore indicate that differences between an individual's nationality and his or her communication skills are unlikely to be driving our results.
Second, we assessed the validity of TOEFL results as proxy for foreign language barriers.TOEFL results may vary across countries due to other factors than English proficiency, such as exam costs or the popularity of the TOEFL exam.While we believe that such confounding effects are less likely to hold, we further probed into our empirical analysis by considering a more individual measurement of foreign language barriers.To this end, we relied on two questions regarding a) the respondent's perception on English proficiency among managers in the subsidiary and b) the respondent's perception on English proficiency among managers in the parent company.In both cases, English proficiency was assessed on a scale from 1 (managers do not speak English) to 5 (managers speak fluent English).As the limitations of communication are determined by the group with lowest English proficiency, we construct a variable that takes on the lower value of English proficiency among (a) subsidiary managers and (b) parent company managers.Substituting the TOEFL-based measurement of foreign language barriers with this proxy of English proficiency, we find a positive and statistically significant impact of foreign language barriers on expatriate staffing, consistent with the findings derived by the TOEFL-based proxy of foreign language barriers.We are therefore confident that the TOEFL-based measurement of foreign language barriers as suggested by Slangen (2011) represents, despite its limitations, a solid proxy of language barriers.Finally, we probed deeper into the validity of our subjective measurement of MNC performance by alternatively considering public financial indicators, which, however, are often not available, as many firms are privately held and do not publish financial data.To this end, we retrieved financial data from the AMADEUS database for all firms in our sample for which such data was available.We chose earnings per employee as an indicator of profitability (which was available for 62 firms in our sample) and mean-imputed missing values.Substituting our subjective performance measure with this objective measure and rerunning our empirical analysis yielded similar results, leading to the same conclusions for the confirmation of our hypotheses on CEO expatriate staffing.Furthermore, similar to the subjective indicator of profitability as a control variable, the objective indicator suggested a statistically significant negative impact of performance on expatriate staffing.We are therefore confident that our subjective indicator represents a solid proxy of firm performance as a control variable.

Theoretical Implications
Whereas communication-based approaches have recently explained strategic decisions such as the choice between acquisitions and greenfield investments (Slangen, 2011), the selection of international alliance partners (Joshi & Lahiri, 2014), or the stake taken in cross-border acquisitions (Cuypers et al., 2015), our study develops a communication-based perspective on foreign subsidiary CEO staffing.The present paper develops and tests a model displaying the effects of a native language barrier, a foreign language barrier, and geographic distance on MNCs' staffing of subsidiary CEO positions, contingent on the corporation's general international experience.Testing our theoretical predictions on a sample of German MNCs, we find support for almost all of our hypotheses.
Our investigation of linguistic communication barriers advances the burgeoning research on language diversity in MNCs (for overviews see Tenzer et al., 2017 or the recent thematic issue in the European Journal of International Management 2018(1/2)) by differentiating the influences of native and foreign language barriers on global staffing decisions.Contrary to our initial expectations, we find that the native language barrier does not affect MNCs' staffing decisions, indicating that firms consider local languages irrelevant for inter-unit communication.This may be due to the high diversity of local languages which HQs of globally dispersed MNCs are facing, making it unfeasible to address each subsidiary in its national language(s) and dictating the use of a business lingua franca, in most cases English (Kankaanranta et al., 2018).The importance of this shared language, which Komori-Glatz (2018: 47) describes as "the dominant language in international business", explains that the foreign language barrier, which captures the English proficiencies of host country nationals, significantly (albeit weakly) influences firms' probability to rely on expatriates.
With these findings, we contribute to the active debate on Business English as a Lingua Franca (BELF, see e.g.Millot, 2017) in modern MNCs.Whereas critical management scholars lament the "expansionist, imperialistic and hegemonic" (Paunova, 2017: 887; also see Boussebaa & Brown, 2017) force of Englishnization, others portray BELF as a "simplified, hybridized, and highly dynamic communication code" (Kankaanranta & Planken, 2010: 380) which is no longer "owned" by a specific group of native speakers (Kankaanranta & Louhiala-Salminen, 2013;Komori-Glatz, 2018).Our sample companies' reliance on English best reflects the stream praising the benefits of a common corporate language in terms of its "democratizing effect" (Steyaert et al., 2011) and extols English as "the shared code used to 'get work done' in international business" (Kankaanranta & Planken, 2010: 380).
However, our findings also indicate that the frequently invoked reality of the modern MNCs as a "multilingual community" (Aichhorn & Puck, 2017: 386; also see Kankaanranta et al., 2018) is not on top of decision makers' minds, who often fail to realize that introducing BELF does not render workplace communication monolingual (Fredriksson et al., 2006).Even after an MNC has established English as the official language for communication within the company, local employees still rely on their native languages.Particularly those with low confidence in their lingua franca skills frequently communicate with compatriots in their native language, even if colleagues from different language groups are present (Tenzer et al., 2014).
Consequently, the implementation of a common corporate language can entail unintended disintegrating effects (Piekkari et al., 2005;Sanden, 2014).Against this background, it is possible that HQs' greater ease of communicating with a PCN as subsidiary CEO comes at a cost in the sense that this expatriate finds it difficult to communicate with local colleagues.If everyone but the expatriate speaks the local language, high-quality exchange between the PCN expatriate and HCNs may be difficult to develop (Woo & Giles, 2017), while HCNs form a cohesive group based on their local language fluency (Vigier & Spencer-Oatey, 2018).The fact that subsidiaries led by expatriate CEOs have more language policies in place than subsidiaries with HCN top managers (Peltokorpi, 2015b) may be a symptom of PCNs' difficulty to remain in control of local communication.Following up on Tenzer & Schuster's (2017) recent conceptual work, future studies may further examine the intricate interplay of native language and lingua franca communication on the HQ and subsidiary level and in the context of expatriate staffing.
Our study also highlights the importance of geographic distance for expatriate staffing decisions by showing that MNCs are less likely to choose PCNs for the CEO positions of distant subsidiaries.This result aligns with Harzing and Noorderhaven's (2006a) observation that PCN CEOs are less used in Australian and New Zealand subsidiaries of MNCs headquartered in the US, Japan, or Europe.It also supports Abdellatif et al.'s (2010) finding that geographic distance negatively influences MNCs' use of expatriates in foreign subsidiaries.Whereas advanced communication technology and cheaper transportation make geographic distance less relevant in other areas of business (see e.g., Ganesan et al., 2005;Petersen & Rajan, 2002), our findings show that it still matters for international human resource management and global staffing.In sum, we contribute to research on global staffing by disentangling the influence of three different communication barriers, helping us to elucidate the complexity of an MNC's staffing strategy (Paik & Ando, 2011) and underlining the importance of expatriates as agents of coordination, control, and knowledge transfer.
Moreover, the present study shows that general international experience affects the relationships between communication barriers and MNCs' staffing decisions for foreign subsidiaries.Considering that the communication barriers we investigated constitute elements of psychic distance (Håkanson et al., 2016), our findings contradict the tenet that only countryspecific experience can mitigate the impact of psychic distance on strategic decisions.Authors endorsing this view argue that psychic distance is caused by differences between specific markets and can therefore only be overcome by market-specific host country experience (Dikova, 2009).In contrast, we show that MNCs with international experience in a broad range of economies are generally less likely to rely on PCN expatriates as communication conduits, possibly because they can manage the distance with standardized processes and language policies instead.Thus, we extend the literature by indicating that international experience helps to reduce uncertainty caused by cultural distance (Delios & Henisz, 2003).Furthermore, our findings suggest that the influence of different distance measures on global staffing decreases as MNCs gather international experience in a variety of country markets.
Experienced corporations manage linguistic communication barriers more effectively, reducing their need to bridge the language gap through expatriation.They are undeterred by the costs of sending PCNs to distant locations and have a larger talent pool of internationally minded candidates willing to accept foreign assignments, relieving them of the need to staff far away subsidiaries with HCNs.Instead, MNCs with extensive overseas experience can select their subsidiary CEOs to match strategic goals.With these findings, our study expands the contingency perspective on expatriation (e.g., Belderbos & Heijltjes, 2005;Delios & Björkman, 2000) with international experience as a new component.It also informs a key question in international business research: does distance still matter in today's interconnected world?Our findings inform this debate in a dialectical way.On the one hand, we show that distance still matters for communication, and, hence, for global staffing.On the other hand, we demonstrate that its importance declines as MNCs gain increasing international experience.

Managerial Implications
Our study has several implications for the management of HQ-subsidiary relations across borders, particularly with regard to subsidiary staffing.Results show that HQ managers should be aware of various communication barriers requiring different approaches to subsidiary top management selection.Whereas HCNs and TCNs may be eligible to establish coordination and control as well as knowledge transfer between HQs and geographically distant subsidiaries, the decision is more complex in the presence of linguistic barriers.Our empirical findings suggest that native language barriers hardly matter for subsidiary CEO staffing, yet previous studies have established connections between PCN expatriates' local language proficiency and their cultural adaptation (Zhang & Peltokorpi, 2016) or their relationships with host country employees (Zhang & Harzing, 2016).
Our findings regarding foreign language barriers highlight the importance of BELF as "the medium through which many business people get their work done" (Nickerson, 2015: 390).
Supporting Kankaanranta and Planken's (2010: 380) view that English competence "can be considered an essential component of business knowledge required in today's global business environment", our study shows that MNCs need to promote the adoption of English in their globally dispersed units and support this policy with on-the-job language training (also see Neeley, 2017).Depending on the HQs' and/or subsidiary's location, other languages besides English might also be relevant for corporate communication (e.g., French in Africa).Finally, an MNC's international experience determines how competently HQs manage their foreign subsidiaries and how confidently they rely on local top managers.Hence, staffing decisions should not be made in isolation, but under consideration of specific firm-or country-related factors.

Limitations and Future Research
As with most empirical research, our study has several limitations.First, we do not differentiate between HCNs and TCNs as alternatives to PCN staffing.Such differentiation was not feasible due to the low sample size on which our analysis is based.It is quite common in the expatriate staffing literature to compare PCN staffing to only HCN staffing and to exclude TCNs from the analysis due to their negligible number in the investigated subsidiaries (e.g., Colakoglu & Caligiuri, 2008), or to consider HCNs and TCNs jointly in one category as we did (e.g., Gaur et al., 2007).Nevertheless, we encourage future research to disentangle the communicationbased choice between PCN, HCN, and TCN staffing.
Second, due to the low sample size, our findings should be interpreted with caution.One may argue that although small samples are not unusual in research on staffing decisions (e.g., Ando et al., 2008;Pérez & Pla-Barber, 2005), our analysis might have low statistical power reducing the chances of detecting true effects or increasing the probability of overestimating identified effects.We therefore urge future studies to replicate our study on larger samples.Third, we only studied staffing decisions of MNCs located in Germany, which restricts the generalizability of our findings.Relying on only one home country might be problematic due to specific characteristics of the German language or managers being German native speakers.
Prior research also finds that German MNCs tend to have a generally higher propensity to staff foreign subsidiaries with expatriates (Harzing, 2001a), which might affect our findings.Against this background, we encourage future research to study how communication barriers impact staffing decisions of MNCs from other countries.
However, foreign assignments can serve a broad range of different purposes, which entail different communicative requirements and linguistic challenges (Tenzer & Schuster, 2017).
Following Hocking et al. (2007), future research may shed more light on the kind of knowledge transferred by expatriates and the reference points of their communication efforts.
Fifth, we acknowledge that an individual's nationality and communication skills may differ, which limits the generalizability of our findings as some firms may have candidates for subsidiary-CEO positions at their disposal who master several languages.Although a robustness check which excludes those individuals indicated that our results are unlikely to be influenced by such effects, we encourage future research to shed more light on the specific role of these possibly ideal bridge-makers between HQ managers and local employees.
Sixth, we acknowledge that the decision to use expatriates as CEOs in foreign subsidiaries depends on a variety of factors.In this paper, we focus on language barriers as an important component of managerial reasoning in international contexts (Slangen, 2011) to derive how language barriers affect CEO expatriate staffing through communication abilities.Future research may further consider other motives for expatriate staffing such as a higher confidence in managers of the same nationality.Those studies may fruitfully investigate the impact of other motives on CEO-subsidiary staffing in consideration of the role of language barriers we have found.
Seventh, we cannot rule out with certainty that our empirical proxy of foreign language barriers (TOEFL scores) may be influenced by country-specific factors such as the popularity of the TOEFL in the respective country.While we find similar results on the impact of foreign language barriers on expatriate staffing when using a more firm-specific proxy, future research may develop and test further measurements of language barriers to validate the TOEFL-based measurement introduced by Slangen (2011).
Finally, we only included different subsidiary roles as a control variable, but did not theoretically distinguish between subsidiaries that were set up for different purposes.Whereas this aspect was beyond the scope of our study, one could argue that the direction and intensity of communication and the subsequent impact of communication barriers on CEO staffing might vary between subsidiaries merely geared to exploit low-cost advantages, subsidiaries established to tap into local knowledge resources, and subsidiaries aiming to access local markets.Future research may fruitfully relate our study's topic to Gupta and Govindarajan's (1991) typology of subsidiary roles.

CONCLUSION
The management of human resources in HQ-subsidiary relationships requires intensive communication (Harzing & Pudelko, 2013), but is frequently hampered by communication barriers.Subsidiary CEOs are at the centre of this challenge, as they are expected to bridge the divide between HQs and foreign subsidiaries (Boyacigiller, 1990;Colakoglu & Caligiuri, 2008).Our communication-based perspective on subsidiary CEO staffing shows that MNCs' "urgent need" (Kaul, 2014: xi) for effective communication across linguistic and geographic communication barriers affects MNCs' choice between PCN expatriates or local employees in those positions.We have connected staffing research to the fast growing stream on language in international human resource management, highlighted the continued importance of geographic distance for staffing decisions, and demonstrated the moderating effect of the MNC's international experience.This model helps us to better understand the complexity of MNCs' staffing strategies and underlines the importance of expatriates as communication agents.

Figure 2a :Figure 2b :
Figure 2a: Interaction plot -Foreign language barrier: Predicted probabilities for high and low international experience

Figure 3a :Figure 3b :
Figure 3a: Interaction plot -Geographic distance: Predicted probabilities for high and low international experience

Table 1 :
Location of subsidiaries in our sample

Table 3 :
Results from binary logistic regression analysis (DV=1 if CEO in the subsidiary is German; =0 otherwise) Interaction plot -Native language barrier: Predicted probabilities for high and low international experience Figure 1b: Interaction plot -Native language barrier: Delta predicted probabilities for high vs low international experience