Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes

The entry mode decision is a critical topic in international business. Several studies have tested the antecedents of entry mode choices and compared the outcomes after entry. However, the results reported in these studies are contradictory and difficult to explain. Furthermore, the reviews using qualitative approaches cannot statistically combine empirical results and fail to fully discuss these relationships. Additionally, the mediating effects remain unexamined in the existing studies. Drawing upon transaction cost economics and the resource-based view, this paper presents a meta-analysis that combines entry mode choices, antecedents and post-entry outcomes to address these issues. This meta-analysis is based on 1499 effect sizes from 230 published empirical studies. This study focuses on 15 antecedents as well as post-entry performance and survival of entry mode choices. In addition to bivariate relationships (i.e. how a single antecedent affects entry mode choices), the paper uses meta-analytic structural equation modelling to analyse the mediating effects of entry mode choices on the antecedent - outcome relationship.


Introduction
A growing number of multinational enterprises (MNEs) are attracted by the opportunities and resources of international markets. In this context, choosing a suitable entry mode into a foreign market is a key decision for MNEs (Giachetti et al., 2019). The entry mode is closely related to an MNE's operations, growth and performance (Buckley, 2004;Delios & Beamish, 1999;Morschett et al., 2010). However, due to unfamiliarity with the markets and the additional burden imposed by operating in foreign markets (Kao & Kuo, 2017;Tihanyi et al., 2005), making entry mode decisions is challenging. Therefore, it is paramount to understand and explore the antecedents and outcomes of entry mode choices for MNEs. Entry modes have been described as a critical part of the international operations of MNEs. The ownership mode choice refers to 'the level of equity ownership determined in a foreign subsidiary' (i.e. wholly-owned subsidiary [WOS] or joint venture [JV]), while the establishment mode choice refers to 'MNEs acquiring an existing enterprise or building a new startup' (i.e. acquisition or greenfield) (Arslan et al., 2015, p. 997). Regarding their differences, Dikova and van Witteloostuijn (2007) advocated that the two entry mode choices cover different aspects of MNEs' foreign direct investment (FDI) decisions. The ownership mode choice refers to the control level the MNE wants to exert, while the establishment mode choice focuses on the combination of local assets with the ones already owned by the MNE (Dikova & van Witteloostuijn, 2007).
Increased attention has been given to the antecedents and outcomes of entry mode choices in recent decades. Regarding the antecedents of entry mode choices, scholars have tested myriad external factors (e.g. cultural distance and political factors) and internal factors (e.g. firm resources and technologies), all of which affect entry mode choices (Laufs & Schwens, 2014;Meyer et al., 2009). Previous qualitative reviews exploring entry mode choices have listed theoretical models and frameworks (e.g. De Villa et al., 2015;Surdu & Mellahi, 2016), identified antecedents (e.g. Laufs & Schwens, 2014;Meyer et al., 2009) and proposed future research directions (e.g. Brouthers, 2013;Javalgi et al., 2011).
However, the qualitative reviews have not been able to explain the inconsistencies among empirical results. These studies have failed to provide clear guidance for researchers' and managers' decision-making processes because they cannot statistically combine conflicting results from individual empirical studies (Geyskens et al., 2006). These conflicting results pose substantial difficulties for both academic research and foreign entry decisions. To reconcile these conflicting results, some scholars have used quantitative meta-analyses to test some propositions (Cooper & Hedges, 1994). These meta-analytical reviews allow for the quantitative integration of existing results (Kirca & Yaprak, 2010;Tan & Sousa, 2013) and provide a better understanding of the current empirical research.
Although the existing meta-analyses have reported some significant results, a comprehensive review of the entry mode choice is still timely and important (Surdu & Mellahi, 2016). A theoretical framework that combines the antecedents and outcomes of entry modes is still needed for four reasons. First, the existing 1 3 Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes meta-analyses have not combined the effects of the antecedents on the ownership mode and the establishment mode and their effects on the outcomes in one single framework. As two of the most frequently examined entry modes, the ownership mode choice and the establishment mode choice are regarded as two decisions that are made together before international entry Dikova & van Witteloostuijn, 2007). Therefore, the antecedents, ownership mode choice, establishment mode choice and their outcomes should be examined together. This review combines transaction cost economics (TCE) and the resource-based view (RBV) to explain the decision process. This provides a significant departure from the previous meta-analyses, as they did not explain the entry mode choices, their antecedents and outcomes under one framework.
Second, the existing studies have only examined a small subset of factors such as parent-firm characteristics and home-host country differences based on a specific theoretical lens (e.g., TCE, RBV); as a result, the effects of many other antecedents remain unclear. By combining the notions of TCE and RBV in one single framework, this review investigates four groups of antecedents (i.e., parent-firm characteristics, host country characteristics, parent -foreign affiliate differences, and homehost country differences). Moreover, the empirical studies have largely failed to test the different effects on various entry mode choices (i.e. ownership mode choice and establishment mode choice).
Third, the differences in the outcomes of the entry mode choices still need to be addressed by researchers. MNEs have different goals when they launch their international operations, and the post-entry outcomes are important indicators for evaluating the appropriateness of an entry mode for achieving those goals. The existing meta-analyses on the entry mode -outcome relationship have analysed how the ownership mode choice affects the outcomes (e.g. Giachetti et al., 2019;Zhao et al., 2017). However, the effects of the establishment mode choice on the outcomes have largely been ignored. As a result, the current studies provide insufficient guidance on the post-entry outcomes for making entry mode choices.
Fourth, different entry modes entail different levels of organisational control, risk exposure, resource requirements and expectations of future returns (Tang & Buckley, 2020). All these characteristics are closely related to firms' outcomes, which means that the antecedents can influence the outcomes by affecting the entry mode choices. Entry mode choices are regarded as an exogenous random selection when evaluating performance in most existing studies (Brouthers, 2002), while the determinants of the entry mode choices are essential in analysing the outcomes (Papyrina, 2007). Though some empirical studies have solved this problem by testing the fitness of the entry modes, it is still unclear how specific antecedents affect the outcomes through the entry mode choices. Despite acknowledging the importance of indirect effects (Zhao et al., 2017), the existing meta-analyses have failed to explain the mediation role of entry mode choice between antecedents and outcomes.
To address the knowledge gaps discussed above, this paper conducts a comprehensive review of the entry mode choices, including their antecedents and outcomes, and contributes to entry mode research in four ways. First, this meta-analysis extends the existing knowledge by reconciling the inconsistent results from previous studies by including the antecedents, ownership and establishment mode choices and outcomes in one single framework. Our study draws upon TCE and the RBV to analyse the effects of the antecedents on the entry mode choices and their outcomes.
Second, this meta-analysis extends the understanding of entry mode choices by examining the effects of more antecedents (15 antecedents) and comparing two entry mode choices rather than one (i.e. ownership mode choice and establishment mode choice). The ownership mode choice and establishment mode choice are two decisions that are made together before international entry Dikova & van Witteloostuijn, 2007). How the antecedents affect the two choices in different ways has not been discussed thoroughly. The inclusion of these additional antecedents allows the classification of these variables into several groups (i.e. parent-firm characteristics, host-country characteristics, parent -foreign affiliate differences and home -host country differences), thereby linking more clearly the entry mode choice and the antecedents. Moreover, this meta-analysis highlights the importance of examining the two choices together and extends the current research by showing which antecedents affect the two choices.
Third, this paper analyses the relationships between the establishment mode choice and different post-entry outcomes (i.e. performance and survival), which have been largely ignored in previous meta-analyses. Performance focuses on the success of an MNE in the international market, while survival evaluates the longevity of the MNE's subsidiaries (Delios & Beamish, 2001). By testing and comparing different post-entry outcomes in relation to the establishment mode choice, this paper explores the entry mode choice through a more complete model. Since all the antecedents and post-entry outcomes are examined together, the paper provides a more comprehensive and integrative model after testing and comparing all possible models (Bergh et al., 2016). This expands the understanding of making entry mode choices from the antecedents to the post-entry outcomes.
Fourth, this paper explores the 'black box' in international entry by examining whether the entry mode choice works as a mediator between the antecedents and the post-entry outcomes. The black box refers to the mechanism through which antecedents affect post-entry outcomes. This paper combines two current streams of research (antecedents on entry mode choices and entry mode choices on outcomes) with meta-analytic structural equation modelling (MASEM) to analyse the mediating effects of entry mode choices and examine the mediating paths from the antecedents to the outcomes. The examination contributes to the understanding of international entry by explaining what effects the antecedents will have on the outcomes with different entry modes. Thus, the analysis enhances the comprehension of the entry mode choice mechanism and explains how the antecedents affect the outcomes through the entry mode choices.

Definitions of Ownership Mode and Establishment Mode
'Entry mode' refers to the method of organising foreign business activities after various factors are evaluated (Hill et al., 1990). This ranges from contractual  Table 2 presents an overview of the published meta-analytical studies that discuss the antecedents of the entry mode choice. A few meta-analyses have examined a group of antecedents of entry mode choices and the moderating effects on the main relationships. Zhao et al.'s (2004) study incorporated 38 empirical articles from 1986 to 2002 and focused on the constructs operating within TCE: asset specificity, research and development (R&D) intensity, country risk, cultural distance, international experience and advertising intensity. Their analysis strengthened the explanatory power of TCE in the ownership mode choice and clarified the boundary conditions from study-setting factors and statistical artefacts. Morschett et al. (2010) focused on the choice between WOS and the cooperative-based mode (equity-based or contractual). Their research covered 67 studies from 1984 to 2008 and incorporated 13 external factors. They found significant relationships for these external factors and also discussed how the relationships were altered by the industry type and the time of study. Tihanyi et al. (2005) focused on cultural distance in the entry mode choice. They strengthened the importance of cultural distance and showed significant moderation effects from country type, industry type and sample time. Magnusson et al. (2008) also examined the relationship between cultural distance and entry mode choice. They extended the knowledge on the boundaries of cultural distance on two levels (country level and individual level). In addition, they clarified the boundary conditions due to measurement, home-country type and sample time. Klier et al. (2017) compared the choice between greenfield and acquisition, particularly how two resource types (knowledge-based and experience-based) affect the establishment mode choice. The research extended the knowledge on the establishment mode choice by using the RBV and revealed the different effects of two kinds of resources on the establishment mode choice. Beugelsdijk et al. (2018) analysed how cultural distance influences the ownership mode and establishment mode choices. They complemented the understanding of cultural distance in various stages throughout the entire process of internationalisation and emphasised the importance of understanding when and for which aspects of the internationalisation process cultural differences are really important. Tang and Gudergan (2018) examined international experience and ownership mode choice by considering multiple-level moderators. They extended the understanding of the dynamic capabilities view and provided new insights regarding the role of the moderators in these relationships. Finally, Tang and Buckley (2020) examined how host-country risk affects the ownership mode choice and the moderation effect of home-country institutions. They broadened the knowledge of entry modes by using institutional theory (IT) and showing the boundary conditions of home-country institutions.

3
Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes

Meta-Analyses of the Outcomes of Entry Mode Choices
Two meta-analyses have been conducted on the outcomes of entry mode choices. We display a summary of the meta-analyses on the relationships between the entry mode choices and the outcomes in Table 3. Both meta-analyses recognised the different outcomes under the different ownership levels and analysed some moderation effects. Zhao et al. (2017) reviewed a number of empirical studies and suggested that TCE provides the most consistent predictions after reviewing the major theories underlying the entry mode -performance relationship. They also extended the understanding of how the relationship varies with context-specific factors. Giachetti et al. (2019) complemented the former study by explaining the boundary conditions and combining TCE and IT to analyse the effects of the host country on the entry mode -performance relationship.

Summary of Two Streams of Meta-Analyses
These published meta-analyses have provided many insightful explanations of both the antecedents and outcomes of entry mode choices. However, the meta-analyses of antecedents have only examined a limited number of antecedents (e.g. Beugelsdijk et al., 2018;Tang & Buckley, 2020). Another limitation is that the existing metaanalyses mainly focus on one entry mode choice (e.g. Tihanyi et al., 2005;Zhao et al., 2004), thereby failing to compare the effects of antecedents on different entry mode choices (i.e. ownership mode choice and establishment mode choice). Our revision indicates that only two meta-analyses have examined the effects of the ownership level on the outcomes (e.g. Giachetti et al., 2019;Zhao et al., 2017). This suggests a lack of attention to other entry mode choices' effects (e.g. establishment mode choice) on the outcomes. Moreover, the entry mode choice is still regarded as exogenous in affecting the outcomes without considering its determinants (Dikova & Brouthers, 2016;Shaver, 1998). The absence of meta-analytical studies that consider both antecedents and outcomes emphasise the need for combining two streams of meta-analyses. Therefore, this study extends the existing meta-analyses on the entry mode choice by analysing two entry mode choices, two post-entry outcomes and a significantly larger number of antecedents in one model.

A Proposed Framework for the Entry Mode Choice
Based on the environment-strategy-performance paradigm (Child, 1972;Zou & Stan, 1998), a firm's performance-achieving strategy is connected to both its internal and external environments. Following this paradigm, this paper examines the mechanism of the entry mode choice relating to its antecedents and outcomes. After reviewing the existing empirical studies, this paper identifies 15 of the most widely examined antecedents (see Table 4).
A large group of empirical studies has been allocated to explaining the entry mode based on various theoretical backgrounds (e.g. TCE, RBV and IT). TCE is Host-country experience Delios and Beamish (1999) Mudambi and Mudambi (2002) Host-market size Mutinelli and Piscitello (1998a); Delios and Henisz (2000) Lu (  the most popular one to explain the role of governance arrangement in minimising costs (Canabal & White, 2008). The RBV is also used in this paper to explain the entry mode choice because costs do not emerge independently. The RBV points out that MNEs keep competitive advantages by using possessed resources or accessing new resources (Anand & Delios, 1997;Barney, 1991). TCE explains the MNEs' motive of cost efficiency, while the RBV emphasises their desire for resource exploitation (Meyer et al., 2009). The combination explains the balance between costs and resources while considering the entry mode choice. The complementarity of these two theories provides a more complete and stable explanation of the entry mode choice. Thus, drawing from the RBV and TCE and using the antecedents previously identified (see Table 4), we developed our framework and classified these antecedents into four groups: parent-firm characteristics, host-country characteristics, parent-foreign affiliate differences and home -host country differences. All antecedents in the four groups have cost and resource implications on the entry mode decision. In the case of parent-firm characteristics, extra costs come from allocating the parent firm's resources to the international market. On the other hand, RBV argues that firms possess resources and capabilities that enable the firm to pursue different entry mode decisions. The host country may also increase operating costs and provide valuable resources. Moreover, one may look at the host-country characteristics to see how entry mode decisions can augment the use of resources in the host country. Parent -foreign affiliate and home -host country differences are sources of valuable resources but also lead to costs from operating in an unfamiliar market. TCE, for instance, considers cultural differences as a key part of transaction costs because it introduces uncertainty making it more challenging for the firm to operate. Based on TCE and the RBV, the antecedents are closely related to the entry mode choice through which an MNE can reduce potential costs and gain valuable resources.
The decision of the entry mode choice is not the final step in the internationalisation process, as MNEs pursue better outcomes through entering the international market. Two outcomes are captured in our framework: performance and survival. In our paper, the influence of the entry mode choice on the outcomes is explained from two perspectives. TCE claims that the costs MNEs face during international entry vary among different entry modes. The RBV explains that MNEs gain different amounts of resources using different entry modes. TCE and RBV as theoretical basis are important as the alignment of resources, transactions properties, and entry mode choice have subsequent impact on performance. Therefore, the entry mode choice significantly affects performance and survival, as MNEs select a proper mode to reduce costs and gain resources (Brouthers, 2002;Kim & Gray, 2008).
Finally, to further understand the entry mode choice, four moderators are included in the framework to examine the boundary conditions (i.e. time of study, homecountry type, journal level and mode choice measurement). To enrich the discussion surrounding TCE and RBV we need to consider the theories boundary conditions. A consideration of the boundary conditions is important for TCE and RBV since the addition of a contextual perspective may affect cost and resource decisions. Different environmental factors (e.g. home-country type) may weaken, strengthen or even convert the main effects of antecedents on entry mode choices (Magnusson et al., 1 3 2008; Sharma et al., 1981). An unfamiliar environmental setting causes information deficits, which in turn influence how firms manage their resources (Sirmon et al., 2007). Similarly, the costs MNEs face may vary with the environments in which the empirical studies are conducted. Therefore, the four moderators describing how the empirical studies are conducted may affect the main relationships. Based on the above discussion, we proposed the following framework (see Fig. 1).

Parent-Firm Characteristics
Experience is a type of knowledge (Klier et al., 2017;Sousa et al., 2021) that is significant in terms of entry mode choice (Klier et al., 2017;Zhao et al., 2004). Although most existing meta-analyses do not distinguish between different experiences (Schwens et al., 2018), this study tests two types: host-country experience and international experience. The host-country experience indicates the intensity of MNEs in a specific country, which is measured in years or the number of subsidiaries in the host country (Cuypers et al., 2015;Larimo & Arslan, 2013). International experience refers to the extent of MNE internationalisation, and it is measured in years or the number of subsidiaries in all foreign markets (Padmanabhan & Cho,

Host-country characteristics
Host-market size  Yamanoi & Asaba, 2018). Both types of experiences may reduce risks in the host market (Larimo, 2003). Hence, MNEs that have more experience prefer WOS (versus JV) and greenfield (versus acquisition).
R&D intensity has been used as a proxy for asset specificity (Yiu & Makino, 2002). Higher asset specificity is associated with higher transaction costs because MNEs need to protect their valuable assets from hazards (Anderson & Gatignon, 1986). In addition, an MNE adopts the entry mode that can better protect its assets in a host environment. Therefore, MNEs with higher R&D intensity prefer WOS (versus JV) and greenfield (versus acquisition). Similar to R&D intensity, advertising intensity represents an MNE's marketing assets. This is due to the high cost of transferring valuable assets into a host market. In this case, a high control mode is better for MNEs that have high advertising intensity (Lu, 2002). Hence, MNEs with higher advertising intensity prefer WOS (versus JV) and greenfield (versus acquisition).
Parent diversification refers to a firm's scope of operations (Morris, 1996). Less diversified MNEs own enough product-specific knowledge and have less desire to partner with other companies (Larimo & Arslan, 2013). Although diversified MNEs may lack this kind of knowledge, they develop the ability to control a sophisticated system that is beneficial for foreign operations (Hennart & Park, 1993). Thus, parent diversification affects an MNE's entry mode choice when selecting partners and utilising the firm's ability. MNEs with higher parent diversification prefer JV (versus WOS) and acquisition (versus greenfield). Parent size refers to the size of a business unit or scale of operations. Large firms operating in international markets have more advantages and are more successful than small firms because large firms can obtain more resources in the international market (Moini, 1995;Wolff & Pett, 2000). Large firms also have a greater ability to absorb available resources. Due to these differences, MNEs with larger scales prefer WOS (versus JV) and greenfield (versus acquisition).
Parent profitability shows a firm's ability to generate profits from activities and affects the firm's entry mode by modifying free cash flows (Yamanoi & Asaba, 2018). Better performing firms may be high risk-taking firms, so they may prefer a high control mode . Thus, MNEs with higher profitability prefer WOS (versus JV) and greenfield (versus acquisition). Parent age can show a firm's resources, capability and networks, so different ages can lead to different entry mode choices (Mudambi & Mudambi, 2002). Older firms can better handle international operations, as they have a larger pool of resources, capabilities and networks. Furthermore, older firms desire more new resources and face fewer risks than younger firms. Therefore, MNEs that have operated longer prefer WOS (versus JV) and acquisition (versus greenfield).

Host-Country Characteristics
Host-market size indicates the number of potential customers, since larger market sizes provide more latent partners and the opportunity for higher returns (Morschett et al., 2010;Solocha & Soskin, 1994). Host-market benefits increase an MNE's propensity to invest and enhance its commitment to a host country. Thus, MNEs prefer WOS (versus JV) and acquisition (versus greenfield) when they enter bigger host markets. Host-market growth is also an indicator of market attractiveness (Morschett et al., 2010). Fast market growth yields more opportunities for firms, while fastchanging environments can also lead to increased risk. A cooperative entry mode is more suitable in markets that have high growth. Thus, MNEs prefer JV (versus WOS) and acquisition (versus greenfield) when they enter host markets with higher growth. Host-country risk is another frequently discussed factor in empirical studies. In their meta-analyses, Zhao et al. (2004) and Morschett et al. (2010) respectively tested the effect of country risk on ownership-based entry mode choices and examined its impact on the choice between WOS and cooperative mode (e.g. JV and licensing). Host-country risk comprises many aspects, including politics and the economy (Delios & Beamish, 1999), which enhance the likelihood of environmental changes and decreased profitability for MNEs (Tsang, 2005). Thus, MNEs prefer WOS (versus JV) and greenfield (versus acquisition) to protect their benefits from higher host-market risk.

Parent-Foreign Affiliate Differences
Unrelated entry is used to describe the differences between the activities of parent firms and subsidiaries. When an investment occurs in a new industry, an MNE chooses an entry mode that helps it obtain specific knowledge (Larimo, 2003). MNEs prefer JV (versus WOS) and acquisition (versus greenfield) when they enter a new industry market. Relative size is the ratio of subsidiary size to parent size, which can be measured using dimensions such as employment size (e.g. Delios & Beamish, 1999) or investment size (e.g. Dikova & van Witteloostuijn, 2007). A large relative size means that a foreign subsidiary may not receive sufficient resources from its parent firm and may experience a shortage of resources (Delios & Beamish, 1999;Hennart & Park, 1993). When the relative size is large, JV (versus WOS) and acquisition (versus greenfield) can improve the flow of resources.

Home-Host Country Differences
Cultural distance is a frequently studied topic in the existing research, particularly as a factor influencing entry mode choices (Pla-Barber et al., 2010;Shane, 1994). Opposing results from the effects of cultural distance on entry mode choices have been found in empirical studies. Although some meta-analyses have examined the opposing results (e.g. Magnusson et al., 2008;Tihanyi et al., 2005), this paper tests this factor using a more integrated framework. Cultural distance leads to a misunderstanding of host countries and may increase an MNE's transaction costs. Thus, MNEs prefer JV (versus WOS) and greenfield (versus acquisition), which decrease costs. Geographic distance is the actual distance between the home and host locations (Malhotra et al., 2011), which generates barriers to transportation and communication (Slangen, 2011). MNEs may choose a high control mode to reduce costs from geographic barriers. Thus, MNEs prefer WOS (versus JV) and greenfield (versus acquisition) when they enter international markets with greater geographic distances.

3
Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes

Time of Study
The studies in our sample were published between 1988 and 2022. Over more than 30 years, the investment and macro-economic environments have changed significantly (Giachetti et al., 2019). A changing environment can affect an MNE's entry behaviour ; therefore, it is critical to examine whether the effects of antecedents change over time (Magnusson et al., 2008). Moreover, it is meaningful to consider relationships using a dynamic perspective that provides more insight and reveals the intervening effects of time.

Home-Country Type
The development of a home country is an important factor affecting an MNE's international entry (Cui & Jiang, 2012). Previous studies have found that a home country can alter the relationships between antecedents and entry mode choices (Magnusson et al., 2008). However, the effects of a home country are different for different antecedents (Zhao et al., 2004). Thus, it is necessary to examine the home country in the context of previously neglected antecedents.

Journal Level
Studies on entry mode choices have been published in a wide range of journals. However, the number of published papers varies across different journals (Dikova & Brouthers, 2016). Giachetti et al. (2019) discussed the need to consider the heterogeneity of the reported results among journals with different characteristics. As such, seeking to further understand the influence of different journals, the moderation effect of the journal level is examined using a bivariate meta-analysis.

Entry Mode Measurement
A measurement error may exist when entry mode choices are measured differently. Not all antecedents' effects vary with different measurements of entry mode choices (Zhao et al., 2004). To find out more about the stability of bivariate meta-analytic relationships, the measurement of the entry mode choice should be examined in connection with both antecedents and outcomes.

Performance
Some empirical studies have only discussed performance after market entry. MNEs select the entry mode that provides the best return on investment (Brouthers et al., 1999), and many empirical studies have demonstrated that entry modes affect MNEs' performance (Ripollés & Blesa, 2012;Woodcock et al., 1994). However, no meta-analysis has examined both antecedents and outcomes when examining entry mode choices. Although some meta-analyses of antecedents have considered performance, they regarded it as part of the international process and only examined the influence of the antecedents on different parts of the international process (e.g. entry mode and performance). This study analyses the relationships between the antecedents and the entry mode choices as well as the relationships between the entry mode choices and performance, thus enhancing the understanding of entry mode choices.

Survival
Although performance is a popular means of evaluating MNEs' operations, survival is a more straightforward measure of successful foreign market entry. Survival is an important dimension of an MNE's subsidiary's performance measurement (Delios & Beamish, 2001). Some studies have examined the relationship between entry mode choice and foreign entry survival. For example, Chung and Beamish (2005) asserted that a fully owned greenfield is more likely to survive than a shared greenfield or fully owned acquisition. As most studies only focused on performance, they were unable to consider which firms failed in a foreign market (e.g. Holtbrügge & Berning, 2018;Trąpczyński & Gorynia, 2017). However, considering the survival of a firm is important, as it complements the analysis of post-entry outcomes by examining both surviving and failed entries. Hence, this study examines whether entry modes affect survival and performance in different ways.

Sample
All relevant studies should be identified when conducting a meta-analysis. Hunter and Schmidt (1990) noted that a meta-analysis should include all available data regardless of the quality of the data. Thus, relevant studies were searched and screened using the procedure outlined below (see also Fig. 2). Specifically, the identification stage included three steps. First, we followed several review articles about entry mode choices to define the key terms, such as entry mode, mode of entry, market entry, entry strategy, entry decision, entry form, entry type and establishment mode. Based on this, online databases (e.g. EBSCO, ProQuest, JSTOR, ScienceDirect and Web of Science) were searched using those keywords. Second, we searched for the key terms across journals focusing on international management, international marketing, general management and marketing (e.g. Journal of International Business Studies, Journal of Management, International Business Review, Journal of World Business and Journal of International Marketing). Third, we also used an AI approach (i.e. researchrabbit) to search for relevant literature and chased backward and forward citations based on the reference lists obtained from the previous steps. With all the above efforts, we obtained 2,159 studies. At the screening stage, we 1 3 Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes first retained the studies that (1) were empirical studies rather than reviews or theoretical papers; (2) focused on international entry rather than domestic market entry; (3) analysed either ownership, establishment mode choice or both; (4) informed the sample sizes and correlation coefficients for empirical analysis; (5) were not duplicate studies; and (6) met the coding criteria explained in the next section. Then, among the retained studies, we excluded those that used overlapping samples and had no constructs for this study. Finally, at the inclusion stage, 230 primary studies were included as the sample for this meta-analysis.

Coding and Analysis Procedures
All the selected studies were carefully read before coding. A coding protocol was developed by reading relevant papers (Lipsey & Wilson, 2001). Three authors discussed the protocol and reached a consensus on the discrepancies. All the selected studies were coded with the protocol to extract data on the correlation coefficients, sample sizes, measurement reliabilities, times of studies, home countries, journal levels, and entry mode measurements. Any further discrepancies were discussed and solved during coding. This study used only the correlation coefficient as the effect size index because correlations were provided in most of the entry mode studies. Some antecedents were not included in the analysis due to insufficient data. The studies on these antecedents were abandoned after coding.  (Delios & Beamish, 2001) Performance Performance is measured with (a) financial measurements (e.g. sales level, profitability and sales growth) (Riaz et al., 2014), (b) non-financial measurements (e.g. market share, marketing, reputation and market access) (Brouthers, 2002), (c) survey items of manager perceived performance (Giachetti, 2016) Entry mode choices (e) dummy variable, 0 for small and medium firms and 1 for large firms (Arslan & Larimo, 2011;Filatotchev et al., 2007) Parent profitability Parent profitability is measured with (a) the ratio of return to assets; (b) the ratio of return on equity; (c) the market-to-book ratio (Wooster et al., 2016) 1 3 Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes  (Lu et al., 2018;Xie, 2014) Host-market size Market size relies on (a) the GDP; (b) the population; (c) survey items of the host market (Sartor & Beamish, 2018;Yamanoi & Asaba, 2018) Host-market growth Host-market growth is measured with (a) the GDP growth of host country; (b) survey questions on perceived market growth; (c) the average annual growth rate of shipments of the four-digit U.S. industry in the 5 years Hennart & Park, 1993) Host-country risk Host-country risk relies on (a) 100 minus Euromoney country risk index; (b) reverse coded international country risk guide published by the PRS Group; (c) survey items of perceived risk (Cui & Jiang, 2009;Cuypers et al., 2015) Unrelated entry Unrelated entry is measured with (a) 1 if the subsidiary operated in a different industry sector from the parent company and 0 otherwise; (b) 1 minus the ratio of the number of industries in which both firms operate to the total number of distinct industries in which they operate (Chang & Rosenzweig, 2001;Cuypers et al., 2015) Relative size Relative size is measured with (a) the ratio of the employment of the foreign affiliate over that of the parent; (b) the ratio of the sales of the foreign affiliate over that of the parent; (c) the ratio of the assets of the foreign affiliate over that of the parent; (d) the ratio of total investment in the affiliate over total assets of the parent firm (Dikova & van Witteloostuijn, 2007;Slangen & van Tulder, 2009)  Geographic distance Geographic distance is measured with (a) the actual distance in kilometres between the capital cities of the home and host country; (b) the great-circle distance in kilometres between the midpoints of the home and host country (Boellis et al., 2016;Cuypers et al., 2015) When multiple studies with overlapping or identical samples provided the same relevant variables for analysis, the study with the larger sample size was included to ensure independence between effect sizes (Klier et al., 2017). However, when a study's multiple correlations were from several independent samples, they were coded separately to prevent bias or missing data (Johnston et al., 2018). As a result, the coding procedure revealed 1499 effect sizes from 230 studies.
In this study, the estimated combined effect was the mean correlation coefficient of all the correlation coefficients in the sample (Kolev, 2016). Before the combined effects were calculated, the average value was used for calculating the combined effects when the empirical studies used multiple measurements for the variables and provided more than one correlation for the same pairwise relationship (Johnston et al., 2018;Kolev, 2016).
Moreover, following Hunter and Schmidt's (1990) method, the artefacts from the measurements and samples were corrected for calculating the combined effects. Measurement errors were adjusted by dividing the correlation coefficients by the product of the square root of the reliabilities of the paired constructs. The reliability of the variables with objective measurements was set at 0.8, which was consistent with the previous studies (Dalton et al., 1999). Most studies using subjective measurements reported reliability. For studies that did not provide reliability, sample weighted mean reliability was used (Geyskens et al., 1998).
Furthermore, the reliability-adjusted correlation was weighted with the sample size and standardised to account for different metrics in the original scales for calculating the combined effect size (Borenstein et al., 2009;Johnston et al., 2018). The random-effect model was chosen for analysing these independent studies because it is appropriate for distributed true effect sizes, while the fixed-effect model is generally used for analysing studies that share a common effect size (Borenstein et al., 2010).
The outliers were identified with iterative methods (Viechtbauer & Cheung, 2010). We calculated the combined effects with and without deleting the outliers, and no significant differences were found (see the Appendices 1, 2). In addition, the removal of outliers may lead to the loss of generalisability of meta-analyses, as they may be legitimate effect sizes drawn by chance from the ends of a distribution and may be attributed to sampling differences (Dalton & Dalton, 2005;Steel et al., 2021;Tabachnick & Fidell, 2019). Therefore, we kept the outliers for the analysis, as Lee and Madhavan (2010) did.
The publication bias was examined with two complementary methods, considering that no single method is able to assess publication bias accurately in any situation (Ferguson & Brannick, 2012). Specifically, the Begg & Mazumdar test and the trim-and-fill r method were used, with the former assessing whether a publication bias existed (Begg & Mazumdar, 1994) and the latter providing the publication bias-adjusted effect size (Duval & Tweedie, 2000a, 2000b. For the judgement calls involved in the trim-and-fill r method using CMA3, we first select to look for missing studies where 'to left of mean' if the estimated correlation coefficient is positive; and select 'to right of mean' if the estimated correlation coefficient is negative (Borenstein et al., 2009). Next, we select random effect model to looking for missing studies as advocated by Tweedie (2000a, 2000b). As shown in Tables 6, 1 3 Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes Table 6 Antecedent-ownership mode correlations Data is calculated using CMA3. k = number of effect sizes; N = combined sample size; r = mean sample size weighted correlation; ρ = estimated true score correlation corrected for measurement error and weighted for sample size; CI = confidence interval, Q = Q-statistic for heterogeneity; I 2 = the ratio of variance from heterogeneity. "UC" means that the estimated effect size remains unchanged as no potential studies are missing. The numbers in the parentheses denote the potential missing studies *p < 0.05, **p < 0.01 Ownership mode (WOS = 1) In addition, the small number of potential missing studies in the trim-and-fill r method indicated that publication bias was not an issue.

Antecedent-Entry Mode Relationship
The correlations between the ownership mode choice and its antecedents are shown in Table 6. Specifically, the effects of host-country risk, unrelated entry, cultural distance and geographic distance on the ownership mode choice (WOS) were significantly negative (ρ = − 0.074, p < 0.05; ρ = − 0.072, p < 0.01; ρ = − 0.060, p < 0.01; and ρ = − 0.051, p < 0.01, respectively). The effects of host-country experience, international experience, R&D intensity, advertising intensity and host-market size on WOS were significantly positive (ρ = 0.073, p < 0.01; ρ = 0.076, p < 0.01; ρ = 0.066, p < 0.01; ρ = 0.085, p < 0.05; and ρ = 0.070, p < 0.01, respectively). All other factors had non-significant effects on WOS. The Q-statistic and its p-value were used to test for the homogeneity of the null hypothesis, the true dispersion is exactly zero. All the Q-statistics in Table 6 were significant, indicating that the true effects were heterogeneous. In addition, the I 2 can address what proportion of an observed variance is real, and an I 2 near zero indicates that almost all of the dispersion may be attributed to random error (Borenstein et al., 2009). As all the I 2 s were far from zero in Table 6, we knew that some of the variances were real and could potentially be explained by examining the moderators via subgroup analysis or meta-regression. The results from the trim-and-fill r estimate and the Begg & Mazumdar test indicated that publication bias was not an issue in our study, as already explained above. Regarding the choice of establishment mode (Table 7), host-country experience and unrelated entry had significant, negative influences (ρ = − 0.122, p < 0.01; and ρ = − 0.084, p < 0.01, respectively) on the establishment mode choice (greenfield). The influences of R&D intensity and host-market growth were positive (ρ = 0.109, p < 0.01; and ρ = 0.160, p < 0.01, respectively). The significant Q-statistic and large non-zero I 2 indicated that heterogeneity existed in all the significant relationships, and some of the variances were real and could be explained by the moderators.
As the results in Tables 6 and 7 indicate, there was a need to explain the variance by detecting the potential moderators of the relationships between the antecedents and the entry mode choices. Given the information in the primary studies, this study only tested four sample-and measurement-related moderators: the time of study, home-country type, journal level and entry mode measurement (see Tables 8,  9). Some of them are commonly included in meta-analytical studies (Magnusson et al., 2008;Zhao et al., 2004). The time of study was measured as the actual publication time of each primary study. The home-country type equalled one for developed home countries and zero otherwise. The journal level was measured based on the academic journal guide (AJG) 2021 from five to one. The measurement of the  Tables 8 and 9 show the moderation effects of the time of study. The moderation effects were significant for the relationship between ownership mode and host-country experience, international experience, parent diversification, parent profitability, host-country risk, unrelated entry and cultural distance, respectively. This meant that these relationships changed significantly over time.
Regarding the relationships between the establishment mode and its antecedents, the moderating effects of the time of study were significant for R&D intensity, parent profitability and relative size. Thus, the influences of these antecedents on the establishment mode choice weakened or strengthened over time.
The second column in Tables 8 and 9 presents the moderation effects of the home-country type. For the ownership mode choice, the moderation effects were significantly positive on unrelated entry but significantly negative on relative size and geographic distance. This meant that unrelated entry had a stronger impact on the ownership mode choice in a developed home country than in a developing home country, whereas relative size and geographic distance had weaker impacts on the ownership mode choice in a developed home country than in a developing home country. For the establishment mode choice, only 12 out of the 15 antecedents were tested for the home-country type's moderation due to a lack of data; among them, three were significant. The home-country type positively moderated the influences of host-country risk and cultural distance but negatively moderated the effects of advertising intensity.
The third column in Tables 8 and 9 presents the moderation effects of the journal level. The journal level negatively moderated the effect of parent profitability on the ownership mode choice. This meant that the impact of parent profitability on the ownership mode choice was weaker in higher level journals. When the establishment mode choice was examined, the journal level negatively moderated the effect of advertising intensity but positively moderated the effect of host-country risk and geographic distance. These indicated that the existence of preferences among different journal levels may explain the non-significant main relationships.
The moderation effect of the entry mode measurement was examined only for the ownership mode choice. When the ownership mode choice was measured as a dummy variable (compared with a continuous variable), the effects of parent size, host-market growth and unrelated entry on the ownership mode choice were stronger, while the effect of the host-market size was weaker.

Entry Mode-Outcome Relationship
Tables 10 and 11 show the meta-analytical effects of entry mode choices and postentry outcomes. Both the ownership mode choice and the establishment choice had significantly positive effects on survival (ρ = 0.086, p < 0.01; ρ = 0.172, p < 0.01), while only the ownership mode choice was significantly positive on performance (ρ = 0.035, p < 0.05). The results showed that WOS and greenfield usually generated higher survival rates. Moreover, WOS also indicated better performance. Tables 12 and 13 display the moderating effects of the time of the study, home-country type, journal level and entry mode measurement. The time of study negatively moderated the effect of the establishment mode choice on performance. The home-country type positively influenced the effect of the establishment mode choice on performance. The journal level had no significant influence on the entry mode choice -outcome relationship. The measurement of ownership negatively moderated the positive effect of the ownership mode on performance.

Mediating Effects
We used MASEM, which combines meta-analysis and structural equation modelling (SEM), to test the mediating effects of the entry mode choices, as meta-analysis only tests direct connections of bivariate relationships and is unable to estimate mediation effects (Bergh et al., 2016). SEM provides an accurate analysis by considering the dependence among variables, analysing a merged larger dataset, and providing information about a model's degree of fit (Bergh et al., 2016;Cheung, 2013Cheung, , 2021. As shown in the published studies, SEM is tested against the average correlation matrix formed with the results from meta-analysis (Cheung et al., 2019). Therefore, relationships neglected in previous empirical studies can be examined with a more comprehensive model with SEM.
Specifically, the existing entry mode studies have established the relationships between entry mode choices and outcomes by testing the fitness of an entry mode for a firm in a specific environment (Brouthers et al., 2003;Kim & Gray, 2008). However, though the entry mode affects the outcomes (i.e. different modes provide different levels of control and resources), it does not affect the outcomes through fitness. Other studies have examined the moderating effects of entry modes but not the mediating effects (Trąpczyński & Gorynia, 2017;Wang & Schaan, 2008). Thus, Table 10 Entry mode-survival correlations Data is calculated using CMA3. k = number of effect sizes; N = combined sample size; r = mean sample size weighted correlation; ρ = estimated true score correlation corrected for measurement error and weighted for sample size; CI = confidence interval, Q = Q-statistic for heterogeneity; I 2 = the ratio of variance from heterogeneity. "UC" means that the estimated effect size remains unchanged as no potential studies are missing. The numbers in the parentheses denote the potential missing studies *p < 0.05, **p < 0.01. MASEM is a useful measure to test the indirect effects of the antecedents on the outcomes through the entry mode. This paper tested the mediating effects, as shown in the framework presented in Fig. 1. The path models were specified in Amos, as shown in Figs. 3, 4, 5, 6. In the mediating models, only seven antecedents were tested (cultural distance, hostcountry experience, international experience, R&D intensity, advertising intensity, parent size and unrelated entry) due to the accessibility of the data and the model's fitness. The sample size weighted correlations that were calculated by the meta-analysis among the constructs served as the basis for SEM (see Table 14). The sample size weighted mean reliabilities (Cronbach's α) on the diagonal were used to correct the measurement error in SEM. The sample size for computing the significance levels of the coefficients was the harmonic mean of the sample sizes for the correlations involved (Bergh et al., 2016;Tan & Sousa, 2013). The two general methods used in modelling are maximum likelihood (ML) and generalised least squares (GLS), which demand fewer restrictions on the data (Kelloway, 1998). In this paper, Amos's default method ML was chosen to estimate the models. GLS was also examined and showed no significant differences from ML.
The simple path models are shown in Figs. 3 and 4. The ownership mode choice and the establishment mode choice were analysed together for three reasons. First, based on the existing empirical studies, both the ownership mode and the establishment mode affect firms' post-entry outcomes. Second, there are interactive influences between the two entry modes. Third, models that test the ownership mode and the establishment mode simultaneously fit the data better.   Furthermore, the full models in Figs. 5 and 6 examined the direct relationships between the antecedents and the outcomes. Considering the direct and indirect relationships improved the examination of the mediation effects. The significant paths from the antecedents to the outcomes through the entry mode choices showed the existence of mediation effects. Figure 5 presents the coefficients of the mediating test for survival, and Fig. 6 presents the coefficients of the mediating test for performance. As the aim was to identify the mediation effects of the entry mode choices between the antecedents and the outcomes, only the significant paths that showed the antecedents affecting the outcomes (direct and indirect effects) were illustrated as solid lines in these figures. The other paths were illustrated as dashed lines.
As shown in Fig. 5, both the ownership mode and the establishment mode had significant effects on survival. Similarly, all the antecedents could affect survival indirectly through the two entry mode choices. All the antecedents also had direct effects on survival except R&D intensity. In Fig. 6, the ownership mode had a positive effect and the establishment mode had a negative effect on performance. All the antecedents could affect performance indirectly through the two entry mode choices. Except for unrelated entry, six antecedents also illustrated direct influences on performance.
The results showed that establishment mode and ownership mode worked differently as mediators. Most of the antecedents had different indirect influences through different entry mode choices. Moreover, most of the antecedents had reverse effects between the direct and indirect influences on the outcomes. Unrelated entry could only affect performance through the entry mode choices. R&D intensity also could only affect survival through the entry mode choices. Among the rest of the antecedents that could affect the outcomes directly, only parent size showed a consistent influence on performance and survival.

Robustness Test
Following Koh et al. (2019), we used full information MASEM (FIMASEM) as a robustness check for the results from the MASEM in the last section. Although FIMASEM does not work well in statistics and goodness-of-fit, it allows us to consider the inevitable heterogeneity among the studies in the SEM stage , which is not considered by MASEM. Specifically, in FIMASEM, the estimated true score correlation (corrected for the measurement error and weighted for the sample size) and standard deviation (tau) were used in 5,000 iterations to analyse the model (see Table 15). The mean path coefficients and 80% credibility interval are shown in Figs. 5 and 6 below the results from MASEM. The mean coefficients from FIMASEM presented similar results as the coefficients from MASEM.

Relative Importance Test
Seven antecedents and two entry mode choices were analysed in MASEM. However, which significant antecedents (or entry mode choices) had stronger influences was Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes  Table 15, the relative importance of the antecedents to the entry mode choices and that of the entry mode choices to the outcomes were examined. The results are shown in Tables 16 and 17. Regarding the ownership mode choice, advertising intensity had the strongest influence, and parent size had the least influence. In terms of the establishment mode choice, advertising intensity also had the strongest influence, while international experience had the least influence. The relative importance of the seven antecedents was more balanced for the ownership mode choice than for the establishment mode choice. In the analysis of the outcomes, the establishment mode choice had a much stronger influence on both performance and survival than the ownership mode choice.

Discussion and Implications
This study provided a comprehensive synthesis of the antecedents and outcomes of entry mode choices to address the current research gaps in the literature. By combining the existing empirical results concerning these antecedents and outcomes, this study found that the choice of entry mode has significant implications that need to be considered. These findings establish a solid foundation for future research on the subject.

Implications for Theory
Based on the meta-analysis, this paper discussed 15 antecedents and two post-entry outcomes related to the ownership mode choice and the establishment mode choice. This paper examined four groups of antecedents: parent-firm characteristics, hostcountry characteristics, parent-foreign affiliate differences and home-host country differences. In contrast to the previously published meta-analyses, this paper compared two entry mode choices (ownership mode and establishment mode), examined a greater number of antecedents and analysed the entry mode choice as a mediator.
To further highlight how this study extends and complements earlier meta-analyses, a comparison is presented in Tables 18 and 19.
The published meta-analyses on antecedent -entry mode choice relationships only tested a small number of antecedents for the two entry mode choices. This study went further by considering a larger number of antecedents (15), which were categorised into four groups (parent-firm characteristics, host-country characteristics, parent-foreign affiliate differences and home-host country differences). Thus, this paper greatly improved both the quantity and range of antecedents considered. The two published meta-analyses on entry mode choice -outcome relationships only examined the choices among different ownership levels. This paper extended the existing analyses by evaluating both the ownership mode choice and the establishment mode choice. Five implications for the existing literature are explained below.
First, this study has contributed to the current research by investigating whether the two entry mode choices are affected by different antecedents. The ownership mode choice is more closely related to MNEs' control (Dikova & van Witteloostuijn, 2007); therefore, more significant parent-firm characteristics of the ownership mode choice revealed the close relationship between these characteristics and MNEs' control level. However, the establishment mode choice describes how MNEs build their foreign entities (Dikova & van Witteloostuijn, 2007), and only two of eight parentfirm characteristics had significant influences on building foreign entities. Additionally, for the 10 antecedents that significantly affected the entry mode choices, only three antecedents were significant for both entry mode choices. The remaining seven antecedents were significant for either the ownership mode choice or the establishment mode choice. The differences in the antecedents' effects emphasised recognising the influential antecedents in different entry mode decisions. Notably, as is shown in Table 18, whereas the majority of our findings are consistent with those of previous meta-analyses, a few differences exist. One obvious difference is about cultural distance. Although our study and Zhao et al. (2004) found that it has negative influence on ownership mode choice, Beugelsdijk et al. (2018) and Tihanyi et al. (2005) found no significant influence; although Beugelsdijk et al. (2018) found that cultural distance has positive influence on establishment mode choice, we found no significant impact. Possible explanations for the difference include (1) our sample size for the cultural distance-ownership mode is larger than previous studies, which allows us to detect a significant effect; and (2) the impact of cultural distance on the establishment mode is positively moderated by home-country type (see Table 8). Our study found that the impact of cultural distance on establishment mode is positive for developed home country and is not significant for developing home country.  As more and more studies focus on developing countries, the sample cases of developing home countries account for a larger proportion of our reviewed studies than that of Beugelsdijk et al. (2018), this may explain why we found non-significant impact of cultural distance whereas they found positive impact. Another difference is that Klier et al. (2017) found parent profitability to be positively associated with establishment mode, whereas we found no significant association. In addition to the difference sample size, this difference may be also due to the different measurement of parent profitability.
Second, this research demonstrated that both the ownership mode choice (WOS versus JV) and the establishment mode choice (greenfield versus acquisition) can affect post-entry outcomes significantly. Though the establishment mode choice was non-significant for performance in the bivariate meta-analysis, it became significant when other variables were controlled in SEM. It showed the importance of analysing not only the ownership mode choice but also the establishment mode choice in post-entry outcomes. Based on the relative importance analyses, the establishment mode choice was relatively more critical than the ownership mode in affecting postentry outcomes. In addition, the effects of the ownership mode choice on performance and survival had the same direction, while the establishment mode choice had opposite influences on two post-entry outcomes. Acquisition, which is beneficial for better performance, decreases the survival likelihood. As MNEs may sacrifice subsidiary survival for overall corporate success (Kim & Hwang, 1992), MNEs can achieve high performance accompanied by low survival likelihood using acquisition. Notably, as is shown in Table 19, inconsistent with this study, Giachetti et al. (2019) and Zhao et al. (2017) found the impact of ownership mode on survival to be non-significant. This may be due to the difference in measures: we clearly distinguished between ownership mode and establishment mode, whereas they did not. In addition, this study has a relatively larger sample size, which makes it more likely to detect significant impact in the analysis.
Third, the test of moderation also revealed some significant effects. The significant time of the study showed decreased influences for some antecedents (e.g. hostcountry experience and R&D intensity) and increased influences for others (e.g. cultural distance). The significant home-country type revealed that a developed home country strengthens the influences of several significant antecedents (e.g. unrelated entry and geographic distance). The home country characteristics that have been generally ignored are critical to international entry. The significant moderation effects of the journal levels demonstrated the resulting stability among different levels of journals. The significant main relationships were stable among journals with non-significant moderation effects. The stability presented less possibility of bias due to the publication of these relationships. The examination of the moderation effects of the entry mode measurement showed the importance of using suitable measurements in empirical studies. The continuous measurement of the ownership mode choice that is regarded as containing more information strengthened the effects of some antecedents (e.g. unrelated entry) but weakened others (e.g. host-market size). These significant moderation effects may have been derived from developments in the international environment, improved institutional policies and diversified international cooperation. The differences may have also originated from diversified and continuously enhanced methodologies. The moderation effects presented a dynamic perspective on entry mode research and supplemented the traditional theoretical view (i.e. IT and RBV). This combined perspective yielded new insights into entry mode choices (Surdu & Mellahi, 2016). Fourth, this study combined antecedents, entry mode choices and post-entry outcomes to present a dynamic process of international entry. As Brouthers (2002) and Shaver (1998) pointed out, an exogeneity problem exists when post-entry outcomes are compared without considering antecedents. A given type of entry mode choice is not a valid predictor of outcomes. The analysis of post-entry outcomes must consider the mechanisms of entry modes and how the antecedents affect the outcomes through entry mode choices. The examination of the entry mode choices' black box revealed the mechanism of the effects. Considering the mediation effects provided an opportunity to examine the mediation role of the entry mode choice in a theoretical model, thereby extending the boundaries of the theories used as moderators. Moreover, the significant indirect effects of the antecedents on the post-entry outcomes expanded the perspectives on some causal relationships that were previously ignored. Some antecedents had positive direct effects but negative indirect effects on post-entry outcomes. The opposite direct and indirect influences showed the importance of examining the two effects together.
Fifth, this research explored the theoretical connection among antecedents, entry mode choices and post-entry outcomes using a comprehensive model. We investigated those connections by combining the TCE and RBV. Moreover, the combination of the two theoretical lenses (TCE and RBV) allows us to make a more comprehensive analysis of the antecedents of entry mode choices and their effects on post-entry outcomes. The results emphasised the importance of including more antecedents in a single model and considering the interactions between them. Most existing meta-analyses use established models that are based on only one theory. For instance, Zhao et al.'s (2004) model was based on TCE and only examined transaction cost determinants, and Klier et al. (2017) analysed two resources using the 1 3 Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes RBV. However, many antecedents are ignored within specific theoretical perspectives. Different theories examine different antecedents and offer varied perspectives for explaining international entry, making it critical to select and combine theories (i.e. using two theories in an empirical study) to analyse the importance of antecedents within a more comprehensive model. A clear purpose for international entry (e.g. better performance, firm strategy or resource acquisition) helps identify significant antecedents, while the combination of international aim, a suitable theoretical perspective and critical antecedents helps to create a more complete model and provides more stable results.

Implications for Practice
This paper also offers some practical implications for managers. First, managers are advised to focus on different antecedents when considering different entry mode choices. When making the ownership mode choice (WOS versus JV), managers should use WOS when MNEs possess high host-country experience, international experience, R&D intensity and advertising intensity or invest in a host country with a large market size. On the other hand, managers are suggested to use JV when there is a high host-country risk, cultural distance, geographic distance or an unrelated entry. In relation to the establishment mode choice (greenfield versus acquisition), managers are advised to use greenfield when MNEs have high R&D intensity or are entering a host country with high growth. Acquisition seems to be more appropriate when MNEs have high host-country experience or are conducting unrelated entries.
Second, managers must be aware that the entry mode choice will have a different impact on the performance and survival of the firm. In the case of the ownership mode choice, WOS is a better choice for both performance and survival. However, managers should be more alert when choosing between greenfield and acquisition. If they wish to focus on higher survival likelihood, greenfield appears to be the more appropriate establishment mode. However, acquisition seems to be a better option if they want to pursue high performance.
Third, it is important for managers to be aware of the mechanisms of the entry mode choices underlying the relationships between the antecedents and the outcomes. Managers cannot achieve better outcomes by focusing only on entry mode choices. The antecedents may lead to worse outcomes if they fail to clarify how such antecedents affect the outcomes through the entry mode choices. When the antecedents have positive (or negative) direct impacts on the outcomes, managers may enhance (or mitigate) the antecedents' direct effects with clear knowledge of the mechanisms of the entry mode choices. However, the mediation effects of the entry mode choices vary with different antecedents. Therefore, every antecedent should be treated carefully on how it affects the outcomes through the entry mode choices.

Limitations and Suggestions for Future Research
This meta-analysis has some limitations, offering directions for future research. First, this study examined 15 antecedents, which is a small subset of all the factors 1 3 affecting entry mode choices. More studies should be conducted to explore more antecedents. Future studies should also reveal the different influences of antecedents on different entry mode choices. The clarification of these relationships will make it easier to recognise influential antecedents in different entry mode decisions.
Second, some antecedents were examined aggregately based on their components, e.g. cultural distance (consisting of Hofstede's dimensions) and host-country risk (consisting of economic, political and structural dimensions), in this paper. However, the examination of aggregate influences may lead to non-significant effects (Brouthers & Brouthers, 2001). A few empirical studies have outlined different components with varied effects on entry modes (Magnusson et al., 2008;Xie, 2014). This paper failed to analyse different components due to data availability. Future studies are suggested to examine the individual effects of different components on entry mode choices for deeper understanding.
Third, this meta-analysis discussed two outcomes based on existing empirical studies. Besides performance and survival, MNEs enter international markets for various international aims (e.g. marketing, distribution, seeking resources and avoiding regulations). Though these international aims have been considered in entry mode choices (e.g. Kim & Gray, 2008;Yiu & Makino, 2002), how the chosen mode is beneficial for fulfilling the aims is still unclear. Future studies should emphasise whether MNEs' aims are achieved after their entry mode choice for better implementation in international strategy.
Fourth, both bivariate meta-analysis and FIMASEM showed the existence of heterogeneity in the main relationships examined in this meta-analysis. This paper tested four moderators, but many other moderators may also affect the process of entry mode choices (e.g. MNEs' strategies, industry clusters and state ownership). Such limitations invite future studies to investigate more moderation conditions for a deep understanding of how and why direct relationships change.
Fifth, only seven antecedents were contained in the path study due to limited data. More empirical studies and meta-analyses should discuss the post-entry outcomes (e.g. survival and performance) while testing the influences of antecedents on entry mode choices. Analysing international entry as a complete process can contribute to the understanding of entry modes and thus improve the ability to make suitable choices.

Conclusion
Previous meta-analyses have only found limited antecedents with regard to entry mode choices. Based on various empirical studies, this paper examined previously ignored antecedents that can affect both the ownership mode choice and the establishment mode choice. By testing the antecedents with MASEM, this paper highlighted the importance of investigating the combined effect of different antecedents. This study outlined two outcomes after entry and discussed the different influences of entry mode choices on outcomes. Moreover, the examination of four moderators 1 3 Entry Mode Choice: A Meta-Analysis of Antecedents and Outcomes revealed significant moderation effects that contributed to mixed empirical results. This study also analysed the mediation effects of entry mode choices. These dynamic insights yielded more accurate results for post-entry operations by examining the whole process of international entry from the antecedents to the outcomes. In conclusion, this paper's findings emphasised the importance of examining international entry using a comprehensive model that considers several aspects of entry (e.g. antecedents, outcomes and mechanisms). Therefore, more critical attention is needed to recognise the antecedents and outcomes and build systematic models.

Appendix
See Appendix 1 and 2.