International diversification translating into financial performance for European listed multinational organizations

(1)

International diversification translating into financial performance for

European listed multinational organizations

Final version January 25, 2016

Faculty of Economics and Business MSc Business Administration Track: International Management

Nathalie Duijvesteijn MSc. Student number: 6030297

Email: nathalieduijvesteijn@gmail.com First supervisor: E. Dirksen MSc.

(2)

Statement of Originality

This document is written by Student Nathalie Duijvesteijn who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

The effect of multinationality on firm financial performance is a central question in international business and strategy literature. Many scholars have studied this relationship, but failed to produce a consistent set of empirical results, with different studies finding evidence for different shapes of the multinationality-performance (M-P) relationship. Some studies find a positive linear model, suggesting that more international diversification increases firm-level performance (e.g. Grant, Jammine & Thomas, 1988; Han, Lee & Suk, 1998), while others find curvilinear models incorporating both costs and benefits associated with multinationality, such as the inverted U-shaped model (e.g. Gomes & Ramaswamy, 1999; Hitt et al., 1997), the U-shaped model (e.g. Capar & Kotabe, 2003), or the horizontal S-shaped model (e.g. Lu & Beamish, 2004).

Where many scholars have tried to identify the M-P relationship in its magnitude or functional form, Verbeke, Li and Goerzen (2009), among others, criticize such approaches and argue that no valid theoretical rationale supports a generalizable M-P relationship. The possible diversity in M-P linkages depends on the variety of strategic motivations for foreign direct investment (FDI), the environmental complexity and the organizational complexity facing multinational enterprises (MNEs). These three key parameters underlie the substance of the multinationality concept, and each of them might be distinctly related to firm performance (Verbeke et al., 2009).

Internalization theory provides an explanation for FDI, or MNEs in general, because MNEs are created when firms can increase their value by internalizing markets for certain of their intangible assets across national borders (Buckley & Casson, 2003). Several studies find evidence for the moderating effect of firm specific advantages (FSAs) on the M-P relationship. In line with internalization theory, intangible assets of individual firms form their FSAs that lead to optimal multinationality paths differing among those individual firms.

(5)

Therefore, it might indeed be the case that there does not exist one generalizable M-P relationship.

Since most studies on the M-P relationship draw upon the internalization theory and transaction cost economics (TCE), this thesis aims to incorporate elements of the resource-based view (RBV) into TCE and internalization theory to identify optimal multinationality levels for individual firms. The RBV believes that firms are able to outperform other firms based on the valuable, rare, inimitable and non-substitutable resources the firms possess (Barney, 1991; Peng, 2001), and that this lies in the ability to coordinate diverse skills and integrate multiple streams of technology. By engaging in FDI, firms are able to acquire these skills and integrate technologies to create new FSAs. Multinationality also results in country-specific advantages (CSAs) and subsidiary country-specific advantages, and firms that successfully transfer these into the parent firm are able to outperform others and positively influence their financial performance.

Since most previous studies focus on multinational organizations from the U.S.A., this thesis uses European MNEs as sample. While incorporating important features of the RBV, this thesis aims to contribute to the M-P literature by answering the following research question:

How does international diversification translate into financial performance? Yang and Driffield (2012) suggest that regression analysis provides more reliable results than other approaches when studying the M-P relationship. Therefore, this thesis uses quantitative research to study the effect of international expansion on financial performance, using listed European multinational firms as sample. In order to capture the time-varying effects of international expansion, the M-P relationship is tested using panel data comprising European listed MNEs and their internationalization activities in the years 2007 until 2014. Several

(6)

control variables, such as firm size, firm age, industry dummies and sales growth, are included to control for firm-specific and industry-specific effects, among other factors.

The dependent variable is firm financial performance, measured by the return on assets (RoA), which is the ratio of a firm’s income to its total assets. The main independent variable is the degree of multinationality, indicating the extent of a firm’s foreign footprint. Two dimensions of the AAA-framework (Ghemawat, 2008) are used, namely adaptation and aggregation activities of firms, in a simplified model of moderated mediation. It is expected that R&D intensity, indicating a firm’s adaptation activities, works as a mediator between multinationality and performance, where the measure for multinationality indicates a firm’s aggregation activities. The control variables included in the model have moderating effects on either the direct link between multinationality and performance or the indirect link with the mediator R&D intensity, or both. The main findings show that multinationality significantly increases firm performance, but higher levels of multinationality negatively impacts on this M-P relationship, providing evidence for an inverted U-shaped model. Also, the results show that R&D intensity has a negligible low influence on the M-P linkage.

This thesis continues as follows. Chapter 2 analyzes the existing literature on the M-P relationship and the previous approaches undertaken to examine this relationship. The main empirical models and their critiques are discussed, after which the step is made away from one generalizable M-P relationship and more towards the different dimensions through which MNEs (attempt to) increase their financial performance. Chapter 2 ends with four hypotheses that follow from the existing literature, which will be tested using the data and methodology as described in Chapter 3. In Chapter 4 the results from the regression analysis are given and discussed, after which Chapter 5 discusses some limitations of this thesis and provides recommendations for future research. Chapter 6 ends this thesis with a conclusion.

(7)

2. Literature review

Many scholars have studied the M-P relationship in multiple ways. Some try to identify a direct relationship, while others examine moderating effects of other factors on the M-P relationship. This chapter covers the existing literature on the M-P relationship and the different approaches undertaken to examine this link. It first elaborates on the conceptualization of multinationality and performance, after which different approaches undertaken to examine the M-P relationship are discussed. It then continues with the different relationship patterns found in the M-P literature, along with their received critiques. The chapter ends with hypotheses following from the discussed literature.

2.1 Conceptualization of multinationality and firm performance

The degree of multinationality refers to the extent to which firms diversify the geographic scope of their business activities, in order to gain or increase competitive advantage. Firms with high levels of multinationality operate in many foreign countries and are thus highly geographically diversified. This construct has been referred to in the M-P literature in different terms: the degree of internationalization, multinationality, a firm’s foreign footprint, international expansion, geographic diversification, and international diversification (Hult, 2011). Since all these terms refer to the same construct, for simplicity this thesis mainly uses the term multinationality. The degree of multinationality is generally measured as the ratio of foreign sales to total sales (F/T). However, some studies (e.g. Rugman & Oh, 2010) challenge this way of conceptualizing, which will be discussed in the following sections.

Firm financial performance or firm profitability refers to a firm’s ability to expropriate value from its assets, and is generally measured as the firm’s return on assets (RoA). In international business research, firm performance can increase if the benefits of doing business abroad are outweighing the costs of it. Therefore, firms should formulate their

(8)

international strategies to create an optimal balance between the exploitation of revenue-increasing FSAs and the limitation of costly foreign operations (e.g. de Jong & van Houten, 2014).

2.1.1 Benefits and costs associated with multinationality

The way multinationality can impact on a firm’s overall performance depends on the benefits and costs associated with international diversification. Firms can increase their performance with multinationality due to exploitation and exploration benefits (Lu & Beamish, 2004). Exploration benefits stem from the advantages from tapping into local resources, so resources in foreign markets that are not or less available in the domestic market, and the use of foreign knowledge (Caves, 1996; Singh, 2008). Exploitation benefits arise when firms exploit their domestically developed FSAs in international markets (Buckley & Casson, 1976; 2003; Hymer, 1976), or when they use arbitrage opportunities across multinational markets (Ghemawat, 2007).

Firms face costs of liability of foreignness and newness when operating abroad, which can become high when they enter increasingly dissimilar markets (Hymer, 1976; Zaheer, 1995). Liability of foreignness (LoF) is the cost of doing business abroad related to the disadvantages that MNE subsidiaries face to local competitors in a host country because of their non-local status (Zaheer, 1995). Firms operating in a foreign country may suffer losses in performance relative to local firms due to differences in for example the institutional environment, social norms and customs, local market practices, or local hostility (Peng, Wang & Jiang, 2008). LoF can be overcome by having certain FSAs (Hymer, 1976), something further elaborated on in the RBV. Liability of newness (LoN) is the cost of doing business abroad related to new operations and setting up new subsidiaries (Lu & Beamish,

(9)

2004). LoN can be more easily overcome when more experience is gained in doing business abroad.

Where the costs of LoF and LoF may diminish with further internationalization, the costs of transaction and coordination increase with higher levels of multinationality (Lu & Beamish, 2004). The costs associated with product diversification, such as incentive misalignment and coordination difficulties, also apply to geographic diversification. Also the costs based on TCE arguments (Williamson, 1979), such as the distortion or loss of information with transactions, fall under this category. In addition, higher degrees of multinationality increase managerial complexity and the associated costs of coordination (Hitt, Hoskisson & Kim, 1997).

2.2 Previous approaches in studying the M-P relationship

An approach that incorporates most of the costs and benefits associated with multinationality is the study of a horizontal S-curve M-P relationship, sometimes referred to as the general three-staged paradigm (Contractor, 2012). Lu and Beamish (2004), among others, hypothesize this shape of the M-P relationship, in which too low or too high levels of multinationality decrease firm performance and intermediate levels of multinationality positively impact on performance. Too low levels of multinationality, e.g. in the initial phase of international expansion, come with high costs of LoF and LoN that outweigh the benefits of internationalization. Further internationalization increases a firm’s knowledge and understanding about how to deal with subsidiaries in foreign countries and thus reduces the costs related to being new and foreign. In this second phase, increasing levels of multinationality positively influence a firm’s profitability or performance. However, a third phase follows in which too high levels of multinationality increase transaction and coordination costs to such extent that they outweigh, again, the benefits of

(10)

internationalization. This horizontal S-curve relationship of multinationality and performance can be moderated by a firm’s intangible assets, such as advertising and R&D intensity, where higher levels of intangible assets increase the performance gains associated with geographic diversification (Lu & Beamish, 2004).

Another approach in testing the M-P relationship considers the regionalization theory of Rugman and Verbeke (2004). This theory suggests that multinational firms have relative sales dominance in a specific regional market rather than a very wide and evenly distributed spread of sales, which has implications for mainstream international business research topics. Rugman and Oh (2010) examine whether this regionalization theory influences the way in which the main literature dealing with the M-P relationship empirically tests this. They propose a new measure of the multinationality conceptualization in which the ratio of regional sales to total sales (R/T) supplements the traditional ratio of foreign sales to total sales (F/T), and find that the R/T ratio may be superior. Therefore, including the R/T ratio as moderating variable in the link between multinationality and performance may lead to a better understanding of the M-P relationship.

Focusing on the regionalization aspect of internationalization, another study examines Chinese multinational firms expanding into the Great China and Asia areas compared to firms internationalizing into a broader geographic scope (Chen & Tan, 2012). This study finds that firms internationalizing within the Asia region gain higher performance gains than firms expanding outside this region. This suggests that regionalization theory indeed influences the relationship between multinationality and performance.

Instead of looking at external moderating factors affecting the M-P relationship, Verbeke and Barkema (2002) argue that the conceptualization of multinationality, namely the process of international expansion, is of great importance for performance outcomes of firms. Due to time compression diseconomies, in which increased pace in process leads to

(11)

diminishing returns (Dierickx & Cool, 1989), differences arise between firms following different expansion paths. In order to compare these different paths and the speed at which firms internationalize, Chetty, Johanson and Martín (2014) develop a conceptualization and measure of the speed of internationalization. The optimal firm-level speed should balance international opportunities and FSAs, and is therefore a key element in a firm’s international strategy, which intends to optimize performance outcomes.

Kirca et al. (2011) provide a meta-analysis of what studies have been conducted so far and thus provide a good overview of past research and directions for future research. This meta-analysis includes the findings of Lu and Beamish (2004) that advertising and R&D intensity can have moderating effects on the M-P relationship, and uses more past research to identify that FSAs strongly influence the M-P relationship under certain conditions, such as R&D and advertising intensity and their stronger effects in different industries (Kirca et al., 2011). R&D and advertising intensity as FSAs affecting the M-P relationship are the main focus of this meta-analysis, consistent with internalization theory, because other FSAs like human assets and management and production skills could not be included due to lack of inclusion in prior studies in the M-P relationship literature. This lack of inclusion of other FSAs indicates a need to capture FSAs that are more related to the resource-based view in international business and strategy literature.

2.3 Different empirical models and their critiques

Although the research about the linkage between multinationality and performance has a large domain, no consistent and interpretable findings have been generated and no general consensus about the M-P relationship has been reached (Cardinal, Miller & Palich, 2011). This largely stems from the different relationship patterns produced by many different studies. This section elaborates on different empirical studies finding the main patterns for

(12)

the link between multinationality and performance, and some critiques posed against these patterns are discussed.

2.3.1 The positive linear model

In line with the basic idea that more international diversification provides opportunities for scale and scope economies and could therefore increase firm performance, some studies find evidence for a positive linear M-P relationship (see Figure 1, where DOI is the degree of internationalization) (e.g. Grant, Jammine & Thomas, 1988; Han, Lee & Suk, 1998). According to the RBV, core competencies developed in the domestic country can be exploited internationally to enhance performance, for example by creating economies of scope advantages (Cardinal et al., 2011; Wiersema & Bowen, 2011). Also, by operating in foreign markets, firms can enhance their knowledge base and capabilities through experimental learning. Firms can therefore gain exploration benefits from international diversification (Buckley & Casson, 1976; Vermeulen & Barkema, 2001). From an industrial organization perspective it is argued that firms can increase their market power over customers, suppliers and other actors in the value chain by expanding operations across several international markets (Kogut, 1985). The benefits from increased market power, organizational learning, and the exploitation of scale and scope economies suggest that firms with higher levels of multinationality should have higher performance levels (Wiersema & Bowen, 2011).

Although the arguments given above supporting a positive linear M-P relationship follow logically from basic business principles, they do not entirely incorporate the fundamental complexity of the dynamics of internationalization (Cardinal et al., 2011). The main critiques against the positive linear model are based on the questionability of two assumptions: that international opportunities are unlimited, and that firms can handle all

(13)

complexities and challenges associated with international diversification on unlimited basis. In reality, international opportunities of growth, economies of scale and scope, and location advantages are limited. Also, firms have limited capacities to absorb the increasing challenges and complexities associated with the management of expanding international activities (Cardinal et al., 2011). Therefore, the positive linear model is theoretically limited.

Figure 1 - Positive linear M-P relation

Source: Gomes & Ramaswamy, 1999: 42

2.3.2 The inverted U-shaped model

In reaction to the limitations of the positive linear model, internationalization research starts focusing on curvilinear models, starting with an inverted U pattern (see Figure 2) (e.g. Gomes & Ramaswamy, 1999; Hitt et al., 1997). The arguments for increasing performance with international diversification as given for the positive linear model still hold for low to moderate levels of multinationality. However, higher levels of international diversification increase the managerial complexity and transaction and coordination costs (Wiersema & Bowen, 2011). Studies finding an inverted U-shaped relationship between multinationality and performance suggest therefore that moderate, and not too high, levels of international diversification form the optimal balance of costs and benefits and lead to higher performance levels (Cardinal et al., 2011).

(14)

The inverted U model shows progress in the move from the limited positive linear model, but it still has its shortcomings when compared to reality. The main critique is that it starts upward sloping immediately, and it keeps downward sloping after a certain level of multinationality. This means that this model does not incorporate the possibility of slow learning and making mistakes at initial phases of international expansion, which would be better explained by beginning with a flat curve. It also does not allow for growth prospects at higher levels of internationalization, possible due to learning by firms from cumulative experience (Cardinal et al., 2011).

Figure 2 - Inverted U-shaped M-P relation

2.3.3 The U-shaped model

Incorporating the initial trial and error phases in the internationalization process, some studies find an U-shaped pattern for the M-P relationship (Figure 3) (e.g. Capar & Kotabe, 2003). At initial entry in foreign markets, firms face challenges such as the liability of foreignness and newness, and they must bear the costs of over overcoming these liabilities (Hymer, 1976; Zaheer, 1995). Firms can do this by gaining experience abroad and acquiring additional knowledge about foreign markets, as explained in the Uppsala model of internationalization (Johanson & Vahlne, 2009). Further internationalization generates such knowledge and expertise according to the U-shaped model, and allows firms to take advantage of the benefits

(15)

associated with international diversification (ID), as described in the positive linear model. After this point, continued expansion should therefore increase financial performance (Cardinal et al., 2011).

As the costs of doing business abroad such as LoF and LoN are widely recognized, the initial downward slope in the U-shaped model of the M-P relationship seems logical. However, as the Uppsala model of internationalization also explains, firms are likely to choose to enter foreign markets that are similar to their domestic markets and are therefore easier to manage (Johanson & Vahlne, 1990; 2009). This lowers the costs in the initial phase of internationalization, suggesting that the initial downturn in performance might not occur.

Figure 3 – The U-shaped M-P relation

2.3.4 The horizontal S-shaped model

In reaction to the shortcomings of the previous models, some studies propose a horizontal S-shaped model, or general three-stage paradigm (Figure 4). This model aggregates important features of the inverted U and U-shaped models. For the same arguments as described in those models, the horizontal S-shaped model starts with a decrease in performance due to LoF and LoN, followed by an increase due to experience from international business. Too high levels of multinationality come with high transaction and coordination costs, having a decreasing effect on performance (e.g. Lu & Beamish, 2004).

(16)

The general three-stage paradigm seems to take away all the shortcomings of the other models and has been described ‘intuitively plausible’ (Verbeke & Brugman, 2009), but it still receives some criticism. First, the same critiques as for the other models apply for this model, although in less severity, as it uses the same arguments to build the model. Second, more empirically, as it combines the inverted U and U-shaped models, the S-shaped model should unfold across the full range of ID, something unlikely to be met in empirical research (Cardinal et al., 2011).

Figure 4 - The horizontal S-shaped M-P relation

Source: Lu & Beamish, 2004

The different findings of different M-P relation patterns across the many studies conducted on this topic suggest that as long as researchers study restricted samples, it is unlikely that understanding and consensus about the M-P linkage will be reached. As it is unlikely that unrestricted samples can be used to study this linkage, and in line with Verbeke et al. (2009) and Powell (2014), it is time to move away from the search for one generalizable M-P relationship. Instead, more research should focus on opening the black box of multinationality and evaluate the different substantive elements that lead to performance.

(17)

2.4 Not one generalizable M-P relationship

The concept of a generalized M-P relationship has also been challenged by Verbeke and Brugman (2009), who use an internalization theory perspective to argue that not so much the degree of multinationality determines firm-level performance, but rather the characteristics of the firm’s FSAs. Their study proposes triple testing the quality of M-P studies and can therefore be useful for current and future research on the M-P relationship, or the conceptualization of multinationality or performance itself.

In line with the theory that FSAs influence the M-P relationship (Verbeke & Brugman, 2009; Kirca et al., 2011), and also with the notion that there might not be one generalizable M-P relationship (e.g. Verbeke et al., 2009), Powell (2014) uses transaction costs theory and internalization theory to test the idea that there are different optimal levels of multinationality for individual firms. This study takes on a new approach of multinationality alignment and performance (MA-P), in which intangible assets of individual firms lead to optimal configurations of international expansion to decrease transaction costs and maximize value, and can be translated into financial performance. Because optimal levels of multinationality differ among firms, no generalizable link between multinationality and performance can be made within this approach. However, it is still providing a theoretical link to performance.

2.4.1 Missing theoretical link

While there is an abundance of theories about international diversification of firms and studies examining the M-P relationship, Wiersema and Bowen (2011) and Hennart (2011) conclude that the current conceptualizations of multinationality persist to be inadequate. In his commentary paper, Hult (2011) shows that the inconclusive findings in the M-P research stream are the result of the absence of a direct link between multinationality and performance. Multinationality can only indirectly be linked to performance through the

(18)

conversion of multinationality into something of value. This is comparable to the missing link in the RBV literature: no direct link exists between resources and performance, but this link can indirectly be made through a firm’s ability to turn resources into something of value.

The RBV focuses on the nature of FSAs, where resources have potential for sustained competitive advantage only when they are valuable, rare, inimitable and non-substitutable (Barney, 1991). This creates concerns (Hennart, 2007), as the RBV suggests that uniqueness of resources does not necessarily imply competitive advantage and/or above normal returns. However, Peteraf (1993) shows that high returns do not depend on uniqueness or rarity of resources per se, because equally efficient firms can theoretically earn rents, as long as there exists an efficiency differential between those firms and others. In this sense, not only a firm’s resources matter, but also its efficiency in exploiting FSAs compared to other firms. A recent study does show, however, that the relation between resources and performance is stronger for resources that are valuable, rare, inimitable and non-substitutable, providing an explanation for the resources-performance link in the RBV (Crook, Ketchen Jr., Combs & Todd, 2008).

From both the RBV and internalization theory it is argued that a firm’s proprietary assets, such as technological know-how, patents, brands, marketing skills, reputation, and production and management skills, are strategic resources that lead to competitive advantage in international markets (Lu & Beamish, 2004). Internalization theory provides arguments for multinationality being the most efficient governance structure to transfer a firm’s FSAs across country borders, while keeping them in the firm, and for these transfers to positively impact on firm profitability (Kirca et al., 2011). Adding arguments from the RBV, including Barney’s (1991) criteria, multinationality is a good channel through which a firm can exploit its FSAs efficiently to increase firm performance in international markets.

(19)

2.4.2 Inadequate conceptualization of multinationality

The contradictory results in the M-P literature lead not only to critiques about the generalizability of the M-P relationship, but also to the questionability of the theory behind multinationality increasing firm performance (Hennart, 2011; Hult, 2011; Wiersema & Bowen, 2011). Nonetheless, multinational firms exist, an increasing number of firms are expanding across country borders, and in general they earn higher profits than firms only active in domestic markets (Contractor, 2012). Therefore, the problem might not be the underlying theory, but rather the inadequate measurements in testing the M-P relationship.

Three essential elements should be considered prior to undertaking empirical research on the M-P relationship (Verbeke et al., 2009). These elements are the heterogeneity in (1) rationales for firm-level, international FDI programs, either in the context of contraction or expansion, (2) the foreign environments facing the MNE, and (3) functions and capability portfolios of foreign affiliates. Taking account of the variety of strategic motivation for FDI, the environmental complexity and the organizational complexity opens the black box of multinationality for researchers, allowing them to study the effect on performance when parameters of multinationality take on particular values instead of the general measure of foreign sales to total sales.

While acknowledging that the search for a generalizable M-P relationship in order to guide managers in their optimal international expansion path is unlikely to be resolved, Verbeke et al. (2009) contribute to this search by pointing out the different parameters of the multinationality conceptualization, in the hope for more effective M-P research. Highlighting these three substantive elements opens up the black box of multinationality, in an attempt to address the theoretical missing link between multinationality and performance as noted by Hult (2011) and Powell (2914). Instead of trying to find a direct link between multinationality

(20)

and performance, the distinct relation of each substantive element of multinationality with performance should be examined.

Overall it seems that international diversification does improve a firm’s profitability, even though many studies testing this linkage provide mixed results. There are two main explanations for the inconclusive findings in the M-P literature (Contractor, 2012). First of all, some firm-specific or country-specific factors change the beneficial effect of multinationality on performance. Such factors differ among firms, industries, and countries, and should be accounted for when further investigating the M-P relationship. Incorporating elements from the RBV provides insights in the differences between firms and their FSAs and are therefore a good solution for the noisy effect such factors have on the positive relation between performance and multinationality.

Secondly, most previous studies use cross-sectional data and methods, therefore testing the M-P linkage at one point in time. From the literature discussed so far, it follows that at certain stages of the internationalization path of firms, the costs (e.g. LoF and LoN) outweigh the benefits of multinationality. Therefore, using longitudinal (panel) data should provide better insights in the M-P relationship for firms along their different expansion paths (Contractor, 2012).

2.4.3 The AAA framework

One way to deal with the conceptualization of multinationality is to use the AAA framework constructed by Ghemawat (2008). The AAA framework describes three dimensions of multinationality, namely adaptation, aggregation, and arbitrage. Each of these dimensions are associated with different challenges and benefits; for example, where country differences can be a source of value with arbitrage activities, they can be a constraint on value with aggregation activities due to for example the associated LoF (Ghemawat, 2007).

(21)

Adaptation activities draw on local resources and knowledge to exploit locational advantages (Nachum & Zaheer, 2005). This relates to the vertical integration motive for multinationality, in which efficiencies can be realized through the integration of value-creating activities (Powell, 2014). Aggregation activities can improve performance by tapping into local markets to realize economies of scale and scope, while facing challenges of differences between countries. This relates to the horizontal integration motivation for multinationality as defined by Powell (2014). Arbitrage activities relate to the diversification motive for multinationality to increase firm stability and/or reduce risk. Firms can exploit differences between countries and leverage country-level advantages across their global operations by creating an integrated network of subsidiaries (Frost, Birkinshaw & Ensign, 2002).

2.5 Hypotheses

Although previous studies provide inconsistent findings, the general consensus is that multinationality increases firm performance, but firm-specific and country-specific factors may alternate this relationship. Therefore, the first hypothesis of this thesis is:

H1: Internationalization of firm activities increases firm profitability.

The three distinct dimensions of multinationality – adaptation, aggregation, and arbitrage – may have different performance implications, but firms may also change the relative emphasis they put on each dimension at different levels of multinationality. At higher levels of multinationality, when firms move from more similar to more distant markets (Johanson & Vahlne, 2009), firms experience less local relevance of their home country knowledge and resources and therefore receive greater value from their efforts to adapt to conditions in foreign markets. This leads to the following hypothesis:

(22)

H2: Higher R&D intensity levels increase the positive effect of multinationality on firm

performance.

On the other hand, when firms increase their multinational presence, the marginal value of their home country FSAs is likely to decline, which lowers average returns. Therefore, the average profitability of aggregation activities may decline as the firm becomes increasingly multinational (Hitt et al., 1997). This leads to the following hypothesis:

H3: Higher levels of multinationality decrease the marginal effect of aggregation activities on

firm performance.

A firm that operates in only a few countries may have limited access to foreign knowledge and recourses, whereas a firm operating in more and diverse foreign locations has access to a larger variety of foreign knowledge and resources. Therefore, the more multinational firms are, the greater opportunities to exploit country-level differences through international diversification (Frost et al., 2002). This leads to the following hypothesis:

H4: Higher levels of multinationality increase the positive effect of arbitrage activities on

firm performance.

Previous studies also highlighted the importance of the country of origin effect (COE) (Harzing & Sorge, 2003). The COE is an indicator for preferential differences for products or services from countries in which companies originate. There may exist differential conditions in the MNE’s home country that would impact the MNE’s performance if certain attributes of that home country were internationalized by the MNE (Elango & Sethi, 2007). Small home countries typically have large trade volumes, providing profitability opportunities outside the home country. Larger home countries have relatively moderate trade levels and therefore tend to have less influence on the M-P relationship. This leads to the following hypothesis:

H5: MNEs with small home countries tend to get more increased performance levels from

(23)

3. Data and methodology

Yang and Driffield (2012) suggest that regression analysis provides more reliable results than other approaches when studying the M-P relationship. Therefore, this thesis will use quantitative research to study the effect of international diversification on firm performance, using listed European multinational firms as sample. The dependent variable is firm financial performance, measured by the return on assets (RoA), which is the ratio of a firm’s income to its total assets, and the main independent variable is the degree of multinationality, indicating the extent of a firm’s foreign footprint. This chapter first describes the data that are used and then elaborates on the methodology with which the M-P relationship is analyzed.

3.1 Data

The firm-level data are obtained from Bureau van Dijk’s (BvD) Orbis database, a global financial database that contains comprehensive information on public and private companies worldwide. BvD’s Orbis database provides up to ten years of detailed information, making the database appropriate for the use of longitudinal data in quantitative analysis.

The sample is selected on the criteria that the companies have at least one foreign subsidiary, and are large (defined by BvD’s Orbis as having over 10 million euros in operating revenue) or very large (having over 100 million euros in operating revenue). Companies should also be originated in one of the 28 European Union member countries, and must have the same country of origin as its Global Ultimate Owner (GUO) (Elango & Sethi, 2007; Harzing & Sorge, 2003). A GUO can be defined as the highest parent company, owning a minimum percentage of 50.01% of a subject company in order it to be a subsidiary and having the knowledge about at least one of its shareholders, which cannot own more than 50.01% (BvD’s Orbis Database, 2015).

(24)

The use of financial performance as dependent variable in ordinary least squares regression analysis may lead to biased estimation results, since financial results of multinational firms typically vary from year to year. Therefore, data are used from the years 2007 to 2014, both to control for such estimation bias and to control for possible effects from the recent financial crisis. To rule out any inferior or superior firm performance a five-year period is generally accepted (de Jong & van Houten, 2014), so the eight-year period covered in this thesis is more than sufficient.

The sample selection according to the above mentioned criteria resulted in 1292 MNEs. From this large number a subsample of 500 MNEs is made. After accounting for missing values, occasional outliers or companies with a date of incorporation later than 2007, the sample selection process resulted in 497 MNEs, which for eight years resulted in 3976 observations. Table 5 in Appendix B shows the distribution of these 497 MNEs across the EU member countries and BvD’s major industry sectors. The United Kingdom is the country with the largest number of publicly listed MNEs (149), followed by France (70), Germany (65), and Sweden (43). The industry sector with the largest number of MNEs is Machinery, equipment, furniture & recycling (99), followed by Other services (85), and Chemicals, rubber, plastics & non-metallic products (59).

In order to account for the possible differences in performance levels for MNEs with small home countries as compared to MNEs with large home countries, the European countries where the 497 MNEs originate from must be defined as either small or large. There are three commonly used variables to measure country size, namely land area, population, and economic performance, which is measured as gross domestic product. Small countries may experience disadvantages like diseconomies of scale or lack of diversification due to limited natural resources, so they typically have larger trade volumes than larger countries to compensate for this (Harzo & Sorge, 2003; Sethi & Judge, 2007). Therefore, this thesis uses

(25)

the variable land area to divide the countries into being small or large. Countries with a land area larger than 100.000 square kilometers are defined to be large countries, whereas countries with a land area smaller than 100.000 square kilometers are defined to be small countries. This leads to the following division of the countries: Germany, Spain, Finland, France, the United Kingdom, Greece, Italy, Poland, and Sweden are considered large countries, whereas Austria, Belgium, Denmark, Hungary, Ireland, Luxembourg, the Netherlands, Portugal, and Slovenia are considered small countries.

3.1.1 Dependent and independent variables

The dependent variable is MNE performance and is measured by a firm’s return on assets (RoA). The RoA is calculated as earnings before interest, taxes, depreciation, and amortization (EBITDA), divided by the total assets of a firm. The data used for EBITDA and total assets are both in thousands of euros. While previous studies use the ratio of a firm’s net income to its total assets (Lu & Beamish 2004; Kirca et al., 2011), this thesis uses income before interest and taxes in order to better capture a firm’s ability to create value from its assets aside from the different tax rates in different countries.

Adaptation activities are measured with a firm’s R&D intensity, calculated as a firm’s R&D expenses divided by the number of employees. This ratio measures a firm’s knowledge activities and is therefore a good indicator of a firm’s adaptation activities with which it exploits locational advantages (Nachum & Zaheer, 2005).

To measure aggregation activities across foreign markets, the commonly used measure for multinationality can be used, which is the ratio of foreign sales to total sales (F/T). Since foreign sales data are largely unavailable for European publicly listed companies and therefore the commonly used measure for multinationality cannot be used, this thesis follows de Jong and van Houten (2014) to construct a multidimensional measure of multinationality,

(26)

as recommended by Hitt et al. (1997). Multinationality is measured by the formula 𝑀 =!_!(_{!"# !}! +_{!"# !}! ), where N is the number of foreign subsidiaries the firm has, max{N} is the maximum observed number of foreign subsidiaries of a firm, K is the number of foreign countries in which the firm has subsidiaries, and max{K} is the maximum observed number of foreign countries in which a firms has subsidiaries. Since the measure for adaptation activities controls for the development of new capabilities abroad, this measure of multinationality captures the extent to which a firm is able to leverage its home country FSAs in foreign markets in order to realize economies of scale and scope (Powell, 2014).

The ratio of intra-firm cross-border products transfers to total sales of the firm measures the extent of integration between firm activities located in different countries. As this captures the extent to which products flow across country borders within the firm, it is a good measure for a firm’s arbitrage activities. However, data on intra-firm cross-border trade is largely unavailable for the companies included in the sample, making it impossible to use such data for arbitrage activities. Therefore, the dimension of arbitrage activities of Ghemawat’s (2008) AAA-framework cannot be included in this thesis, unfortunately.

Table 1 - Descriptive statistics dependent and independent variables

Variable Mean Std. Dev. Min. Max. Obs.

Return on Assets 12.32 7.79 -36.77 86.68 3976

R&D intensity 5.73 18.64 -1.22 457.24 3973

Multinationality 0.10 0.10 0 0.83 3976

Table 1 shows the descriptive statistics for the dependent and main independent variables. The average return on assets for all MNEs included in the sample is 12.32, meaning that on average MNEs are able to turn their assets into more than 12 times of their value. However, there are large variations with a lowest return on assets of -36.77 and a highest of 86.68. The way multinationality is measured results in values ranged between 0

(27)

and 1. The average value of 0.10 with a standard deviation of 0.10 for multinationality in the sample indicates that European multinationals are generally not very internationally diversified. R&D intensity shows a minimum of -1.22 and a maximum of 457.24, while the average is 5.73. This suggests that there are only a few firms with high R&D intensity levels, while the lion’s share of MNEs keeps R&D expenses as ratio to total employees relatively low.

3.1.2 Control variables

Several control variables are included in the model. First of all, whether a firm has a small or large home country may lead to different performance outcomes. Therefore, the division of the countries between being small and large is included as a dummy variable, meaning that MNEs with a small home country have the number 1 assigned to the small home country variable, whereas MNEs with a large home country have the number 0 assigned.

From the competitive-forces approach (Porter, 2008) and the industrial organization perspective it is argued that industry profitability, among other external industry factors, determines firm performance. Also, large firms are typically diversified over several industries, and the average profitability varies considerably between industries. Therefore, industry dummies are included to control for industry effects.

The firm’s organizational capacities and other FSAs play a more dominant role in increasing firm performance in the RBV as compared to TCE and internalization theory. One commonly used variable used to control for firm-specific factors is firm size, since larger firms are typically more capable of exploiting economies of scale (Chao & Kumar, 2010). This in turn allows a larger firm to increase its return on assets. Firm size is therefore included as control variable and is measured by the natural log of the total number of employees. The natural log (Ln, the logarithm with base e) is used and not a ‘normal’ log,

(28)

because regression coefficients on the natural log scale can be directly interpreted as approximate proportional differences (Gelman & Hill, 2007).

Another size-related measure is used to control for firm size in terms of domestic footprint. The number of domestic subsidiaries, especially for service firms, may be related to firm performance (Hitt, Bierman, Uhlenbruck & Shimizu, 2006). Previous research does not indicate a clear positive or negative influence on performance, so a firm’s domestic footprint, measured by the natural log of its domestic subsidiaries, is included as control variable without expectations.

Similarly, the age of the firm may result in more knowledge stocks and improved performance (Qian, Li, Li & Qian, 2008). On the other hand, older firms may have lower performance when they still use outdated management practices or obsolete technologies (de Jong & van Houten, 2014). Since the impact of firm age on firm performance is ambiguous, firm age is included as control variable without expectations about its effect. Firm age is measured by detracting the year of incorporation of a firm from the years 2008 to 2014, as used in the analysis.

The last control variable included in this thesis is sales growth. Since higher sales growth may result in a more evenly spread of fixed costs over higher revenue and thus increased performance, this measure accounts for the efficient use of firm capacities and resources, i.e. its FSAs. Sales growth is measured by the difference between sales in year t and sales in year t – 1, divided by sales in year t – 1.

As noted above, Table 5 in Appendix B shows the distribution of the 497 MNEs across countries and industries. Table 2 shows the descriptive statistics of the control variables other than industry sector and country of origin. The average age of a firm is 57.73 years, where the youngest firm is only 1 year old in 2007, and the oldest firm is 349 years old in 2014. The large standard deviation of 51.39 suggests that there is large variation in the age of firms,

(29)

indicating that the sample includes a good selection of young, medium-aged and old firms. Firm size in terms of employees is on average almost 3 times as large as firm size in terms of domestic subsidiaries, but no derivations can be made from these numbers, as they are natural logarithms in order to be able to interpret their coefficients in regression analysis. Average sales growth is 10 per cent, with large variations between -0.83 and 84.23 per cent.

Table 2 - Descriptive statistics control variables

Variable Mean Std. Dev. Min. Max. Obs.

Firm Age 57.73 51.39 1 349 3976

Firm Size (employees) 9.40 1.48 2.64 13.19 3973

Firm Size (domestic subsidiaries)

3.31 1.26 0 8.01 3968

Sales Growth 0.10 1.39 -0.83 84.23 3976

3.2 Methodology

In order to analyze the effect of multinationality on firm performance, while incorporating a firm’s capabilities to exploit locational advantages and while controlling for firm-specific and industry-specific factors, a simplified model of moderated mediation is used. Moderated mediation occurs when the same model contains both one or more mediator variables and one or more moderator variables (Hayes, 2012). A mediator variable works between the independent variable and the dependent variable, such that the effect of the independent variable on the dependent variable passes through the mediator variable. This is also known as the indirect effect. A moderator variable is involved with an interaction with another variable and the dependent variable, such that the effect of that other variable on the dependent variable depends upon the value of the moderator variable. Figure 5 shows the model of moderated mediation from Hayes (2012) that is the most applicable to the analysis in this thesis.

(30)

Figure 5 - Model for moderated mediation analysis

Source: Hayes, 2012: 35

It is expected that R&D intensity, indicating a firm’s adaptation activities, works as a mediator between multinationality and performance, where the measure for multinationality indicates a firm’s aggregation activities. Relating to the model in Figure 5, this means that R&D intensity is denoted as M, multinationality is denoted as X, and firm performance is denoted as Y. The control variables work as moderator variables, denoted as W in Figure 5. The industry dummies and sales growth are expected to have a moderating effect on the direct M-P linkage, whereas firm size in terms of domestic subsidiaries is expected to have a moderating effect on the indirect link between multinationality and performance. The small-home-country dummy, firm size in terms of employees and firm age are expected to have moderating effects on both the direct and indirect M-P linkage.

Although Hayes (2012) created a computational tool for SPSS to analyze moderated mediation, or PROCESS what he calls it, this tool is only compatible with cross-sectional data and can therefore not be used in this thesis with longitudinal data. Therefore, a simplified model is used to analyze the mediating and moderating effects in the M-P relationship. Following previous studies (e.g. de Jong & van Houten, 2014), including an interaction term between the mediator and the independent variable makes it possible to (partly) analyze the indirect effect in linear regression. Squared terms of the independent variable and the mediator variable are included to analyze the more complex relationships. The control variables enter the regression to control for industry-, country-, and firm-specific

(31)

effects, which are simply assumed to analyze the moderating effects on the direct and indirect M-P linkage.

In order to analyze the precise effect of multinationality on performance, and the moderating effect of R&D intensity on this relationship, first a benchmark model is constructed which only includes the control variables. Then a second model follows in which the linear terms of multinationality and R&D intensity are added. In a third model the analysis is extended with the inclusion of the squared terms of the independent and the mediator variables. A fourth and final model follows adding also the interaction term between R&D intensity and multinationality, making this the fullest model including all variables and their squared and interaction terms. Analyzing the M-P relationship by adding the different parameters in stages also checks the model for robustness.

(32)

4. Results and discussion

This chapter elaborates on the results of the regression analysis, see Table 3. The estimation results include multiple hierarchical regression models to estimate the effect of multinationality, R&D intensity, their squared terms, their interaction term, and several control variables on firm financial performance. The first model serves as a benchmark model and includes only control variables. The main independent variables sequentially enter the regression analysis in the following three models.

The model fit measures at the bottom of Table 3 show that the fuller models, models 3 and 4, significantly fit the data better than the benchmark model and model 2. The R-squared increases from 12.9 per cent in the benchmark model to 14 per cent in models 3 and 4, whereas the Wald Chi-Square statistic increases from 120.85 to 142.26 in model 3 and 142.24 in model 4. This slightly lower Wald Chi-Square statistic in model 4 as compared to model 3 suggests that model 3 fits the data best.

4.1 Direct and indirect results for the M-P relationship

In all regression models the main independent variable of interest, multinationality, has a highly significant positive effect on firm performance. This provides clear evidence for hypothesis 1, stating that internationalization of firm activities increases firm profitability, which is reflected in a positive relationship between multinationality and performance. In Model 2, a linear regression model only including multinationality and R&D intensity besides the control variables, a 1% increase in the degree of multinationality increases a firm’s return on assets by 9.3%. For models 3 and 4 no such clear expectation can be made, since the squared term of multinationality and the interaction term with R&D intensity are included, making the regression non-linear. However, the regression coefficient increases from 9.30 to 24.19 and 24.12 in models 3 and 4, respectively, suggesting that adding those

(33)

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Model 1 Model 2 Model 3 Model 4

Constant 21.52*** (1.969) 23.17*** (2.048) 23.41*** (2.044) 23.42*** (2.046)

Independent variables

Multinationality 9.30*** (3.455) 24.19*** (6.753) 24.12*** (6.816)

R&D intensity -0.01 (0.009) -0.04*** (0.016) -0.05** (0.021)

Multinationality squared -30.96** (12.44) -30.91** (12.45)

R&D intensity squared 0.00** (0.000) 0.00* (0.000)

Multinationality * R&D intensity 0.01 (0.106) Control variables: RBV Firm Age -0.02*** (0.006) -0.02*** (0.006) -0.02*** (0.006) -0.02*** (0.006) Ln Employees -0.29* (0.172) -0.45** (0.181) -0.52*** (0.182) -0.52*** (0.182) Ln Domestic Subsidiaries -0.65*** (0.246) -0.94*** (0.268) -1.06*** (0.270) -1.06*** (0.270) Sales Growth 0.08 (0.055) 0.08 (0.055) 0.08 (0.055) 0.08 (0.055)

Control variables: Country

Small Home Country -0.28 (0.868) -0.31 (0.865) -0.37 (0.863) -0.37 (0.864)

Control variables: Industry

Industry sector 2 -2.32 (1.829) -2.63 (1.824) -2.65 (1.818) -2.64 (1.818) Industry sector 3 3.61 (2.783) 3.33 (2.773) 3.03 (2.766) 3.04 (2.766) Industry sector 4 -6.60*** (2.102) -6.72*** (2.093) -6.84*** (2.086) -6.84*** (2.086) Industry sector 5 1.73 (2.217) 1.82 (2.208) 2.12 (2.202) 2.12 (2.203) Industry sector 6 -1.84 (1.501) -2.12 (1.505) -2.00 (1.506) -2.00 (1.506) Industry sector 7 -5.33*** (1.683) -5.34*** (1.676) -5.36*** (1.670) -5.36*** (1.670) Industry sector 8 -3.50** (1.411) -3.64** (1.414) -3.52** (1.415) -3.51** (1.416) Industry sector 9 -5.90*** (1.751) -5.52*** (1.749) -5.37*** (1.744) -5.37*** (1.744) Industry sector 10 -7.95*** (1.649) -7.44*** (1.653) -7.07*** (1.654) -7.07*** (1.654) Industry sector 11 -5.25*** (1.599) -4.85*** (1.599) -4.58*** (1.598) -4.58*** (1.599) Industry sector 12 0.64 (2.777) 0.47 (2.765) 0.21 (2.757) 0.21 (2.758) Industry sector 13 -6.38*** (1.831) -6.26*** (1.824) -6.11*** (1.820) -6.12*** (1.820) Industry sector 14 1.03 (1.831) 1.06 (1.824) 1.47 (1.820) 1.47 (1.824) Industry sector 15 -6.37 (6.172) -5.92 (6.148) -5.32 (6.130) -5.33 (6.131) Industry sector 16 -2.25 (1.427) -2.18 (1.421) -1.95 (1.420) -1.95 (1.420) Industry sector 17 5.53 (6.191) 1.64 (6.332) 4.18 (6.402) 4.20 (6.410) Industry sector 18 -7.48** (3.732) -6.44* (3.737) -5.80 (3.733) -5.80 (3.734) N 3,965 3,965 3,965 3,965 Number of MNEs 496 496 496 496 R2 0.129 0.135 0.140 0.140 Wald chi2 _120.85 _129.79 _142.26 _142.24

(34)

squared and interaction terms amplifies the positive effect of multinationality on performance.

In the simple linear regression (Model 2), the coefficient of R&D intensity is negative, low, and far from significant. This means that in linear regression, the level of R&D intensity has no direct effect on firm performance. This is in line with expectations from theory, where R&D intensity as a proxy for a firm’s adaptation activities fulfills the role of mediator between multinationality and firm performance, and therefore by itself does not relate to performance. Adding squared terms of both multinationality and R&D intensity in Model 3 changes the coefficient of R&D intensity into a highly significant negative one, although still low in value. This result still holds when adding the interaction term in Model 4, which only makes it less significant, but still significant at the 5% level. This means that when MNEs increase their R&D intensity levels, this slightly decreases their financial performance. Although this is against expectations, no further derivations from this result can be made before looking closer at the squared and interaction terms.

In Model 3 the squared terms of multinationality and R&D intensity are added to analyze the more complex relationships. Looking at multinationality, the squared term has a negative coefficient that is significant at the 5% level (β= –30.96). This together with the highly significant positive coefficient for the linear term of multinationality (β= 24.19) provides evidence for an inverted U-shaped relationship between multinationality and firm performance. This means that international diversification first increases firm performance, but higher levels of multinationality negatively affect firm performance, providing evidence for hypothesis 3.

The parameter estimates for R&D intensity have the opposite signs as those for multinationality, but are so low in value that this might as well be the other way around. The squared term for R&D intensity has a positive coefficient with a value of 0.00, which is

(35)

significant at the 5% level, and the linear term has a highly significant negative coefficient, but also has a value very close to zero (β= –0.04). Therefore, the parameter estimates for R&D intensity neither confirm nor contradict the inverted U-shaped relationship between the dependent and independent variables, indicating that R&D intensity has very little influence on the M-P relationship.

To analyze the indirect effect of multinationality on firm performance through R&D intensity, an interaction term between R&D intensity and multinationality enters the regression in Model 4. The estimation coefficient of this interaction term is low (β= 0.01) and not significant, implying that the effect of multinationality on firm performance does not depend on the level of R&D intensity. This is not in line with expectations stating that higher R&D intensity levels increase the positive effect of multinationality on firm performance, leading to rejection of hypothesis 2.

Due to lack of data on intra-firm cross-border trade for publicly listed European MNEs, the impact of a firm’s arbitrage activities on the M-P relationship cannot be analyzed. Therefore, it is not possible to test whether hypothesis 4, stating that higher levels of multinationality increase the positive effect of arbitrage activities on firm performance, holds or should be rejected.

4.2 Results for the control variables

The estimation results for the control variable for country effects, namely whether an MNE has a small or a large home country, is close to zero and not significant. This means that the country of origin effect has no influence on the M-P relationship. Therefore, hypothesis 5 stating that MNEs with small home countries tend to get more increased performance levels from multinationality than MNEs with larger home countries is being rejected.

(36)

The results for the control variables from the RBV are all negative, except for sales growth, and significant mostly at the 1% level (β= –0.02 for firm age, β= –0.5 (approximately) for firm size in terms of employees, and β= –1.0 (approximately) for firm size in terms of domestic subsidiaries). Since firm size in terms of domestic subsidiaries and firm age were included as control variables without expectations, their results are not further discussed here. From theory (e.g. Chao & Kumar, 2010) it was expected that larger firms would be able to better exploit economies of scale and therefore have higher performance levels. However, the highly significant negative coefficient for firm size in terms of employees indicates that a one percent increase in firm size decreases a firm’s return on assets by approximately 0.5%. Recall that natural logs were used to measure firm size, making it possible to make such approximations (Gelman & Hill, 2007). A possible explanation for this negative effect of a larger firm on financial performance could be that a higher number of employees increases coordination costs to such extent that they outweigh the benefits of internationalization. Looking at the sample used in this thesis, the average firm size is 9.40, with a minimum of 2.64 and a maximum of 13.19 (see Table 2 at page 27). This suggests that a larger share of the 497 MNEs have a relatively high number of employees, leading to the negative coefficient for firm size in terms of employees.

Sales growth was expected to positively moderate the M-P relationship, since a higher growth in annual sales would enable a firm to spread its fixed costs more evenly, which in turn leads to higher performance. The coefficient for sales growth is positive but low (β= 0.08), and not significant. The positive sign is in line with expectations, but the lack of significance and low value make that sales growth does not significantly improve firm financial performance.

Table 3 shows that there exist industry differences for several industry sectors. The coefficients for industry sectors 4, 7, 8, 9, 10, 11, and 13 all have a negative sign significant

(37)

at at least the 5% level. This means that MNEs from the industry sectors Wood, cork & paper (4), Metals & metal products (7), Machinery, equipment, furniture & recycling (8), Gas, water & electricity (9), Construction (10), Wholesale & retail trade (11), and Transport (13) perform worse than those in the benchmark industry. The benchmark industry is the one left out of the regression, which in this analysis is industry sector 1 (Primary sector). Since the time window of observation (2007-2014) in this thesis is during and after the recent economic and financial crisis, the below average performance of MNEs from these industries aligns with the low economic activity during the crisis and slow pick-up after it in the sectors to which all these industries can be related.

5. Limitations and recommendations

This chapter discusses the limitations of this thesis and provides recommendations for future research. The empirical results derived in the previous chapter are not completely in line with expectations that followed from theory and previous studies. The differences in outcomes may be the result of the different sample used, or even that fact that the sample comes from a different population, i.e. European MNEs instead of American. They may also be the result of possible threats to internal and/or external validity. Section 5.1 discusses the limitations of this thesis, after recommendations for future research are given in section 5.2.

5.1 Limitations

The internal validity of a study refers to the degree to which statistical inferences can be made based on the measures and the empirical design used in the study. The external validity of a study concerns the extent to which the (internally valid) results are applicable to other settings and populations (Stock & Watson, 2012).

International diversification translating into financial performance for European listed multinational organizations