THE INFLUENCE OF DEVELOPING COUNTRY SUBSIDIARIES ON INTERNATIONALIZATION PERFORMANCE
van Lare, B.
S3013324
Rijksuniversiteit Groningen Faculty of Economics and Business
Nettelbosje 2, Groningen, 9747AD Tel: (00) 31 50 363 3741 b.van.lare@student.rug.nl
Abstract
In this paper, the influence of developing market subsidiaries on the relation between internationalization and performance is studied. Thereby this study builds on the great body of literature on the internationalization and performance relationship. All variables used are measured on the firm level. Return on assets is used to measure performance with the cause of internationalization expressed in foreign direct investment. Vast amounts of literature describe many factors which influence this relationship. In meta-analysis, this is a slightly positive relationship. However, developing market subsidiaries, measured as a percentage of total subsidiaries, is not one of the independent variables. Developing markets subsidiaries could have a positive effect when arguing that: it increases innovation (reverse innovation), eased rules and regulations allow for different approaches, and market opportunities as developed markets are often saturated and have smaller growth rates. Nevertheless, negative factors are also present like: greater levels of liability of foreignness, and the smaller market size in terms of value and volume. In the end, it is found that developing market subsidiaries have a slightly negative effect on performance.
Key terms: foreign subsidiaries, internationalization, performance Seminar supervisor: P.J. (Paulo) Marques Morgado
Second supervisor: E. (Esha) Mendiratta, PhD Seminar: Master’s Thesis IB&M
Table of Contents
Introduction ... 1
Literature review ... 3
Industries ... 3
Theories ... 4
I-P relationship ... 6
Disadvantages of conducting business abroad ... 6
The gains of going abroad ... 7
The relationship ... 7
Developing market subsidiaries ... 8
Reverse innovation ... 8
Institutional environment ... 9
Market opportunities ... 9
Hypothesis ... 10
Methodology ... 12
Data ... 12
Variables ... 13
Models ... 14
Results ... 16
Descriptive statistics and data analysis ... 16
Multiple regression analysis ... 18
Discussion ... 21
Conclusion ... 24
Appendix A ... 26
References ... 28
Introduction
“It has been said that arguing against globalization is like arguing against the laws of gravity” – Kofi Annan (United Nations, 2006)
The rise of globalization is reflected in the global trade numbers, from 5.2 trillion in 1995 to 16.5 in 2015 (Worldbank, 2016). Via this phenomenon, many firms are now involved in international activities. Dealing with an international environment influences firms, in for instance: their strategy, operations and business models (Majocchi & Zucchella, 2003). These influences are bound to have an impact on performance. Therefore, scholars are interested in the relationship between firm internationalization and performance (I-P) (Bausch & Krist, 2007). Current research entails many different results (Hennart, 2007), whereby this field is still an ongoing discussion (De Jong & van Houten, 2014). The effect of developing markets subsidiaries (DMSs) has not been given much attention. Therefore, this study aims to shed a light on the effect of DMSs on I-P. In doing so, the intention of this study is to aid the ongoing debate.
Three different industries are used to test the influence of DMS on the I-P relationship.
The chemical, metals, and pharmaceutical industries are used. All of these industries represent high capital industries, whereby much investment in assets is conducted. Two of them, the chemical and pharmaceutical industries, are more high-tech, what potentially results in higher performance due to innovation. Return on assets will be the proxy of performance, and its value is analysed via the resource based view. The resource based view theory, argues firms need resources superior to their competitors to internationalize successfully.
Existing research on I-P proposes that internationalizing firms have greater opportunities due to economies of scale and scope (Yang & Driffield, 2012). These opportunities increase firm performance (Dunning, 1988; Helpman, Melitz, & Yeaple, 2004).
Conversely firms also suffer from increased coordination which impacts firms in a negative way (Zaheer, 1995). With these positive and negative pressures, this field experiences many mixed results. Even within meta-analytical studies conflicting, but slightly positive results are found (Bausch & Krist, 2007; Kirca et al., 2011; Marano, Arregle, Hitt, Spadafora, & van Essen, 2016). Firm size, age, country of origin, product diversification, and R&D are just a few of the many factors which have been found to influence the relationship.
Although a great body of literature exists on the I-P relationship very little research has been conducted on the effect of DMSs. Studying this effect might lead to the discovery when internationalization is fruitful, which is according to Bausch and Krist an important topic for future study in the field (Bausch & Krist, 2007). Developing market subsidiaries are interesting due to their unique institutional environments and markets. To aid the ongoing debate this article sets out to answer the following question:
How do developing country subsidiaries influence the relationship between internationalization and performance?
No prior studies are found on the proposed effect, hence the expectations being uncertain. There are positive factors by which DMSs increase performance like: reverse innovation (innovation sourced in a developing country), the easier rules and regulations, and the growing market. However, the foreseeable negative influences are: the institutional environment (developing economies have generally less developed institutions), and the smaller market size. In the end, a positive association between DMSs and the I-P relationship is hypothesised.
This expectation has the potential to make DMS an important explanatory variable in the ongoing I-P discussion. Moreover, when DMSs are found to be a significant factor, scholars will be able to build on this study and explore the effects in depth. Not only is an answer of interest for scholars, firms are able to use this information to adapt their internationalization strategy. Due to the fact that no prior study was found on the proposed effect, significant results will lead to unique literature.
Literature review
Internationalization is seen in this study as moving across the firm’s own border. In this regard firms which internationalize cross borders via subsidiaries and exports. In this section the concepts of internationalization, the I-P relationship and DMS are provided. This section is divided into seven parts. The used industries in this study are discussed first. Second, main theories of internationalization are introduced. Trade theories, internationalization theories and the recourse based view are discussed. third, the costs and gains of internationalizing are discussed. The I-P relationship will be discussed in the fourth part focussing on the different findings. Reverse innovation, institutions and market opportunities of DMS are discussed in the fifth, sixth, and seventh part, respectively. Finally, a theoretical framework is offered to clarify the aforementioned effects.
Industries
The first industry is the metal industry. Most of the value in this industry comes from making basic iron and fellow-alloys (Eurostat, 2013a). Iron and steel represent 70% of the metal industry (Marketline, 2014). In making these products much energy is required, representing 40% of the costs (European Commision, 2017). Metals being elements, have the potential to be indefinitely recycled. However, as furnaces are required to be on all the time and reach economies of scale, producing new metals has a significant advantage (Marketline, 2014). The industry requires large initial investments, discouraging newcomers to the market (Marketline, 2014). Reasons to internationalize in the metals market are mostly driven by the increasing costs of transport and energy in the developed markets. DMS will likely have a lower impact on this industry as this industry is not very innovative and reliant on key recourses. Still, these resources are found in developing as well as developed markets (U.S.
Department of the Interior & U.S. Geological Survey, 2016), offering firms a choice. Second is the pharmaceutical industry. This industry is largely concentrated in the USA with 40.3% of the market share. The four greatest players own 25% of the value, and with growth rates of around 5% this is the quickest growing industry in this study (Marketline, 2013). Competition in this market is fierce and companies compete over drug approvals. The industry operates in a unique setting in which governmental and private spending finance the products. Having high quality materials, equipment, and personal is key in the pharmaceutical industry (Marketline, 2013). These key resources might be the reason for pharmaceutical companies to internationalize, or the motive may be the new market. The last industry used in this research is the chemical industry. 57% of the global revenue is generated in the Asia-Pacific region
(Marketline, 2013). Most of the value comes from basic chemicals (Eurostat, 2013b). The industry is known for its high sunk costs whereby it is hard to exit the industry. Entry is also hard, as the initial investment are high (Marketline, 2013). Although highly innovative, the European market is stagnant due to the high energy and labour costs (Marketline, 2013). This could drive the need to internationalize to developing or cheaper markets. For an overview of the exact business structures of the selected industries see appendix A: Figure 1, 2, and 3.
Theories
Fundamental to internationalization theories are trade theories. One of the first promotors of trade is Adam Smith, who argued countries should specialize in their commodities in which they have an absolute advantage (Schumacher, 2012). An absolute advantage refers to being able to produce a commodity for lower costs than the trading partner.
Years later David Ricardo theorized that a country does not need an absolute advantage, the trade needs to be mutually beneficial, which he referred to as a relative advantage (Costinot &
Donaldson, 2012). Eli Heckscher and Bertil Ohlin build on the concepts of Ricardo by putting it in a mathematical model. In essence this model predicts that products flow from places where production factors are cheaper and abundant to places where they are scarce (Leamer, 1995).
Location theory, first proposed by von Thiinen’s 1842 analysis, suggests the same when only looking at location to maximize utility (Krugman, 1993).
Firm perspective
Hence there are different benefits to diverse locations. A question which arises is ‘why do companies not establish themselves in every country to gain all locational advantages?’ The first theory which deals with this question is the Eclectic paradigm, or OLI framework, by John H. Dunning. Dunning’s theory states that: ownership, location and internationalization need to be fulfilled for successful internationalization (Dunning, 2001). The first factor, Ownership advantages, refer to products of production processes which are unique to the given firm. This resource is often intangible. Examples are: trademarks, patents, and reputation. Ownership advantages give benefits to a given firm, which should be greater than the disadvantages of internationalization. Second, location advantages are often the fundamental argument to internationalize. As mentioned above, different locations have different endowments which could be transferred into key resources. Examples of location advantages are favourable governmental conditions and cheap labour. The third factor is key in Dunning’s argumentation.
The great benefit of internationalization is being able to make transactions in-house which are
cheaper than the external market (Dunning, 1988). DMS are potentially a source of competitive advantage from this theories point of view. DMS have the potential to fulfil the entire framework in a value adding way. Especially the ownership is key as tacit knowledge is hard to mimic, and is therefore harder to substitute. The importance of this argument is supported later on in this paper by the resource based view theory.
Product life cycle theory by Vernon (1966) is the second discussed theory which has influence on locational decisions. The basic concept of the theory is that the comparative advantage of the production location shifts. Hence maturing products’ manufacturing places shift to other places. Following this model, it is logical to internationalize to developing markets, after the developed markets are more saturated. DMS are then of a positive influence on the I-P relationship. This argument’s logic is found at ‘Market opportunities, under Developing market subsidiaries’, later in the current section.
The last branch of theories influencing locations are stage theories which look at the experience of firms as a determining factor to internationalize. The most famous of these theories is the Uppsala theory by Johanson and Vahle. They propose that companies take incremental steps to internationalize. Via these incremental steps companies get the needed international experience and are able to produce themselves abroad (Johanson & Vahlne, 1977). In relation to DMS this theory has either positive or negative influence on the I-P relationship. If a company has the experience, it will be more positive, and vise-versa.
Recourse based view
Based upon the preceding theories in the field of internationalization, this study uses the resource based view, to explore the concepts of internationalization. Pioneering the RBV is Penrose, who found in 1959 that resources are important for a firms’ competitiveness.
Popularity, however, was gained later when Barney introduced his study ‘Firm Resources and Sustained Competitive Advantage’ in 1991. This theory explains how a competitive advantage of a company is built around its capability to use value creating strategies. A sustained competitive advantage is gained if other companies are unable to duplicate it. The RBV is built around the resources of the firm, the tangible (e.g. land) and the intangible (e.g. reputation).
The RBV assumes that the resources are heterogeneous (different from others) and immobile (not mobile). According to Barney the firm resources should be VRIN: valuable, rare, in- imitable, and non-substitutable, to gain a competitive advantage (Barney, 1991).
This theory has gained much support due to the limitations of stage theories (Barney, Wright, & Ketchen, 2001). One of such limitations is the inability to explain the concept of
‘Born-Global’. In the context of this study the RBV is applied to argue for the unique resources which are gained from internationalizing to developing markets. According to the RBV companies who enter new markets use their existing resources. Entry mode and foreign operation management are dependent on firm specific resources.
I-P relationship
In this subsection, the disadvantages and advantages of conducting business abroad are discussed. These factors are needed to understand the wide variance of results in the I-P relationship, which is discussed afterwards.
Disadvantages of conducting business abroad
The extra costs and complexities which arise when an MNE enters a foreign country is often referred to as liability of foreignness (LOF) (Zaheer, 1995). Most of the used sample’s companies originate from developing markets, therefore higher levels of LOF are expected.
According to Zaheer (1995) the costs of LOF is classified into four categories: 1 geographical distance, due to travel, transportation and coordination over distance; 2 firm specific, due to the lack of local knowledge; 3 costs from the host country environment, for example lack of legitimacy; 4 costs from the home country environment, like export/import restrictions. To overcome these costs she argues that a MNE should transfer its firm specific advantages or that it should mimic local partners (Zaheer, 1995). In a later study she argues that LOF falls with in-country experience (Zaheer & Mosakowski, 1997). Two years later Kostova & Zaheer (1999) apply institutional theory to the concept and argue MNEs get rewarded for local isomorphism with legitimacy in the foreign market.
Connecting LOF and institutional theory was a logical step, as institutions are “the rules of the game in a society or, more formally, are the humanly devised constraints that shape human interaction.” (North, 1990). As such the institutional environment, potentially explains a great deal of the LOF. MNEs are embedded in the institutional environment, institutional theory emphasises the importance to adapt to practices, policies, and structures within the institutional preferences in order to be legitimate (Meyer & Rowan, 1977). According to DiMaggio and Powel (1983), there are three ways in which the institutional environment influences a firm to gain legitimacy: coercive isomorphism (via other firms), mimetic isomorphism (via copying other firms), and normative isomorphism (via norms) (DiMaggio &
Powell, 1983). Firms investing abroad also face different standards of isomorphism, whereby they should make an extra effort (Eden & Miller, 2004). Ultimately institutions are a cause of LOF and cause extra costs when going abroad.
The gains of going abroad
Internationalization gives greater uncertainty, employing strategic assets in such a situation will lead to harder to imitate resources (Amit & Schoemaker, 1993).
Internationalizing also allows a firm to unlock resource pools and unique opportunities (Bausch
& Krist, 2007; Thomas & Eden, 2004). Via these additional resources multinational firms are able to appropriate more value from their assets (Goerzen & Beamish, 2003). According to Hennart (2011) international firms also have greater returns on their intangible assets, via the spread of fixed costs over a larger market (Hennart, 2011). Via these additional resources multinational firms are able to: reduce taxes (Thomas & Eden, 2004), have greater learning opportunities (Kim, Hwang, & Burgers, 1993), reduction of risk (De Jong & van Houten, 2014;
Kim et al., 1993).
The relationship
A great deal of literature in international business is dedicated to the relationship between internationalization and performance. The I-P field is still in an ongoing debate with varying results. Even comprehensive meta-analysis report different results, from slightly positive (Bausch & Krist, 2007; Marano et al., 2016), to U shaped (Yang & Driffield, 2012) relationships. When looking at individual studies a multitude of different outcomes is seen, positive linear (Chan Kim, Hwang, & Burgers, 1989), negative linear (Wan & Hoskisson, 2003), positive U shaped (Lu & Beamish, 2001), negative U shaped (Hitt, Hoskisson, & Kim, 1997), S shaped (Ruigrok, Amann, & Wagner, 2007), and no relationship (Hennart, 2007).
The great variety of result is likely due to a combination of theories and perspectives linked to the debate. To structure the theories and perspectives, they are classified in negative and positive international performance theories. First liability of foreignness (Zaheer, 1995) as a negative theory, second the three positive streams of literature proscribed by Contractor, Yang, & Gaur, (2016), which argue why companies invest abroad in spite of liability of foreignness, namely: internationalization theory (Hennart, 1982), the knowledge-based perspective (Nelson & Winter, 2002), and the resource based view (Barney, 1991).
Developing market subsidiaries
Developing economy subsidiaries are subsidiaries located in developing economies.
This type of subsidiaries face another kind of institutional environment, and thereby different rules of the game and growth potential in the economy (North, 1990, 1991). A great body of literature is dedicated to the performance of individual subsidiaries. Examples of this are: local embeddedness (Halaszovich & Lundan, 2016), knowledge transfer (Chang, Gong, & Peng, 2012), and country characteristics (Christmann, Day, & Yip, 1999). However, measuring performance on the subsidiary level might not be as useful as it seems, as it may be beneficial for MNEs to manipulate subsidiary performance for tax purposes (Christmann et al., 1999). In testing the effects of developing market subsidiaries on the firm level this literature review goes into potential factors via which developing market subsidiaries influence the I-P relationship.
MNEs investing in developing economies have relatively high returns (McKinsey Global Institute, 2003). However risk is also high, which shows in high pull out rates (Hansen
& Gwozdz, 2013). With the higher risk, and thus being harder to copy, one could argue that the resources would be more VRIN in the RBV, whereby these companies develop a competitive advantage. Potential forces out of which the competitive advantage arises in the case are reverse innovation, the institutional environment, and market opportunities. The latter two factors also carry potential negative influence, moreover under the perspective headers.
Reverse innovation
Reverse innovation is an innovation of a product in a developing country and later used in a developed country (Radojevic, 2015; Von Zedtwitz, Corsi, Søberg, & Frega, 2015). This phenomenon was first pointed out by Immelt et al. (2009), when describing how GE used its products developed in India and China at home (Immelt, Govindarajan, & Trimble, 2009).
Now, major players like Deloitte uses the term in their reports (Morrison, Pearce, Kounkel, Szuhaj, & Gantcheva, 2013). To develop reverse innovation existing products are modified.
Such modifications make it better suited for developing markets. Practically this comes down on 15% of the price and 50% of the performance (Radojevic, 2015). After improvement of the product in the developing market, the product is brought back to the developed markets. One of the original authors of the term, Vijay Govindarajan, stated in an interview that this is to serve the bottom of the pyramid, as the top 10% in developing countries are able to buy the original (Govindarajan & Euchner, 2012). In a blog Govindarajan illustrates how the reverse innovation works as follows:
FIGURE 1
Phases of reverse innovation (Govindarajan, 2009)
Institutional environment
With the less developed institutions in developing economies, one could argue that it allows firms to conduct business outside their domestic rules and regulations. This is especially beneficial for the selected manufacturing industries in this study. The used industries, specifically the chemical industry, have the potential to profit from less environmental regulations (Hoffman, 1999). When regulations are low dumping of the waste is likely cheaper than to process it.
However, under the pressure of global economic shifts, institutions in developing markets are improving (Hoskisson, Eden, Lau, & Wright, 2000). This makes sense as better developed institutions attract more FDI to a country (La Porta, Lopez-de-Silanes, Shleifer, &
Vishny, 1998). The cost of doing business in developing economies is still higher than the OECD high income economies (Solf, 2011). It might be therefore that developing countries which pursue foreign capital are democratizing and thereby changing their institutions for the positive (Li & Resnick, 2003).
Market opportunities
Income differences between developed and developing markets is still significant.
Therefore, the market sizes differ significantly. However, nowadays economic growth is shifting to developing economies (IMF, 2016, 2017). This is in line with the Solow model (Juchem Neto & Claeyssen, 2015). Due to the saturation of the established markets MNEs are increasingly investing in developing markets (Getachew & Beamish, 2017). A characteristic of developing economies is the use of economic liberalization, which leads to high growth potential and low income (Hoskisson et al., 2000). MNEs entering developing markets have
advantages to the local firms like technological advancement (Li, Lin, & Arya, 2008), organizational capital and access to credit (Chari, Chen, & Dominguez, 2012).
To summarise the effects, a theoretical framework is presented on below.
FIGURE 2 Theoretical Framework
Hypothesis
Much is gained via internationalizing. This is also reflected by the literature, which proposes a slightly positive I-P relationship. To support existing literature the following hypothesis is set:
Hypothesis 1: The higher a firms’ foreign direct investment, the higher its return on assets
With the establishment of pro and against arguments for developing subsidiaries, much potential for firm performance via developing market subsidiaries is seen. I propose that the positive aspects will way off the negative ones and hence offer the following hypothesis:
Hypothesis 2: Foreign direct investment delivers higher return on assets when the percentage of developing market subsidiaries relative to developed market subsidiaries is higher
In accordance with the hypothesis and the literature, a conceptual model is made and shown below. For the completeness of the model the way the factors are measured are also included under the respective factor. The measures will be elaborated on in the next section.
FIGURE 3 Conceptual model
Methodology
Positivism is the research philosophy this research sets out with. An objective view is taken and only observable data will be included. The statistical outcomes of the research are independent of the researcher and are structured to the best of the researcher’s ability. To assure independence of the researcher a deductive approach with the use of a statistical database is used. The manner of analysis is according to a mono method. Due to the use of the data base at a snap shot moment the study is of a cross-sectional nature.
Multiple statistical analysis have been used when studying the I-P relationship (Ruigrok et al., 2007). In this study, a multiple regression is conducted. However, before doing so the data needs to be tested for some assumptions. A set of parametric tests are used to check for the normality of the data. First of all, before conducting the tests the outliers will be removed, as extreme scores can bias the results (Field, 2013). To do so all entries which fall out of 2.5 times the standard deviation will be deleted. It is very common to delete 2.5 standard deviations (Ratcliff, 1993). Second, there will be checked for heteroskedasticity and skewness via the descriptive statistics. This is done to check the normal distribution of the sample. In addition, there will be looked at the histogram. Third, there is looked at the autocorrelation, which is checked via the Durbin-Watson test. Finally, to check for multicollinearity, a correlation matrix is made.
Data
Sources
For this research only secondary data is used. This decision was made due to time constraints and the completeness of existing databases. All of the data used is derived from Orbis a database by bureau van Dijk. Access to the database is gained via the Rijksuniversiteit Groningen.
Selection and design
The sample has many selection criteria: first of all, it consists out of the largest companies in the pharmaceutical, chemical, and metals industries of which Orbis has data. The choice of these industries was made due to their high-tech capabilities, high asset control, and being relatively similar when looking at the classification of industries by Eurostat (Eurostat, 2008). Second, the selected companies need to have known data on: developing market subsidiaries, foreign subsidiaries, a known value of ROA, active in one of the aforementioned
industries, long term debt, assets, and turnover. All of the values were measured in 2015 as this is the most recent year of which Orbis has data on for all companies. Thereafter, companies without any developing subsidiaries or foreign subsidiaries were omitted from the sample.
Finally, due to the immense size the sample size was cut off at 1097 companies. Later on, this value was lowered to 1007 due to the removal of outliers. The cut off was initially around the 3000 companies as the number of subsidiaries dropped drastically. 1097 was reached after further removal of zero values of the used variables. This was partially due to Orbis’ incomplete dataset, and due to companies not having developing market subsidiaries.
Variables
A closer look is taken at the used variables. First the dependent variable is discussed followed by the independent variable and ending at the control variables. Afterwards, the complete model will be presented.
Dependent
The dependent variable for performance is measured as ROA. As ROA reflects to what extend a firm is deploying its assets in a good way it is a sign of how a firm performs. Especially when looking at DMSs looking at ROA becomes interesting as the interest of this study is the effect on the total firm performance. ROA is the most popular measure of performance (Bausch
& Krist, 2007), and as ROA easy to use and widely applied in the I-P field, it is easy comparable with other studies (Ruigrok & Wagner, 2003). ROA is calculated by dividing net income by total sales, and is directly supplied by the Orbis database.
Independent
To measure the independent variable internationalization FDI is used. FDI, is measured via a ratio of foreign subsidiaries over total subsidiaries. The same measure has been used in previous studies (Lu & Beamish, 2006; Vermeulen & Barkema, 2002). Measuring internationalization in this way is likely a better measure than a percentage of exports, as one cannot find out where that percentage is actually based upon.
Moderator
The moderator in this study is developing market subsidiaries is measured as a percentage of total subsidiaries. Orbis supplies the countries in which the given MNE has subsidiaries. The calculation is done by hand and is just a simple percentage of the total. The classification of the United Nations will be followed to determine which countries should be
marked as developing economies (United Nations, 2014). In this classification economies in transition are also mentioned. As these economies are not part of the study, they were omitted from the sample. However, another regression was done with the computation of developing and transition market subsidiaries. There will be a small section attributed to this in the next chapter.
Control variables
In this study three control variables are used: company size, leverage and the industry.
These variables were also used in similar studies to control for external effects (Gomes &
Ramaswamy, 1999; Majocchi & Zucchella, 2003). For these variables, the following proxies are selected: turnover, leverage calculation, and dummy variables. These dummy variables consist out of 1 or 0 to show whether it is in its perspective industry or not. This results in three dummy variables of which two will be used, as statistics proscribe a control group with k-1 degrees of freedom for the total model. This control group will be the chemical industry as most of the entries were from the chemical industry.
The choice of these three industries is, as mentioned before, based upon the fact that they are relatively similar industries when looking at their Nace Rev. classifications. These industries are set as control variables as the dependent variable ROA is connected to industries.
The bases of the other control variables are now explained.
First company size, there are two sides to this story. Fist, a larger company has a broader resource base what is useful when internationalizing (Wolff & Pett, 2000). Second, smaller firms have a more specific focus and less bureaucracy due to their structure (Wolff & Pett, 2000). Moreover, differences in significance were found between large and other companies in the meta-analysis by Marano et al. (Marano et al., 2016).
Second leverage, leverage measured as long-term debt over assets, increases a company’s risk level but also raises its capital. As capital directly influences the dependent variable ROA, leverage is included. Furthermore, risk was found to be a significant factor in the study by Majocchi and Zucchella (2003).
Models
The following regression models will be conducted:
𝑴𝒐𝒅𝒆𝒍 𝟏 ROA= α+ β1(LEV) + β2(LnSize) + β3(Metal) + B4(𝑃ℎ𝑎𝑟𝑚𝑎)+ 𝜀
𝑴𝒐𝒅𝒆𝒍 𝟐 ROA= α+ β1(LEV) + β2(LnSize) + β3(Metal) + B4(𝑃ℎ𝑎𝑟𝑚𝑎)+ β5(FDI) + 𝜀 𝑴𝒐𝒅𝒆𝒍 𝟑 ROA= α+ β1(LEV) + β2(LnSize) + β3(Metal) + B4(𝑃ℎ𝑎𝑟𝑚𝑎)+ β5(FDI)+
β6(DEV) + 𝜀
𝑴𝒐𝒅𝒆𝒍 𝟒 ROA= α+ β1(LEV) + β2(LnSize) + β3(Metal) + B4(𝑃ℎ𝑎𝑟𝑚𝑎)+ β5(FDI)+
β6(DEV) + β7(FDI*DEV)+ 𝜀
In these model α is the intercept term. The error term: ε is assumed to be normally distributed. To make sense of the abbreviations used the following table is presented:
TABLE 1 Used abbreviations
Variable Definition Abbreviation
Level of FDI activities
Number of foreign subsidiaries over total
subsidiaries FDI
Profitability Return on Assets ROA
Developing market subsidiaries
Number of developing market
subsidiaries over total subsidiaries DEV
Leverage Long term debt over assets LEV
Firm size (Ln) Turnover LnSize
Chemical industry (dummy) 1 for Chemical industry/ 0 for other Chem Metals industry (dummy) 1 for Metals industry/ 0 for other Metal Pharmaceutical industry (dummy) 1 for Pharmaceutical industry/ 0 for other Pharma
The logic behind the use of these models is explained now. First model one, model one only consists out of the control variables. Second, model two is comprised out of the control variables in combination with the independent variable internationalization. Third, model three consists out of the control variables, and the independent variables internationalization and developing market subsidiaries. Finally, model four, the complete model which is build out of all the aforementioned factors in combination with the interaction term. The results of these models are presented in the subsequent section.
Results
In this section, the results of the models will be presented. First basic information is given about the validity of the model, along with statistical testing. Second, the multiple regression table is presented.
Descriptive statistics and data analysis
The data set contains 1007 MNEs in the chemical, metal and pharmaceutical industry.
The data set was reduced to 1007 from 1096 after the removal of outliers. From these 1007 MNEs 322 are in the metal industry, 215 in the pharmaceutical industry, and 470 in the chemical industry, which correspond to 32%, 21.4%, and 46.7% respectively. The descriptive statistics are presented below:
TABLE 2 Descriptive statistics
Now this section continues with the analysis of the data. As shown in the appendix (Appendix A: Table 1) the Kolmogorov-Smirnova and Shapiro-Wilk tests both indicate that the sample is not normally distributed as the values are significant. However, according to Field (2013) these tests are likely significant in large samples even for small and unimportant effects.
Hence, to check for normality the histogram of the dependent variable ROA is checked. In the figure below, a relative normal distribution is seen with kurtosis. However, the data is still seen as reliable as in sufficiently large samples normality is not a prerequisite (Lumley, Diehr, Emerson, & Chen, 2002).
Variable
N Minimum Maximum Mean Std. Deviation
FDI 1007 1.45 100.00 50.6515 29.38560
ROA 1007 -19.41 24.98 3.2790 6.32646
DEV 1007 1.45 100.00 51.5733 33.42588
LEV 1007 .00 56.89 15.0245 12.67291
LNSize 1007 12.32 17.23 14.0097 1.18731
FIGURE 4
Histogram Return on Assets
TABLE 4
Skewness and Kurtosis
Variable Skewness Kurtosis
Statistic Std. Error Statistic Std. Error
FDI .095 .077 -1.202 .154
ROA -.477 .077 2.134 .154
DEV .198 .077 -1.466 .154
LEV .888 .077 .213 .154
LNSize .605 .077 -.504 .154
Autocorrelation Durbin-Watson statistic of 1.879, which is close to 2 hence no problems with autocorrelation are assumed (see Appendix A: Table 2). As the data has been accepted to this point the correlations are checked via the correlation table. In the table below the correlations are presented. Please note that the dummy variables are not incorporated, as they are build out of the categorical variable ‘industry’ consisting out of only three factors. Due to this low number of categories, these variables are correlated, however as they are used to control for industry alone, this should not pose any problems.
TABLE 3 Correlations
As seen in the table no values are above or below the standard statistical threshold values of -.5, -.7, .5, and .7. Hence it is assumed that there are no problems with multicollinearity. In addition, variance inflation factor (VIF) is looked at. Found below, all of the VIF values are within the range of 1-10, and the tolerance does not drop below .20. Hence, no multicollinearity problems are assumed. As we can now assume that the data is good for the multiple regression, the multiple regression is conducted in the next part of this section.
TABLE 4
Variance Inflation Factor Variable Tolerance VIF
FDI .886 1.128
DEV .882 1.134
Metals .857 1.167
Pharma .845 1.183
LEV .882 1.134
LNSize .891 1.122
Moderator .838 1.193
Multiple regression analysis
In this subsection, the four proposed models are tested and evaluated, by their unstandardized Betas. The models are accompanied by R² statistics and the F statistics. R² indicates the amount of variability in the model. The adjusted R² statistic shows how good the
FDI DEV ROA LEV LNSize
FDI Pearson
Correlation 1 -.153** .076* .148** .027
Sig. (2-tailed) .000 .016 .000 .386
DEV Pearson
Correlation -.153** 1 -.083** -.107** -.136**
Sig. (2-tailed) .000 .008 .001 .000
ROA Pearson
Correlation .076* -.083** 1 -.182** -.008
Sig. (2-tailed) .016 .008 .000 .811
LEV Pearson
Correlation .148** -.107** -.182** 1 .296**
Sig. (2-tailed) .000 .001 .000 .000
LNSize Pearson
Correlation .027 -.136** -.008 .296** 1
Sig. (2-tailed) .386 .000 .811 .000
**significant at 1%, *significant at 5%
model is generalizable. The significance of the R² statistics is tested via the F statistic, which tests if the parameters are not too close to zero (Field, 2013).
TABLE 5
Multiple regression models
Variable Model 1 Model 2 Model 3 Model 4
(Constant) 0.883 0.271 1.441 1.445
Metalss -4.501*** -4.441*** -4.414*** -4.411***
(-.332) (-.328) (-.326) (-.325)
Pharma 2.079*** 2.011*** 1.954*** 1.949***
(.135) (.130) (0.127) (0.126)
LEV -0.086*** -0.090*** -0.092*** -0.092***
(-.172) (-.181) (-.185) (-.185)
LNSize 0.334** 0.339** 0.303** 0.302*
(.063) (.064) (.057) (.057)
FDI 0.012* 0.010* 0.011
(.056) (.049) (.049)
DEV -0.011** -0.011*
(-.057) (-.058)
Moderator -.014
(-0.002)
R² .195 .198 .201 .201
Adj R² .192 .194 .196 .195
F statistic 60.571*** 49.347*** 41.886*** 35.868***
N 1007 1007 1007 1007
***significant at 1%, **significant at 5%, *significant at 10%.
All of the proposed models have an adjusted R² value of between .192 and .196, this suggests that the models predict the value of ROA between 19.2% and 19.6%. These values are not particularly high, which can likely be explained by many other factors which can influence the dependent variable ROA. To control for the unexplained variance the two most used control variables in the I-P relationship are used: leverage and size. These two factors are significant in every model.
FDI as proposed in the literature and the first hypothesis, FDI has a slightly positive effect on performance. In the second model FDI just fell out of the 5% significance region with a value of .052. Only in the fourth model was FDI not significant anymore. It just fell out of the accepted significance region with a value of 102. This is likely explained by the moderator, which had such an insignificant value of .937 that it distorted the model. Therefore, hypothesis
one is accepted. Due to the insignificance in combination with the minus sign is sufficient to say that the second hypothesis cannot be supported.
As mentioned before, the transforming economy subsidiaries were not included. With an average percentage impact of only 1,3% not much deviations from the standard models should be found. To study the impact the transforming market subsidiaries are added to the developing market subsidiaries, the results of this regression is found in the appendix (Appendix E). As expected this did not change much, the only change was the significance of FDI, which is logical as FDI would become slightly more accurate.
In the next section these findings are discussed.
Discussion
In this section, the results are compared with the existing body of literature whereby building towards theoretical and practical implications. In doing so, it is important to consider the main question this research tries to answer. The research question is: ‘how do developing country subsidiaries influence the relationship between internationalization and performance?’.
To answer this question there is first looked at the I-P relationship in combination with the used control variables. Secondly, the effect of DMS is discussed accompanied with possible explanations. Finally, this section ends with discussing the theoretical and practical implications of the findings.
Uniform with other I-P relationship studies, this study finds the I-P relationship to be significant and slightly positive. Hence the advantages of going abroad like: reduction of risk and more value appropriation, win over the additional costs due to LOF. Therefore, the first hypothesis is accepted. As this is in line with literature, this section continues with the used control variables.
Fist, the industries. The pharmaceutical, chemical and metal industry are used to control for the industry type. The metal industry is found to have the most negative relationship on performance, relative to the other industries. This could be explained by the fact that metals are a relatively generic resource to produce, without much innovation happening. Therefore, metals are quite an easy substitutable product, whereby it is hard to gain a competitive advantage. The furnaces to make the metal are, however, very expensive and costly to run.
Hence companies in the metal industry could likely better focus on scaling up the process at one location to gain economies of scale, resulting in a lower production cost, which could lead to a competitive costs advantage. The other two industries are more technology reliant and face more heterogeneous products. Thus, it would get easier in these industries to gain a competitive advantage and gain higher performance, relative to the metal industry.
Second, leverage. Increasing the debt to asset ratio is found to be a highly significant negative factor on performance. This is in line with the findings of the meta-analysis by Marano et al. (2016), who also found debt to be a significant negative factor on performance. It could be the case that companies with poorer performance lend more in order to keep up. What the exact reason for the finding exactly is up for speculation.
The last used control variable is company size. Company size is found to be a significant positive influence. This finding is, again, in line with the study by Marano et al.
(2016). A possible explanation for the positive relationship to performance might be economies of scale and scope.
With significant control variables and the general relationship between internationalization a performance in line with existing literature (H1), a closer look is taken at the added value of this research. The second hypothesis is: ‘foreign direct investment delivers higher return on assets when the percentage of developing market subsidiaries relative to developed market subsidiaries is higher’. First the moderating effect of DMS are discussed, followed by the lose effect of DMS on performance, as this was found to be significant.
The initial expectation for DMS was to moderate the I-P relationship in a positive way.
However, the contrary is found. With a negative sign DMS seem to be a negative influence on the I-P relationship. Nevertheless, no significance was found, hence making more inferences about the moderating effect is not useful. Continuing with the effect of the relative amount of DMS on performance, a significant slightly negative relationship is found. It is of even more significance than the internationalization factor. In addition, the beta value offers more explanation relative to the internationalization factor. In the end, it is concluded that the second hypothesis cannot be supported. Not only is the hypothesis rejected by the results, the contrary (although insignificant) is found to be the case. Possible explanations for the effects of DMS on performance are offered below.
The first explanation originates from innovation and knowledge. When everyone offers basically the same and it is easy to switch, transactions are made on the market (Dyer & Singh, 1998). As reverse innovation only first appeared in 2009, many companies could not have thought about the benefits of such an innovation. If MNEs ‘just’ decide to be in a developing market for profit only, it could be the case that they see their knowledge as superior to the local knowledge. Via this effect less responsibility could be given to the DMS whereby reverse innovation is harder. This ‘lack’ of internal knowledge development will make it harder to make a resource which complies with Barney’s (1991) VRIN framework, hence not gaining a competitive advantage.
Second, liability of foreignness. As mentioned in the literature review, the extra cost incurred via the complexities of another country (Zaheer, 1995). The used sample is especially prone to LOF as 58% of the companies are from developing countries, keeping in mind that it is harder to deal with a less developed institutional environment than vice versa. These extra costs and complexities might fulfil the rare and in-imitable dimensions of the VRIN
framework. However, the extra costs are likely to destroy the value aspect, foreshadowing the lack of a competitive advantage.
Finally, the market size. There are still major differences between the developing economies and developed economies. Via these differences more profits are gained in developed markets relative to developing markets. Economic value and better institutions go together (Dollar & Kraay, 2003). Hence, LOF is likely to be greater in less developed economies. Via the smaller economic value and the likely value reduction do to LOF, the value prospect of the VRIN framework cannot be fulfilled, not resulting in a competitive advantage.
In the current days economic growth is shifting to developing nations (IMF, 2016, 2017; United Nations, 2013), so in the future better institutional development is to be expected. This could in turn the tables and increase the potential value from DMS.
The theoretical implication of this study is that in future I-P studies DMS should be included as a variable. This is as DMS are found to be a more significant factor than internationalization to predict performance. The country to which a company internationalizes seems to be of great importance. If scholars keep DMS in mind in future research in the I-P field, more significant results will potentially be found. With this contribution, a more complete model can be made to predict the I-P relationship, assisting the current ongoing debate in the field.
In a more practical sense, the advice to companies is to not expect great results from investing in developing markets. On the longer term however, higher results are expected. As the economies are growing, more value is added and less value will be lost via institutional differences. When going to a developing market, do not expect great reverse innovations to happen suddenly. Invest in the people, and share knowledge so that an environment is created in which reverse innovations has to potential to cultivate.
Conclusion
This study set out to discover what the effects of DMS are on the I-P relationship. From the literature, a slight positive relationship between I-P was assumed. This slight positive relationship is supported by many trade theories. The first hypothesis was set in line with these arguments. The results supported the first hypothesis, and therefore it is concluded that internationalization positively affects performance.
The second hypothesis was focussed on the moderating effect of DMS on the I-P relationship. As no other study on DMS in the I-P relationship was found, this study had to made its own arguments for and against DMS with respect to performance. The positive arguments were based on: reverse innovation, the institutional environment, and market opportunities via less saturation. LOF was argued to be of a negative influence.
Comparing these factors, the relationship between more DMS relative to developed market subsidiaries, was proposed to be of a positive influence on performance. The results however, predicted the opposite. It was found that DMS moderate the I-P relationship in a negative way. Although this moderation is not significant, the sign pointed in the opposite direction. The significant contribution of this study is the direct effect of DMS on performance.
It is found that a higher percentage of DMS directly influences performance in a negative way.
Possible reasons for this being: reverse innovation is not sufficiently present, the market size is not large enough, and the liability of foreignness is too great. In the VRIN framework, the value would be lacking, therefore not resulting in a competitive advantage, hence DMS having a negative impact on performance.
The theoretical implication of this finding is that DMS play a key role in the I-P relationship. Therefore, scholars who research the I-P relationship will have a more complete model if they include DMS. Where a company internationalizes to via FDI seems to be a very important factor. Thus, include a categorisation of countries like DMS in future studies. In a more practical perspective this study suggested companies to not expect profits from DMS on the short term. Rather, see them as a future investment as the economies are growing, the institutional environment is developing, and reverse innovation takes time.
Limitations and future research
This study has however some limitations. The first one being that in the sample only three industries are included, so it might not be generalizable. Second, the study is a snapshot
study, as only 2015 was used in the data. The year 2015 could have been completely different from other years. Third, there are many more factors influencing performance, with so few variables explanatory value is easier to be gained. Fourth, the database Orbis might not include all potential businesses in the industries, what could distort the results. Finally, as this study is based on a statistical analysis, the reasoning behind the effects comes from the literature. In real life, other factors might offer a better explanation than the ones offered.
These issues could be addressed in future studies. Future studies could dive deeper into the effects of DMS by applying a large timeframe to see whether having a subsidiary for a longer period in a given country effects performance. Such a study could be done over case studies, in this study literature was linked to the phenomenon from the literature. However, there might be other factors not found in the literature which can explain the relationship.
Another interesting field for future studies is how other industries are affected by DMS.
Appendix A
Table 1: Kolmogorov-Smirnov & Shapiro-Wilk test
Tests of Normality
Kolmogorov-Smirnov Shapiro-Wilk
Statistic df Sig. Statistic df Sig.
FDI .084 1007 .000 .950 1007 .000
ROA .117 1007 .000 .949 1007 .000
DEV .116 1007 .000 .896 1007 .000
LEV .118 1007 .000 .919 1007 .000
LNSize .084 1007 .000 .946 1007 .000
Table 2: Durbin-Watson test
Model Summarye
Model R R Square
Adjusted R Square
Std. Error of the
Estimate Durbin-Watson
1 .441a .195 .192 5.68854
2 .445b .198 .194 5.68067
3 .448c .201 .196 5.67252
4 .448d .201 .195 5.67534 2.028
a. Predictors: (Constant), LNSize, Pharma, LEV, Metalss b. Predictors: (Constant), LNSize, Pharma, LEV, Metalss, FDI c. Predictors: (Constant), LNSize, Pharma, LEV, Metalss, FDI, DEV
d. Predictors: (Constant), LNSize, Pharma, LEV, Metalss, FDI, DEV, Moderator e. Dependent Variable: ROA
Table 3: Transforming subsidiaries and developing subsidiaries combined regression models
Variable Model 1 Model 2 Model 3 Model 4
(Constant) .883 .271 1.193 1.287
Metalss -4.501*** -4.441*** -4.418*** -4.398***
Pharma 2.079*** 2.011*** 1.981*** 1.950***
LEV -0.086*** -0.090*** -0.092*** -0.092***
LNSize 0.334** 0.339** 0.310* 0.303*
FDI 0.012* 0.011* 0.012*
TRANSDEV -.008 -.009
MODTRANSDEV -.085
R² 0.195 0.198 0.2 0.2
Adj R² 0.192 0.194 0.195 0.194
F statistic 60.571*** 49.347*** 41.551*** 35.619***
N 1007 1007 1007 1007
***significant at 1%, **significant at 5%, *significant at 10%.
