The internationalization of EU
SMEs from performance perspective
Master Thesis International Economics and BusinessUniversity of Groningen Faculty of Economics and Business
Author: Diana Rittgasser
E-mail: d.rittgasser@student.rug.nl S-number: 2852624
Supervisor: Dr. Ortega Argilés, R.
University of Groningen
Co-Assessor: Dr. M. S. Sorin Krammer
2
Abstract
Previous empirical evidence has provided mixed results on the relationship between internationalization and firm performance. This research tests the horizontal S-shaped relationship on SMEs from the European Union. The theoretical framework of this relationship involves both benefits and motives associated with internationalization for firms in different stages of geographical diversification. Papers investigating the S-shaped relationship were focusing on MNEs or distinguished between the manufacturing or service sector but knowing that SME internationalization is not identical to MNE internationalization, this research tests this effect on SMEs. Additionally, it investigates the impact of FDI activity and operations beyond the European Union on this relationship. Using firms from the EU makes it possible to investigate this effect on a region where integration is advanced. Using a panel data analysis for a novel period between 2010 and 2014, the findings indicate that EU SMEs internationalization follows a horizontal S-curve pattern. The results show that FDI has no moderating effect on the S-shaped non-linear relationship between internationalization and performance. Against our expectations there is no significant effect of operations beyond the EU.
3
Table of Content
1.Introduction ... 5
2. Literature Review and Hypotheses ... 8
2.1. Different theories and models of SME internationalization ... 8
2.2. Determinants of SME internationalization ... 10
2.2.1. Motives of internationalization ... 10
2.2.2. Costs of internationalization ... 11
2.3. Internationalization and performance ... 12
2.3.1. Degree of internationalization ... 12
2.3.2. Moderating effect of FDI activity ... 14
2.3.3. Moderating effect of FDI activity beyond the EU ... 15
3. Methodology ... 16
3.1. Sample and data collection ... 16
3.2. Variables ... 16 3.2.1. Dependent Variables ... 16 3.2.2. Independent Variables ... 16 3.2.3. Control Variables ... 17 3.3. Model ... 18 3.4. Assumptions ... 19 3.4.1. Normality ... 19 3.4.2. Multi-collinearity ... 19 3.4.3. Homoskedasticity ... 20 3.4.4. Autocorrelation ... 21 4. Results ... 21 4.1. Descriptive Statistics ... 21 4.2. Regression results... 23 4.3. Robustness Check ... 27 5. Discussion ... 31 6. Conclusion ... 34
6.1. Limitations and Future Research ... 34
Appendices ... 36
Appendix 1: Previous research on S-shaped relationship ... 36
Appendix 2: Normality... 37
4
Appendix 4: Homoskedasticity ... 40
Appendix 5: Autocorrelation ... 41
Appendix 6: Descriptive statistics... 42
Appendix 7: List of countries ... 43
5
1.Introduction
Globalization from the business perspective refers to “the tendency (of firms) to configure their activities on a worldwidebasis and to co-ordinate and integrate their strategies and their operations across national boundaries” (Stonehouse et al., 2007, p. 5). In the globalized world whether or not a firm wants to join to the global business, it cannot avoid the effects of the already joined firms. As a result, firms are all affected by these interactions on different levels whether they focus on the local market or already entered the international market (Gabrielsson et al., 2004).
In this globalized world the future growth strategy of the European Union (EU), Europe 2020, aims the EU to become a smart, sustainable and inclusive economy. According to the European Commission publication international trade flows play a crucial role in this process. Moreover, the importance of small and medium-sized enterprises (SME) in the region is also decisive and the Small Business Act for Europe focuses on SME supports. The growth and performance potentials of SMEs are important determinants of the future of the European economy (European Commission, 2011). Traditionally, SMEs have been considered as weak participants among firms with international operations, but as it can be seen in previous analyses (Calof, 1994 and Zucchella, 2001) firm size and export intensity are not correlated, meaning that SMEs can be also strong participants in international business (Majocchi and Zucchella, 2003). Therefore, this paper aims to investigate the relationship between international diversification and economic performance of SMEs from the EU. By doing so, we get a picture about how SME internationalization influences firm’s performance and help the EU to reach the future goals.
6
business to the international market need to develop new knowledge and further capabilities to overcome the differences between the different markets (Lu and Beamish, 2001).
Several approaches discuss the diverse modes of internationalization and the reasons behind the chosen forms of internationalization. This paper discusses the stage, learning, contingency, foreign direct investment (FDI) and network approach and applies the learning approach as a base of hypothesis development. The learning approach imagines the international expansion as a dynamic process during which organizational learning develops in the step-by-step expansion (Johanson and Vahlne, 1977).
In the literature there are already papers investigating the relationship between international diversification and firm performance in developed and also in emerging countries. Emerging market firms from China (Child and Rodriges, 2005), from India (Contractor et al, 2003) and from Mexico (Thomas, 2006) has been already the focus of researchers. In developed economies Lu and Beamish (2001, 2004, 2006) did an extensive research on Japanese firms, Capar and Kotabe (2003) on German firms from the service sector and Ruigrok and Wagner (2003) on German manufacturing firms. The focus of this paper is the European Union which behaves differently from individual countries but at the same time EU gets closer to become an integrated macro-regional area focusing on common future plans and growth perspectives (Majocchi and Zucchella, 2003).
7
The novelty of this paper is the investigation on the S-shaped relationship between internationalization and performance in SMEs. The applied database comprises information about the internationalization of European SMEs for a very recent period between 2010 and 2014. EU SMEs are a special case because we can distinguish between two cases, namely when we focus on the geographical diversification of individual countries in the EU and the second one where we make a difference between the internal and external market of the EU. If we consider global internationalization it is worth to investigate the effect of EU SMEs going beyond the EU border. So this paper focuses first on the effect of FDI on performance where there is no distinction between internationalization within or beyond the EU and then the effect of going beyond the EU where the focus is on the effect of not only operating in the EU but moving beyond the border.
Thus, the research question of this paper is: What is the shape of the internationalization-performance relationship in SMEs from the EU? To answer this question the following three sub-questions will help: (1) Is the relationship between internationalization and performance following a horizontal S-curve pattern? (2) Is this relationship influenced by the presence of FDI activity? (3) Is this relationship moderated by operations beyond the EU border? To be able to test these hypotheses the research is using the Orbis database by Bureau van Dijk (BvD). Using a panel data, for the 5 year period between 2010 and 2014 for EU SMEs from manufacturing industry, it makes possible to capture the effect of internationalization on economic performance. The sample contains 4075 SMEs from the European Union.
The results indicate an S-shaped curvilinear relationship between geographical diversification and economic performance for SMEs from the EU but no statistically significant moderating effect of FDI activity on the S-curve. The effect of FDI on performance shows positive and significant results. Moreover, based on the results there is no significant effect of FDI operations beyond the EU.
8
2. Literature Review and Hypotheses
2.1. Different theories and models of SME internationalization
Research on internationalization of large firms has a long history relative to the SME internationalization. There are several approaches about the internationalization of SMEs, like stage, learning, contingency, network approach discussed in the paper of Korsakienė and Tvaronavičienė (2012) and the foreign direct investment theory. In the following we can see the different theories and methodologies applied by the previous mentioned approaches which lead to a better understanding in the motives behind internationalization.
The stage approach is the earliest among the theories of internationalization. In this approach firms start with entry modes where there is low level of resource commitment. After gaining experience about the foreign market they start to involve in activities with more intensive resource commitment (Korsakiené and Tvaronavičienė, 2012). Researchers supporting this approach were Bilkey and Tesar (1977) with the theory of diffusion of innovation and Reid (1991) with the three stage model where firms move from awareness of export potentials to intention to acceptance where firm thinks of exporting as a positive thing.
9
cumulative, path-dependent process where firm behaviour is a function of its previous international experience and knowledge.
The contingency approach states that firms first evaluate and then respond to new opportunities. The decision of the firm is not influenced by the physic distance or the barriers of new market entry. There are internal and also external factors which lead firms to leapfrog stages of internationalization and start international activity on the stage which is the most fruitful for the organization (Okoroafo, 1990).
Recent research investigating the network approach elaborates on a non-hierarchical system and in this system it becomes more important for firms to build and monitor their role in the international networks (Coviello and McAuley, 1999). The main focuses of this approach are the relationships and linkages of internationalization and by starting operations on foreign markets these factors help firms to become successful (Johanson and Mattsson, 1993). Different possible ways of internationalization are: building relationships on the new market, with increasing the commitment in already existing networks and with position integration in different countries (Chetty and Blankenburg-Holm, 2000). Johanson and Mattsson (1988) list the three different entry modes in terms of the position in the international network. The first is the international extension where the network is new to the company. The second one is the international penetration where firms building upon an existing status in the international network. The last one is international integration where the firm expands the coordination in different national levels. Johanson and Mattsson (1988) note that internationalization is led by not only outward-driven activities but also inward-driven ones. Furthermore, they are involved in activities where import and export operations are related to each other. So firms have to collect information about other companies on the market. To conclude, in this theory not the firm- specific advantags are in centre, it is more important to focus on the network relationships in a given company.
10
advantages are specific assets that make the differentiation from other firms possible, like patent, brand, technology etc. The location specific advantages are natural or human resources, commercialization benefits or better taxation rates in the host country. Lastly, internationalization advantages are the decisions of firms to generate and exploit the advantages themselves, like overcoming risk aversion or avoiding knowledge transfer (Dunning, 2001).
2.2. Determinants of SME internationalization
Several advantages and disadvantages are associated with geographical diversification. Researchers have investigated the effects of internationalization on national and also on firm level. An increase in national exports enhances productivity performance, more precisely the social prosperity, accumulation of foreign exchange and employment levels (Morgan and Katsikeas, 1997). The following subsections discuss the motives and costs of internationalization on firm level.
2.2.1. Motives of internationalization
11
and operational flexibility (Caves, 1996; Kogut, 1985; Kim et al., 1993; Kogut and Zander, 1993).
2.2.2. Costs of internationalization
There are also difficulties associated with SMEs when they enter the foreign market: liability of foreignness (Zaheer, 1995; Zaheer and Mosakowski, 1997) newness (Stinchcombe, 1965; Lu and Beamish, 2004) and liability of smallness (Lee et al., 2012). Starting operations on a foreign market results in new challenges for managers, like in purchasing, searching for new employees, building up internal and external networks (Barkema et al., 1996).
12
2.3. Internationalization and performance
2.3.1. Degree of internationalizationResearch on international diversification shows mixed results and the results can depend on the temporal stage being considered as a starting point of the observation (Thomas, 2006). According to Lu and Beamish (2004) there is a group of researchers who found positive linear relationship (Grant, 1987; Grant et al., 1988; Han et al., 1998), then another group argued that there is a U-shaped curvilinear relationship (Qian, 1997; Ruigrok and Wagner, 2003) and another reports about an inverted U-shaped relationship (Chiao et al., 2006; Qian, 2002; Hitt et al., 1997). Other group of researchers found a horizontal S-shaped relationship (see Appendix1).
The expression ‘unified three-stage theory of international expansion’ used by Contractor et al. (2003) denotes the absence of integrity in previous works, namely that they focus on different stages of internationalization by hypothesizing linear, curve or inverted curve relationship. Linear relationship focuses only one stage and the two different types of U-curve relationships on two stages of the three-stage theory. Thomas (2006) states that the sigmoid or three-stage models show that at low level of internationalization firms experience negative effects because they face with liability of newness and foreignness, after getting more experienced firms realize positive returns of internationalization but only to the optimal point where the cost can outweigh the benefits (Contractor et al., 2003; Lu and Beamish, 2004). Previous studies found linear positive relationship (which is capturing stage 2), U-shaped relationship which includes stage 1 and 2 combined and inverted U-shape which is stage 2 and 3 combined ( Contractor et al., 2003).
13
The second stage is the second half of the U-shape relationship where firms face with positive returns on internationalization. Thomas (2006) summarizes these findings like early stages can cause negative returns because of the costs at the beginning in connection with lack of knowledge and experience. By operating on the foreign market firms can adapt to the new market and they are able to overcome the challenges and be able to use the positive effect of exploration and exploitation (Rugman, 1979). The learning approach (Johanson and Vahlne, 1977) suggests that by operating on a foreign market organizational learning occurs and next to the general knowledge, market-specific knowledge can be also gained. As a conclusion, with knowledge about the new market firms are able to overcome the difficulties associated with liability of foreignness and newness. The U-shaped and the inverted U-shaped models are connected at this positive tendency of internationalization on firms’ performance.
The third stage is the second part of the inverted U-shape where the benefits associated with internationalization reach the optimal point and further increase in diversification will result in higher managing locational diversity costs (Geringer et al., 1989; Gomes and Ramaswamy, 1999). Hitt et al., (1997) and Tallman and Li (1996) argues that although the initial liability of foreignness costs are diminished by experience on the new market transaction and coordination costs are increasing with higher level of international diversification. They also state that the growth of resource commitments and internal capabilities cannot be slower than the rhythm of the internationalization otherwise the costs outweigh the possible benefits. If this expansion means geographic diversification firms have to face the following challenges: higher coordination costs because of the need to manage export operations, managerial capabilities are hard to spread and there is a need of culturally diverse information-processing capabilities (Aulakh et al., 2000). Bergh and Lawless (1998) found that there are limits in the efficiency of hierarchical governance and that environmental uncertainty increased the costs.
Following the previous mentioned arguments a horizontal S-shaped relationship is hypothesized between the level of geographical diversification and economic performance.
H1: Internationalization has a non-linear effect on explaining performance, the slope is negative at low level, positive at medium level and negative at high level of
14
2.3.2. Moderating effect of FDI activity
The learning approach, as discussed before, argues that internationalization process is a slow, step-by-step expansion during which organizational learning occurs (Johanson and Vahlne, 1977). Erminio and Rugman (1996) call export activity as a "stepping-stone" toward foreign direct investment and exporting is seen as a platform for other forms of internationalization (Kogut and Chang, 1996). As already mentioned before, it can lay down the basics and improve capabilities for further, more risky internationalization forms (Root, 1994). SMEs starting with export operations have the chance to collect information about the market (Root, 1994) and they can use the existing production facilities which make them possible to broaden the customer base before starting more risky forms of internationalization, like FDI (Lu and Beamish, 2006). Firms starting with exporting are able to learn new methods and can learn from the feedback of other participants on the market (Jovanovic and Lach, 1991). This ‘learning by exporting’ refers to the different mechanisms that can increase productivity when after starting an export activity, the firm invests in marketing, develops the product quality and innovates (Loecker, 2013).
Many mistakes of FDI can be avoided by gaining experience first through exporting and helps the firm perform better (Erminio and Rugman, 1996). The entry mode of firms depends on their resource abundance and according to Dalli (1995) SMEs are tend to start export activity against FDI if they have limited resources. According to these findings firms gain valuable experience during their export operations which helps them in further internationalization steps, like FDI activity, so following the S-curve, FDI will cause a negative moderating effect in stage 1 where firms have not laid down the basic experience and knowledge on the foreign market. In stage2 where the internationalization-performance relationship tends to be positive, the effect of FDI on the more experienced firms’ performance will be positive. This influence of FDI activity on the internationalization-performance relationship in stage 3 will be negative, following the negative effects of surpassing the optimal level of international expansion because additional FDI activity increases the costs more intensive.
H2: The non-linear relationship between internationalization and performance is negatively moderated by FDI activity at low level, positively on medium level and
15
2.3.3. Moderating effect of FDI activity beyond the EU
According to a European Commission (2011) publication 14% of SMEs import from third markets; 13% export to third markets; 3% are engaged in different forms of international (technical) cooperation;less than 1 % have own establishments in third markets. Most of the firms expand their operations within the EU but they are not going beyond the EU. Operating on the internal market of the European Union can be also seen as a stepping stone for SMEs to perform successfully on the external market (European Commission, 2007). As argued by Majocchi and Zucchella (2003) the definition of ‘export market’ has changed in the EU because of the steps toward an integrated market. As a conclusion the EU behaves as an internal market and recently export activities are considered exports beyond the EU “borders”.
These firms which follow international operations beyond the EU need to possess significant and abundant internal resources (Chen and Chen, 1998) but operations on new markets mean for these firms the exploitation of economies of scale and scope (Majocchi and Zucchella, 2003). More precisely, firms starting operations in foreign countries can increase their competitiveness which leads to higher economic performance and internationalization correlates strongly with increase in turnover growth (European Commission, 2010). Internationalized SMEs have an access to the know-how and technology, economies of scale not only on the domestic market but also in the host country and their competence is increasing by starting operations on a foreign market (European Commission, 2007). If we think about the EU as the “host country” of the SMEs or considering the EU as an integrated macro-regional area (Majocchi and Zucchella, 2003) countries beyond the EU border can provide something new in the previous mentioned advantages.
As mentioned in the paragrah about the costs of internationalization, the European Commission (2011) lists the most important barriers for internationalization beyond the internal market of EU. These are: payment risks, the difficulty of paperwork, lack of financing and adequate market information, the different regulations and national technical standards.
Firms which are able to overcome the barriers associated with third markets have an access to the advantages in competitiveness and turnover growth. Moreover, Majocchi and Zucchella (2003) found that SMEs exporting the the USA are more profitable than firms operating only on the EU market. Based on these findings the following hypothesis has been created:
16
3. Methodology
3.1. Sample and data collection
To test the hypotheses this research uses the Orbis database which contains data about global companies, presented by Bureau van Dijk (BvD). Orbis provides company information across the globe. There are financial, market and other information on companies worldwide and the number of private companies reached by the dataset is over 175 million. We can choose regions in countries, countries or world regions as a base of our analyses. It includes financial data about companies in standardized formats and detailed company ownership structure which enables me to test my hypothesis. This paper is observing the time period between 2010 and 2014 which means a five year panel of data. The focus is on active manufacturing firms in the EU.
Based on the fact that this research is interested in SMEs the number of employees has been set to be between 10 and 250 employees. The criteria for defining SMEs differs from country to country. This paper follows the definition of the European Commission Publication (2009), namely that small and medium sized enterprises have 250 or less and 10 or more employees. The lower limit with 10 employees aims to exclude micro firms. The whole sample contains 4076 SMEs.
3.2. Variables
3.2.1. Dependent Variables
Previous studies applied different measures to capture firm’s performance (Hult et al., 2008). 96 different research papers has been examined in the work of Hult et al. (2008) and the results on firm-level shows that the two most commonly used methods are sales-based measures and return on assets measurements. As mentioned before the initial source of the advantage of exploitation and exploration originates from the possible exploitation of market imperfections in the use of intangible assets across countries (Caves, 1971). Sales-based measures are applied and therefore this research applies the natural logarithm of operating revenue and also the natural logarithm of sales to capture firm’s performance.
3.2.2. Independent Variables
17
internationalization (DOI) in their research which includes a ratios about the foreign indicators divided by the total indicators (foreign sales/total sales (also used by Majocchi and Zucchella, 2003)); number of foreign employees/number of total employees and number of foreign offices /number of total offices). Based on this idea internationalization will be measured not simply by export revenues but by the ratio of export revenues divided by operating revenues. Data has been collected for every year between the period 2010 and 2014. To be able to investigate the S-shaped curvilinear relationship between internationalization and economic performance first-, second- and third-order terms has been created for the different stages of internationalization. Also the interaction effect with FDI and Beyond EU dummies has been created for different stages of internationalization.
For FDI activity a year dummy has been created between 2010 and 2014, coded 1 if there was FDI activity in a given year and 0 otherwise. FDI has been also used as a moderator on the relationship between geographical diversification and firm performance.
Another variable is the so called “beyond the EU” variable which is also a year dummy between the period 2010 and 2014. It is coded 1 if there was FDI beyond the EU borders and 0 if firms are only operating on the internal market. The moderating effect of operations beyond the EU has been also investigated on the relationship between internationalization and firm performance.
3.2.3. Control Variables
This paper uses the following control variables to investigate more deeply the relationship between the dependent variable and independent variables: firm size and year effect.
18
3.3. Model
The investigated period of this research is between 2010 and 2014 which means that our sample contains time-series and also cross-sections, so this paper uses panel data to detect the effect of internationalization on firm performance. As can be seen in other paper (e.g. Contractor et al., 2003, Lu and Beamish, 2004) using panel data is valid.
To run the regressions we have to decide whether it is more appropriate to use the fixed effect model (FE) or the random effect model (RE). FE allows for different intercepts for the individuals and also controls for time-invariant, in other words omitted variables, so the model captures the real effects on the dependent variable. It is preferred if we need inferences about the individual units. On the other hand with this model because of the fixed intercepts we are not able to draw inference about the population. FE is also less efficient if the variables have low variance within (Plümper and Troeger, 2007).
In contrary, RE implies that individuals are randomly selected and deals with differences among individuals as random and uncorrelated with the variables. If the differences between random individuals have an impact on the dependent variable RE model is preferred. This model makes possible to draw inferences about the population. The results can be biased if there is a presence of endogeneity (Hill et al., 2012).
By using the Hausman test, we are able to choose and apply the more suitable model in our research. Null hypothesis states that the RE is preferable. The results show that we can reject the null hypothesis (see Table 4.2 and Table 4.3), meaning that this research will apply fixed effect models.
This paper investigates the effect of different independent variables on the two dependent variables, namely revenue and sales which are indicators of firms’ performance. The research will apply the following models for the two dependent variables:
19
Sales𝒊t = 𝛽0 + 𝛽1Internationalization𝒊t + 𝛽1(Internationalization)2𝒊t+ 𝛽2(Internationalization)3𝒊t + 𝛽3FDI𝒊t+ 𝛽4 (Internationalization*FDI) 𝒊t + 𝛽5 (Internationalization2*FDI) 𝒊t + 𝛽6 (Internationalization3*FDI) 𝒊t + 𝛽7 BeyondEU𝒊t+ 𝛽8 (Internationalization*BeyondEU) 𝒊t + 𝛽9 (Internationalization2*BeyondEU) 𝒊t + 𝛽10 (Internationalization3*BeyondEU) 𝒊t+ 𝛽11 Firm Size𝒊t+ 𝛽12 Year𝒊t+ εit
3.4. Assumptions
To avoid assumption violations which can lead to biased estimates of the investigated relationship we have to test our models for normality, multi-collinearity, homoscedasticity and autocorrelation (Osborne and Waters, 2002).
3.4.1. Normality
The first assumption is the normal distribution of the error term. The test of normality investigates whether or not the variables in the model have a normal distribution (Osborn and Waters, 2002). Without having normal distribution the p-values can be biased. In this paper the Skewness and Kurtosis test for normality has been used to detect the possible problem of non-normality, which occurs if the p-value is below 5% significance level and we have to reject the null hypothesis. Skewness and Kurtosis test for the dependent variables had a p-value 0.000 (see Appendix2) which indicates that our distribution is non-normal. To solve this problem the natural logarithm of the dependent variables has been used which provides a more bell-shaped distribution of revenue and sales (see in Appendix2). In order to get a more bell-shaped distribution for firm size, the natural logarithm has been applied (see in Appendix2). In small databases it is more important to have a normal distribution because the residuals in large databases average around the mean value, so normal distribution is not required in sufficiently large data samples, like the database used during these analyses (Lumley et al., 2002).
3.4.2. Multi-collinearity
20
Table 3.1: Variance Inflation Factor
Revenue and Sales
Variable VIF 1/VIF
Internationalization 1.03 0.967899 FDI 3.34 0.299089 Beyond the EU 1.53 0.652018 Firm Size 1.01 0.989535 Year 2011 1.61 0.621900 Year 2012 1.70 0.587515 Year 2013 2.51 0.398135 Year 2014 3.45 0.289479 Mean 2.02
Note: Correlation-matrix in Appendix3 Table 5.1.
Table 3.1 indicates a problem in the VIF value of FDI activity and Year 2014 but based on the fact that these variables are dummy variables a high VIF value is also acceptable (Allison, 2012). As a conclusion all of the independent variables can be used in the models without variables being redundant with one another.
3.4.3. Homoskedasticity
The third assumption, homogeneity, assumes that the variance is constant across the levels of the predicted variables (Weisberg, 2005) .In the case of heteroskedasticity this variance is not constant, so the model is not well-fitted (Chen et al., 2003). This effect will be tested by the modified Wald test for groupwise heteroskedasticity which result can be seen in Table 3.2 and Table 3.3 for Model 1.
Table 3.2: Modified Wald Test for Groupwise Heteroskedasticity with the Dependent Variable Revenue
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model
H0: sigma(i)^2 = sigma^2 for all i
chi2 (4075) 6.9e+06
Prob>chi2 0.0000
Table 3.3: Modified Wald Test for Groupwise Heteroskedasticity with the Dependent Variable Sales
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model
H0: sigma(i)^2 = sigma^2 for all i
chi2 (4075) 1.1e+07
21
As also in other models (see Appendix 4) there is a violating effect of heteroskedasticity in the models because we have to reject the null hypothesis because the values for the Wald test are below the 5% significance level. To be able to mitigate the effect of heteroskedasticity the models will be run with robust standard errors.
3.4.4. Autocorrelation
The assumption of autocorrelation assumes correlation of firms over time. This paper uses panel data which includes time-series and cross-sectional data so it might be the case that a good performing firm will have a tendency to perform well in the upcoming years (Hill et al., 2012). To check the presence of autocorrelation the Wooldridge test for autocorrelation has been used, where the null hypothesis stands for no autocorrelation.
Table 3.4: Wooldridge Test For Autocorrelation in Panel Data with Dependent Variable Revenue
Model 1: Wooldridge test for autocorrelation in panel data H0: no first-order autocorrelation
F( 1, 4074) 331.438
Prob > F 0.0000
Note: Results for other models are shown in Appendix5
Table 3.5: Wooldridge Test For Autocorrelation in Panel Data with Dependent Variable Sales
Model 1: Wooldridge test for autocorrelation in panel data H0: no first-order autocorrelation
F( 1, 4074) 388.386
Prob > F 0.0000
Note: Results for other models are shown in Appendix5
As shown in Table 3.4 for revenues and Table 3.5 for sales we have to reject the null hypothesis which means that there is autocorrelation in our model. By applying robust standard errors we can deal with the presence of autocorrelation.
4. Results
4.1. Descriptive Statistics
22
maximum and minimum values of the data. For the variables there is a distinction between the overall, the between and the within values.
Table 4.1: Descriptive Statistics
Mean Std. Dev. Minimum Maximum Observations
Revenue overall 8.762106 1.167558 3.876412 15.40453 N = 20375 n = 4075 T-bar = 4.99975 between 1.143602 5.437184 15.24984 within 0.2359601 4.842468 11.25491 Sales overall 8.732624 1.171718 3.11985 15.40325 N = 20373 n = 4075 T-bar = 4.99951 between 1.146404 4.813424 15.24898 within 0.2427775 4.82111 11.92317 International.overall 0.3578813 0.3322673 0 1.290071 N = 20374 n = 4075 T-bar = 4.99975 between 0.3252622 0.0006299 1.095971 within 0.0680376 -0.4323085 1.126056 FDI overall 0.0107981 0.1033538 0 1 N = 20374 n = 4075 T-bar = 4.99975 between 0.0701116 0 1 within 0.0759598 -0.7892019 0.8107981 Beyond EU overall 0.1783784 0. 3831763 0 1 N = 555 n = 111 T = 5 between 0.226196 0 0.8 within 0.3098853 -0.6216216 0.9783784
Firm Size overall 3.830371 0.8299536 2.302585 5.521461 N = 20375 n = 4075 T = 5 between 0.8147297 2.302585 5.51742 within 0.1586466 1.644865 5.02983 Year overall 2012 1.414248 2010 2014 N = 20380 n = 4076 T = 5 between 0 2012 2012 within 1.414248 2010 2014
Note: N=number of observations, n=number of firms, T=average number of years a firm was observed More detailed descriptive statistic in Appendix 6.
As can be seen in Table 4.1 the mean of the dependent variable revenue, was a positive value 8.762 ranging from 3.876 to the maximum value 15.405. The maximum value of revenue belongs to a French firm, Scamark, in 2013. The other dependent variable, sales, has a mean value of 8.733 and it ranging from 3.12 to 15.403 in the same company. Using a natural logarithm of these variables makes the interpretation more difficult but the positive mean values are in the line with the results from the normality test (see Appendix2).
23
firms with FDI activity). It can be seen that only 555 observations are for this variable, this indicates that only firms with FDI are observed with this variable.
There are two control variables in the model, namely the firm size and the year dummy. The firm size is measured by the number of employees in the given firm. The natural logarithm of firm size has been applied to get a more bell-shaped distribution for normality. That is shown in the table but using the absolute values, the minimum value of firm size is 10 employees and the maximum is 250. The absolute average value of firm size is 64, indicating that the firms in this sample have on average 64 employees.
This table includes next to the overall variation the between and within variations of variables which contains extra information if the between or the within variation is zero. If the within variation would show zero, this would indicate that the variable does not vary between panels so it is time invariant and it will be dropped in the FE model. All variables used in the analyses have higher value than zero. The minimum value of FDI and Beyond EU dummies are negative but these values refers to deviation from each firm’s average, so negative values are acceptable. The between variation by year is zero, the minimum value is the same as the maximum value. In the case of year every variation is within (between the individuals).
4.2. Regression results
24
(1000$) increase in the second- order term of internationalization would cause 152.6% increase in the revenues of the firm. For the last stage where the cubic term is applied, 1 unit (1000$) increase would cause 93.8% decrease in the revenues.
Hypothesis 2 is tested in Models 4-7 where the effect of FDI activity on the relationship between internationalization and performance has been investigated. More precisely, in Model 4a the single effect of FDI activity on revenues ( in Table 4.2) is investigated and the results show positive effect meaning that if a firm has FDI activity it results in 2.74% increase in revenues but these results are statistically not significant. Model 5a focuses on the effect of FDI on the linear relationship between internationalization and revenues. The regression results do not support hypothesis 2. The next two models, Model 6a and 7a, captures in hand with Model 5a the effect of FDI activity on the S-shaped non-linear relationship. The signs of the interaction effects in Model 7a show the same signs as it was in the case of internationalization, so the interaction with FDI would strengthen the trends in the S-curve if the results would show significance.
Hypothesis 3 is captured by Model 8a-11a where the effect of FDI operations beyond the EU has been investigated on the internationalization - revenues relationship. Model 8a includes the effect of beyond EU operation on revenues but as it can be seen in the table the positive effect of these operations are not significant. The next Model, Model 9a, focuses on the moderating effect of FDI activity beyond the EU on the linear relationship and Model 10a-11a also with Model 9a investigates the S-shaped curvilinear relationship. According to the regression results neither the linear nor the non-linear effect indicates statistically significant results for the moderating effect of FDI operations beyond the EU.
25
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 4.2 Ordinary least square regressions of revenues between 2010-2014
Model(1a) Model (2a) Model (3a) Model (4a) Model (5a) Model (6a) Model (7a) Model (8a) Model (9a) Model (10a) Model (11a)
FE FE FE FE FE FE FE FE FE FE FE Internationalization 0.146*** 0.0321 -0.482*** 0.145*** 0.145*** 0.0299 -0.480*** 0.0317 0.00839 0.257 0.583 (0.0498) (0.0952) (0.163) (0.0498) (0.0499) (0.0955) (0.163) (0.125) (0.130) (0.249) (0.453) Internationalization^2 0.122 1.526*** 0.124 1.517*** -0.312 -1.168 (0.110) (0.378) (0.110) (0.379) (0.221) (1.198) Internationalization^3 -0.938*** -0.932*** 0.542 (0.241) (0.242) (0.739) FDI 0.0274 0.00603 0.00592 0.0106 (0.0216) (0.0373) (0.0482) (0.0533) Internationalization* 0.0514 0.0529 -0.00643 FDI (0.0641) (0.228) (0.530) Internationalization^2* 0.000356 0.175 FDI (0.227) (1.374) Internationalization^3* -0.147 FDI (0.950) Beyond EU 0.00125 -0.0707 -0.149* -0.0763 (0.0360) (0.0645) (0.0763) (0.0696) Internationalization* 0.151 0.566** -0.382 Beyond EU (0.0972) (0.277) (0.814) Internationalization^2* -0.410 2.072 Beyond EU (0.260) (2.161) Internationalization^3* -1.701 Beyond EU (1.458) Firm Size 0.633*** 0.634*** 0.634*** 0.633*** 0.633*** 0.634*** 0.634*** 0.394*** 0.389*** 0.384*** 0.377*** (0.0290) (0.0289) (0.0289) (0.0290) (0.0290) (0.0289) (0.0289) (0.144) (0.144) (0.143) (0.143) Constant 6.276*** 6.284*** 6.309*** 6.276*** 6.276*** 6.284*** 6.309*** 8.359*** 8.392*** 8.388*** 8.397*** (0.110) (0.111) (0.111) (0.110) (0.110) (0.111) (0.111) (0.634) (0.636) (0.635) (0.636) Wald year-dummies
joint significance test
26
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 4.3 Ordinary least square regressions of sales between 2010-2014
Model(1b) Model (2b) Model (3b) Model (4b) Model (5b) Model (6b) Model (7b) Model (8b) Model (9b) Model (10b) Model (11b)
FE FE FE FE FE FE FE FE FE FE FE Internationalization 0.352*** -0.0524 -0.372** 0.352*** 0.351*** -0.0550 -0.370** 0.295** 0.278** 0.303 0.195 (0.0527) (0.0997) (0.172) (0.0527) (0.0528) (0.100) (0.173) (0.125) (0.131) (0.258) (0.479) Internationalization^2 0.435*** 1.309*** 0.437*** 1.297*** -0.0438 0.230 (0.111) (0.397) (0.111) (0.398) (0.213) (1.269) Internationalization^3 -0.584** -0.575** -0.179 (0.249) (0.249) (0.801) FDI 0.0351* 0.0168 0.0127 0.0283 (0.0200) (0.0367) (0.0480) (0.0505) Internationalization* 0.0438 0.0743 -0.153 FDI (0.0646) (0.211) (0.478) Internationalization^2* -0.0265 0.624 FDI (0.210) (1.275) Internationalization^3* -0.487 FDI (0.903) Beyond EU -0.00780 -0.0628 -0.113 -0.0755 (0.0343) (0.0662) (0.0824) (0.0790) Internationalization* 0.116 0.392 -0.0202 Beyond EU (0.0973) (0.270) (0.801) Internationalization^2* -0.276 0.748 Beyond EU (0.243) (2.120) Internationalization^3* -0.689 Beyond EU (1.438) Firm Size 0.641*** 0.644*** 0.644*** 0.641*** 0.641*** 0.644*** 0.644*** 0.428*** 0.424*** 0.423*** 0.423*** (0.0299) (0.0297) (0.0297) (0.0299) (0.0299) (0.0297) (0.0297) (0.144) (0.144) (0.144) (0.144) Constant 6.143*** 6.171*** 6.187*** 6.143*** 6.143*** 6.172*** 6.188*** 8.052*** 8.077*** 8.084*** 8.092*** (0.113) (0.114) (0.114) (0.113) (0.113) (0.114) (0.114) (0.628) (0.631) (0.636) (0.639) Wald year-dummy
joint significant test
27
revenues. Model 2b with the dependent variable sales shows significant results for the second-order term. There is also a change in the significance level in Model 3b where internationalization and the cubic term are only significant at 5% instead of the previous 1%. The effects are also weaker comparing them with the effect with the dependent variable revenue.
There is also a difference if we compare Model 4a and Model 4b with the two different dependent variables. The presence of FDI activity has not shown significant effect on revenues but the results indicate 3.51% increase in sales if the firm has FDI activity and this result is significant at 10%.
There is a distinction between the overall, the within and the between R-squared results. Generally we can conclude for both dependent variables that Model 8a-11a have lower R-squared percent than Model 8b-11b, indicating that the explanatory power of the results with sales are weaker. The within R-squared scores are above 20% in both cases. Using revenues as dependent variables in Model 1a-7a the overall R-squared revenues are close to 42% and the between R-squared values score close to 43%. In the case of sales in Model 1b-7b, the overall explanatory power of the models ranges approximately between 38%-40% and the between R-squared values approximately between 39%-40.5%.
4.3. Robustness Check
Empirical studies run robustness check to examine the possible changes in the original regression model after modifying it by including or excluding variables. Structural validity of the original regression is proved if the coefficients are plausible and robust, meaning that they do not change much after the adjustments (White and Lu, 2010).
The robustness checks applied by this paper will focus mainly on Hypothesis1 because the S-shaped curvilinear relationship between internationalization and performance gave statistically significant results.
28
the export values of firms without considering their total values for different indicators can result in changes.
The results in Table 4.4 indicate that the first-order term of internationalization in Model 1c and 2c are in both cases is positive and significant. In Model 2c using the export revenues instead of the ratio from Table 4.2 in Model 2a the results become statistically significant if we use only the export revenues. The signs are as it was expected, minus in the first-order term and plus for the squared term. The next model, Model 3c, where the whole S-curve is captured is not significant anymore, only for the cubic term. In Model 1c the effect of export revenues on revenues and sales is stronger than in Table 4.2 in Model 1a, but weaker than in Table 4.3 where sales are applied as dependent variable. In Model 4c where the effect of FDI activity on revenues is investigated shows positive and significant results as it was also in Table 4.3 with the ratio and sales. Model 6c in Table 4.4 shows significant results for the first-order term and the square term but in Model 7c the moderating effect of FDI activity on the S-shaped non-linear relationship is not supported. In Model 8c the effect of FDI activities beyond the EU is still not significant and in Model 9c-11c the moderating effect of beyond EU activities on the linear and on the S-shaped non-linear relationship is still not supported. In Table 4.5 where sales is applied as dependent variable and export revenues are used instead of the ratio, the main effects are the same. Neither the S-shaped non-linear relationship nor the moderating effect of FDI and FDI beyond the EU are supported. An interesting fact is that by investigating the effect of FDI operations on the S-shaped curvilinear relationship between export revenues and performance in Model 11c and Model 11d from Table 4.4 and 4.5 the moderating effect of beyond EU operations is not supported but the S-shaped curvilinear relationship between export revenues and performance is significant with the expected signs. The first-, and second-order terms in Table 4.4 are significant at 5% and the cubic term at 10%, in the case where sales are the dependent variable all stages of internationalization captured by export revenues are significant at 5%.
29 Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.4 Robustness check of Revenue –FE Ordinary least square regressions between 2010-2014
Model(1c) Model (2c) Model (3c) Model (4c) Model (5c) Model (6c) Model (7c) Model (8c) Model (9c) Model (10c) Model (11c) Export 0.184*** -0.164*** 0.0134 0.184*** 0.183*** -0.164*** 0.0116 0.239*** 0.236*** -0.275 -1.591** (0.0107) (0.0238) (0.0453) (0.0107) (0.0107) (0.0238) (0.0451) (0.0645) (0.0645) (0.253) (0.769) Export^2 0.0321*** -0.00471 0.0321*** -0.00429 0.0329** 0.200** (0.00229) (0.00902) (0.00229) (0.00897) (0.0162) (0.0969) Export^3 0.00219*** 0.00217*** -0.00681* (0.000550) (0.000547) (0.00394) FDI 0.0353* -0.0797 1.404** 1.217 (0.0199) (0.166) (0.715) (1.363) Export*FDI 0.0129 -0.324** -0.270 (0.0181) (0.158) (0.490) Export^2* 0.0188** 0.0140 FDI (0.00880) (0.0596) Export^3* 0.000114 FDI (0.00243) Beyond EU 0.00506 -0.248 0.511 -1.876 (0.0324) (0.187) (0.691) (1.856) Export* 0.0273 -0.162 0.710 Beyond EU (0.0197) (0.179) (0.729) Export^2* 0.0114 -0.0906 Beyond EU (0.0113) (0.0921) Export^3* 0.00384 Beyond EU (0.00377) Firm Size 0.497*** 0.432*** 0.430*** 0.497*** 0.497*** 0.432*** 0.430*** 0.276*** 0.274*** 0.284*** 0.275*** (0.0243) (0.0239) (0.0240) (0.0243) (0.0243) (0.0239) (0.0240) (0.102) (0.102) (0.0971) (0.0928) Constant 5.578*** 6.572*** 6.363*** 5.578*** 5.579*** 6.571*** 6.363*** 6.761*** 6.792*** 8.642*** 11.93*** (0.0996) (0.115) (0.116) (0.0996) (0.0997) (0.115) (0.116) (0.614) (0.617) (1.048) (1.937) Wald year-dummies
30 Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.5 Robustness check of sales- FE Ordinary least square regressions between 2010-2014
Model(1d) Model (2d) Model (3d) Model (4d) Model (5d) Model (6d) Model (7d) Model (8d) Model (9d) Model (10d) Model (11d) Export 0.194*** -0.170*** 0.00175 0.194*** 0.194*** -0.170*** -0.000187 0.262*** 0.259*** -0.309 -2.193** (0.0113) (0.0248) (0.0464) (0.0113) (0.0114) (0.0248) (0.0462) (0.0760) (0.0762) (0.277) (0.894) Export^2 0.0337*** -0.00197 0.0337*** -0.00154 0.0365** 0.276** (0.00242) (0.00928) (0.00242) (0.00923) (0.0177) (0.112) Export^3 0.00212*** 0.00210*** -0.00973** (0.000568) (0.000565) (0.00447) FDI 0.0439** -0.0129 1.455* 1.363 (0.0178) (0.165) (0.756) (1.333) Export*FDI 0.00637 -0.326* -0.309 (0.0180) (0.167) (0.468) Export^2* 0.0184** 0.0183 FDI (0.00920) (0.0566) Export^3* -7.20e-05 FDI (0.00231) Beyond EU 0.000714 -0.260 0.437 -3.195 (0.0299) (0.184) (0.739) (1.949) Export* 0.0281 -0.147 1.188 Beyond EU (0.0197) (0.190) (0.751) Export^2* 0.0106 -0.146 Beyond EU (0.0119) (0.0932) Export^3* 0.00596 Beyond EU (0.00375) Firm Size 0.503*** 0.435*** 0.433*** 0.503*** 0.503*** 0.435*** 0.433*** 0.287*** 0.285*** 0.296*** 0.283*** (0.0261) (0.0259) (0.0260) (0.0261) (0.0261) (0.0259) (0.0259) (0.103) (0.102) (0.0984) (0.0930) Constant 5.448*** 6.490*** 6.288*** 5.448*** 5.449*** 6.489*** 6.288*** 6.463*** 6.495*** 8.552*** 13.27*** (0.106) (0.123) (0.123) (0.106) (0.106) (0.123) (0.123) (0.709) (0.714) (1.151) (2.184) Wald year-dummies
31
So, we can conclude that using simply the export revenues instead of the ratio of export revenues divided by the total revenues the models capturing the S-shaped relationship between internationalization and performance is not significant anymore.
5. Discussion
For the discussion this paper analyses the findings with the guidance of the research question: What is the shape of the internationalization -performance relationship in SMEs from the EU? and the sub-questions of this paper: (1) Is the relationship between internationalization
and performance following a horizontal S-curve pattern? (2) Is this relationship influenced by the presence of FDI activity? (3) Is this relationship moderated by operations beyond the EU border?
The results of the first sub-question (1) Is the relationship between internationalization
32
diversification on performance can be realised until a certain point where internationalization reaches the optimal level and the benefits of increasing internationalization will be outweighed by the costs. This phenomenon can be seen in the third-order term where the sign changes again and indicates a negative effect for the third stage of internationalization. In general, based on Table 4.2 and 4.3 we can state that the positive square term in Model 3a than in Model 3b has the strongest effect 152.6% increase in revenues and 130.9% increase in sales if we increase the square term with 1 unit.
In Model 1a and 1b the sign of the first-order term is positive which is against our expectations. The possible reasons behind this effect has been discussed in the previous work of Lu and Beamish (2006). They investigated the positive effect of export activity (used in the ratio in Table 4.2 and 4.3 which has been used as an indicator of internationalization) and argued that this effect is because of the less complex form (compared to FDI). Exporting is a fast way to start operations on new markets and through this strategy firms can use their already existing facilities and they do not need to build new ones. Exporting contributes to sales growth because of the new market entry which provides new customer base for firms. Higher sales and production contributes to the achievement of economies of scale and scope (Kogut, 1985). This effect helps to increase also the profitability (Lu and Beamish, 2006).
33
stage. This means that after the firm succeed to overcome the first difficulties and the performance becomes positive, the moderating effect of FDI activity will also turn positive. The sign of the moderating effect in stage three is negative, indicating that after firms reach the point where increasing the level of internationalization will result in decreased economic performance, and the moderating effect of FDI activity strengths this negative effect after the optimal level.
The sign of the findings as it can be seen in the first- and second order term are also in line with the learning approach which is the base of this hypothesis because the sign is first negative which turns positive for the second stage of internationalization. Considering export activity as a ‘stepping-stone’ toward FDI (Erminio and Rugman, 1996) firms in the first stage are not ready to expand the internationalization modes because according to Johanson and Vahlne (1977) internationalization is a slow, step-by-step expansion during which organizational learning occurs so the sign is negative. In the second stage, firms gained both general and market-specific knowledge and these are important aspects of successful foreign operations (Johanson and Vahlne, 1977) so the positive sign supports this idea but unfortunately the results are not significant.
34
6. Conclusion
This research contributes to the understanding of the relationship between internationalization and performance. Previous research focused on MNEs or distinguished between firms from manufacturing or service sector, this paper investigates this issue on SMEs from manufacturing sector. This is important because SMEs are not simply smaller prototypes of MNEs, they have special characteristics. Moreover, the influential effect of FDI activity and the positive moderating effect of operations beyond the EU have been examined.
As a result, the findings of this paper contributes to the existing literature. The results indicate that the internationalization -performance relationship in SMEs tends to follow a horizontal S-curve. This means that the relationship tends to show negative pattern at low level of internationalization, positive on mid-level and the sign turns negative at high level. The findings show that FDI activity does not have significant influence on the S-curve relationship but FDI activity has positive influence on the performance (captured by sales) is we consider the direct effect instead of the moderating effect. Notable is that, however, EU supports operations beyond the EU, based on these results there is no significant positive effect of operations beyond the EU ‘border’.
6.1. Limitations and Future Research
35
would make it possible to detect whether there are some groups of countries which have stronger effect on explaining performance.
36
Appendices
Appendix 1: Previous research on S-shaped relationship
Table 1.1 Horizontal S-shaped relationship between international expansion and performance (Zhang and Toppinen, 2011)
Author Publication Year Data Source and Types Main Results
Contractor et al. 2003
Service Industries Panel Data, 1983-1988
with pooled OLS
Performance decrease at the early stage then profit overwhelm, after the
optimum point, additional operating costs
are detrimental to performance.
Thomas and Eden 2004
U.S manufacturing industry Panel Data, 1990–1994
with pooled OLS
Initial benefits at beginning then is outweighed by costs, overtime, the
long-run benefits dominate. Lu and Beamish 2004 Japanese firms Panel Data, 1986–1997 with RE
At high and low levels of internationalization, the performance was negative, but positive at
moderate level
Li 2005
U.S service industry Panel Data, 1997–2001 with FGLS A horizontal S-curved relationship between multinationality and financial performance Ruigrok et al. 2007 Swiss MNEs in manufacturing industries Panel Data, 1998–2005
with pooled OLS
A nonlinear sinus curve relationship between internationalization and
performance Source: presented in the paper of Zhang and Toppinen (2011. p.5.)
37
Appendix 2: Normality
Skewness/Kurtosis test for Normality:
Skewness/Kurtosis tests for Normality
variable Number of observations Pr(Skewness) Pr(Kurtosis)
revenue 2.0e+04 0.0000 0.0000
sales 2.0e+04 0.0000 0.0000
Histograms after taking natural logarithm:
I: Revenue
39
Appendix 3: Multi-collinearity
Table 5.1: Correlation Matrix
Correlation matrix
40
Appendix 4: Homoskedasticity
Modified Wald test for groupwise heteroskedasticity
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model
Dependent variable: Sales H0: sigma(i)^2 = sigma^2 for all i
Model 2: chi2 (4075) 1.1e+07
Prob>chi2 0.0000
Model 3: chi2 (4075) 9.0e+06
Prob>chi2 0.0000
Model 4: chi2 (4075) 1.1e+07
Prob>chi2 0.0000
Model 5: chi2 (4075) 1.1e+07
Prob>chi2 0.0000
Model 6: chi2 (4075) 1.1e+07
Prob>chi2 0.0000
Model 7: chi2 (4075) 8.9e+06
Prob>chi2 0.0000
Model 8: chi2 (111) 2.4e+05
Prob>chi2 0.0000
Model 9: chi2 (111) 1.8e+05
Prob>chi2 0.0000
Model 10: chi2 (111) 1.2e+05
Prob>chi2 0.0000
Model 11: chi2 (111) 1.6e+05
Prob>chi2 0.0000
Table 5.2: Modified Wald test for groupwise heteroskedasticity
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model
Dependent variable: Revenue H0: sigma(i)^2 = sigma^2 for all i
Model 2: chi2 (4075) 6.5e+06
Prob>chi2 0.0000
Model 3: chi2 (4075) 7.5e+06
Prob>chi2 0.0000
Model 4: chi2 (4075) 6.4e+06
Prob>chi2 0.0000
Model 5: chi2 (4075) 6.7e+06
Prob>chi2 0.0000
Model 6: chi2 (4075) 6.3e+06
Prob>chi2 0.0000
Model 7: chi2 (111) 7.3e+06
Prob>chi2 0.0000
Model 8: chi2 (111) 2.5e+05
Prob>chi2 0.0000
Model 9: chi2 (111) 5.6e+05
Prob>chi2 0.0000
Model 10: chi2 (111) 2.0e+05
Prob>chi2 0.0000
Model 11: chi2 (111) 1.6e+05
41
Appendix 5: Autocorrelation
Table 5.3: Wooldridge test for autocorrelation
Wooldridge test for autocorrelation in panel data Dependent variable: Revenues H0: no first-order autocorrelation
Model 2: F( 1, 4074) 390.963 Prob > F 0.0000 Model 3: F( 1, 4074) 389.098 Prob > F 0.0000 Model 4: F( 1, 4074) 388.386 Prob > F 0.0000 Model 5: F( 1, 4074) 388.435 Prob > F 0.0000 Model 6: F( 1, 4074) 391.076 Prob > F 0.0000 Model 7: F( 1, 4074) 389.210 Prob > F 0.0000 Model 8: F( 1, 4074) 388.386 Prob > F 0.0000 Model 9: F( 1, 110) 25.811 Prob > F 0.0000 Model 10: F( 1, 110) 26.470 Prob > F 0.0000 Model 11: F( 1, 110) 26.036 Prob > F 0.0000
Wooldridge test for autocorrelation Wooldridge test for autocorrelation in panel data Dependent variable: Sales H0: no first-order autocorrelation
42
Appendix 6: Descriptive statistics
Table 4.6: Expanded descriptive statisticsMean Std. Dev. Minimum Maximum Observations
Revenue overall 8.762106 1.167558 3.876412 15.40453 N = 20375 n = 4075 T-bar = 4.99975 between 1.143602 5.437184 15.24984 within 0.2359601 4.842468 11.25491 Sales overall 8.732624 1.171718 3.11985 15.40325 N = 20373 n = 4075 T-bar = 4.99951 between 1.146404 4.813424 15.24898 within 0.2427775 4.82111 11.92317 International. overall 0.3578813 0.3322673 0 1.290071 N = 20373 n = 4075 T-bar = 4.99975 between 0.3252622 0.0006299 1.095971 within 0.0680376 -0.4323085 1.126056 FDI overall 0.0107981 0.1033538 0 1 N = 20374 n = 4075 T-bar = 4.99975 between 0.0701116 0 1 within 0.0759598 -0.7892019 0.8107981 Beyond EU overall 0.1783784 0. 3831763 0 1 N = 555 n = 111 T = 5 between 0.226196 0 0.8 within 0.3098853 -0.6216216 0.9783784
Firm Size overall 3.830371 0.8299536 2.302585 5.521461 N = 20375 n = 4075 T = 5 between 0.8147297 2.302585 5.51742 within 0.1586466 1.644865 5.02983 Year overall 2012 1.414248 2010 2014 N = 20380 n = 4076 T = 5 between 0 2012 2012 within 1.414248 2010 2014 Export overall 6.982712 1.97319 0. 0138001 12.85315 N=20346 n=4075 T-bar = 4.99288 between 1.906766 0. 5240388 12.5182 within 0.5079929 1.283036 10.9714
Table 4.7: Detailed descriptive statistics of the dummy moderators
Observation Mean Std. Deviation Minimum Maximum
43
Appendix 7: List of countries
EU countries (selected in the database)
Countries remained after filters and adjustment
AT-Austria HR-Croatia
BE-Belgium FR-France
BG-Bulgaria GR-Greece
HR-Croatia HU-Hungary
CY-Cyprus IT-Italy
CZ-Czeck Republic GB-United Kingdom
44
References
Allison P (2012) When Can You Safely Ignore Multicollinearity? Retrieved from
http://www.statisticalhorizons.com/multicollinearity on 21.11.2015
Aulakh PS, Kotabe M, Teegen H (2000) Export Strategies and Performance of Firms from Emerging Economies: Evidence from Brazil Chile, and Mexico. Academy of Management Journal 43(3), 342-361.
Barkema HG, Bell JHJ, Pennings JM (1996) Foreign entry, cultural barriers, and learning. Strategic
Management Journal, 17: 151-166.
Bergh DD, Lawless MW (1998) Portfolio restructuring and limits to hierarchical governance: The effects of environmental uncertainty and diversification structure. Organization Science, 9(1): 87-102.
Bilkey WJ, Tesar G (1977) The Export Behaviour of Small-sized Wisconsin Manufacturing Firms.
Journal of International Business Studies, Spring/Summer, pp. 93-98.
Calof JL (1994) The relationship between firm size and export behaviour revisited. Journal of
International Business Studies 25(2): 367-387.
Capar N, Kotabe M (2003) The relationship between international diversification and performance in service firms. Journal of International Business Studies 34(2): 345–355.
Caves RE (1996) Multinational Enterprise and Economic Analysis, 2nd ed., Cambridge: Cambridge
University Press.
Caves RE (1971) International corporations: The industrial economics of foreign investment.
Economica, 38: 1-27.
Chen H, Chen TJ (1998) Network Linkages and Location Choice in Foreign Direct Investment. Journal
of International Business Studies 29(3): 445–68.
Chen X, Ender P, Mitchell M, Wells C (2003). Regression with Stata. Retrieved from
http://www.ats.ucla.edu/stat/stata/webbooks/reg/default.htm on 10.10.2015
Chetty S, Blankenburg-Holm D (2000) Internationalization of small and medium-sized manufacturing firms: a network approach, International Business Review 9(1): 77–93.
Chiao YC, Yang KP, Yu CMJ (2006) Performance, Internationalization, and Firm-Specific Advantages of SMEs in a Newly-Industrialized Economy. Small Business Economics, Vol. 26, No. 5, pp. 475-492.
45 Contractor FJ, Kundu SK, Hsu C-C (2003) A three-stage theory of international expansion: the link between multinationality and performance in the service sector. Journal of International Business
Studies 34(1):5–18.
Coviello NE, McAuley A (1999) Internationalization and the Smaller firm: A review of contemporary Empirical Research. Management International Vol. 39, No. 3 (1999 3rd Quarter), pp. 223-256.
Dalli D (1995) The organization of exporting activities: relationships between internal and external arrangements. Journal of Business Research, 34(2): 107–115.
Dana LP, Etemad H, Wright RW (1999) The Impact of Globalization on SMEs. Global Focus, 11(4): 93–105.
Das S, Chatterjee S (2011) Multicollinearity Problem - Root Cause, Diagnostics and Way Outs. Retrieved from http://ssrn.com/abstract=1830043 on 02.12.2015
Denis DJ, Denis DK, YOST K (2002) Global diversification, industrial diversification, and firm value.
The Journal of Finance, Vol. LVII, 5(10): 1951–1979.
Dunning JH (1988) The electic paradigm of international production: A restatement and some possible extensions. Journal of International Business Studies, 19, pp. 1-31.
Dunning JH (2001) The Eclectic (OLI) Paradigm of International Production: Past, Present and Future.
The International Journal of the Economics of Business 8 (2), pp. 173-190.
Erminio F, Rugman AM (1996) A test of internalization theory and internationalization theory: the Upjohn Company. Management International Review 36(3): 199–215.
European Commission (EC) (2007) Supporting the Internationalisation of SMEs, final report of the Expert Group, DG Enterprise and Industry, Brussels.
European Commission (EC) (2009) COMMISSION STAFF WORKING DOCUMENT on the implementation of Commission Recommendation of 6 May 2003 concerning the definition of micro,
small and medium-sized enterprise. Retrieved from
file:///D:/Let%C3%B6lt%C3%A9sek/sme_report_2009_en.pdf on 14.10.2015
European Commission (EC) and EIM Business & Policy Research (2010) Internationalisation of European SMEs, final report, DG Enterprise and Industry, Brussels.
