Composition of top-level management and international firm diversification The influence of top management characteristics on international diversification
of large European firms
Master Thesis International Business and Management Wouter Michiel Hansen
S2533758
w.m.hansen@student.rug.nl Supervisor: O. Lindahl Co-assessor: R. Drogendijk
MSc International Business and Management 2015-2016
Faculty of Economics and Business, University of Groningen, the Netherlands
Abstract
International expansion is starting to become a significant matter in the field of international business. This study investigates top management characteristics and the firm’s relation to foreign policies by examining the international firm diversification. These aspects are a result of the international experiences as described in the Upper Echelon theory. This study will try and discern to what extent board and CEO characteristics influence international diversification of large European firms. International diversification focuses on the effects top management characteristics on international expansion through an analysis of the geographical footprint.
Several assumptions, like the share of non-nationals in the board, number of different
nationalities and CEO duality specify that certain board and CEO characteristics lead to higher
international firm diversification. Depending on a sample of the 100 largest European firms, it
was founded that a higher share of non-nationals on a firm’s board positively influenced the
international firm diversification. Elaborating on the outcomes that indicate only a significate
and positive effect of the share of non-national in the board of directors, it can be assumed that
the different corporate governance regulations across the European continent could have a
significant importance in influencing top management characteristics and subsequently the
international diversification strategy. This could create an interesting start of further research.
ACKNOWLEDGEMENT
First of all, I would like to thank my thesis supervisor Olof Lindahl from the Department of
Business Studies at the Uppsala University / University of Groningen. Even though mister
Lindahl was working in Sweden, he was always willing to answer my questions by e-mail or
Skype. He consistently helped to navigate me in the right direction throughout the writing
process and encouraged me to make this work my own. Besides, I would like to thank the/my
co-assessor R. Drogendijk for her input. Lastly, I would like to thank dr. J. M. Hallegraeff,
another expert in this field of research, who was willing to discuss the statistical and empirical
parts of my research. He gave me a lot of advice when testing my data and helped me to master
SPSS. Without the passionate help of the aforementioned people, and their participation in the
process of writing and doing research, this paper could not have been presented in its present
form.
Table of contents
TABLE OF CONTENTS 4
1. INTRODUCTION 5
2. THEORY 8
2.1 U
PPERE
CHELONT
HEORY8
2.2 F
UNCTION OF THE BOARD ANDCEO 8
2.3 I
NTERNATIONAL FIRM DIVERSIFICATION9
3. HYPOTHESES 9
4. CONCEPTUAL MODEL 12
5. METHODS 13
5.1 D
ATASET13
5.2 T
IMEFRAME14
5.3 V
ARIABLES15
6. R
ESEARCH MEASURE16
7. E
VALUATION OF METHOD ASSUMPTIONS17
8. E
MPIRICAL RESULTS20
8.1 D
ESCRIPTIVE STATISTICS20
8.2 R
EGRESSION RESULTS21
9. DISCUSSION 24
10. IMPLICATION 27
11. LIMITATIONS AND SUGGESTIONS FOR FURTHER RESEARCH 29
12. REFERENCES 30
13. APPENDICES 35
A
PPENDIXA – E
UROPEAN TOP100
FIRMS36
A
PPENDIXB – C
OOK’
S DISTANCE37
A
PPENDIXC – T
EST FOR HOMOSCEDASTICITY37
A
PPENDIXD – S
CATTERPLOT ANDN
ORMALP-P
PLOT38
A
PPENDIXE – C
ORRELATION MATRIX38
A
PPENDIXF – ANOVA
ANDC
OEFFICIENTSE
NTER-M
ETHOD39
A
PPENDIXG – ANOVA
ANDC
OEFFICIENTSS
TEPWISE-M
ETHOD40
1. Introduction
International business research aims at identifying strategic motivations, intentions and characteristics of entering new foreign markets and industries (Hamel & Prahalad, 1989).
Consequently, internationalization of the global markets has been identified as one of the most significant developments in the business environment over the last decade (Earley and Gibson, 2002). Especially in Europe, hosting several important world economies and financial hubs, the effects of international diversification can be highly significant for firm growth. According to the resource-based view (Barney, 1991), international firm diversification exposes firms to new resources and capabilities that can be leveraged across new markets (Hitt et al., 2006; Geringer, Tallman, and Olsen, 2000; Penrose, 1959; Teece, 1980). International diversification is considered to be the diversification of business activities across national borders. This complex expansion is a result of international firm diversification and creates opportunities and threats which top management encounters in international markets (Hitt et al., 2006; Tihanyi et al., 2000). Moreover, the transaction-cost theory declares that international diversification can improve resource allocation by exploiting and exploring in international markets (Eisenmann, 2002; Williamson, 1975). Consequently, both mechanisms suggest the importance of international diversification as a driving force behind and motive for firm and market growth (Bowen and Wiersema, 2009; Buckley and Casson, 2009; Hitt et al., 2006). In addition, in top management, the board and CEO are of high influence on firm strategies and decision-making.
Besides, they function as the central subject of the international diversification strategy.
Consequently, international diversification is an important aspect of the business strategy, as it provides the firm with growth potential and access to new knowledge and foreign locations (Contractor, et al., 2007). Therefore, understanding the decision-making in the international diversification process is critical according to many recent researchers (Wiersema and Bowen, 2011; Hutzschen- Reuter and Groene, 2009; Peng and Delios, 2006; Peng, Lee, and Wang, 2005; Wiersema and Bowen, 2007).
This study attempts to analyze the key elements of top-level management in the international
diversification strategy. Extensive effort has been made to elaborate on the work of Hambrick
and Mason (1984) to connect the Upper Echelon theory to the role of top management and
corporate governance strategies in the international organization. Since corporate governance
concerns the structure, rights, responsibilities and strategies enhanced by the firm, it can be seen
as a result of its top management characteristics (Aoki, 2000). Therefore, this study contributes
to the ongoing debate on international diversification in relation to board and CEO characteristics. This study tries to answer whether or not certain top management characteristics influence international diversification, and how they influence international diversification. In addition, this study extent the current efforts of many researchers (Finkelstein and Hambrick, 1990; Cho and Chen, 1996; Tihanyi, 2000; Aguilera and Jackson, 2003; La Porta et al., 1999;
Aoki, 2000; Roy 1997), who still leave unsatisfying findings and unnoticed explanations on multinational board characteristics as an essential key element in order to understand the corporate governance - international diversification relation (De Jong and Van Houten, 2014).
Other influences on the differences in top management characteristics are (1) the way a country its property rights, the financial system and the inter firm network influence and shape the role of capital; (2) union organizations and representation right which are of influence on the role of labor in the society or firm; and (3) the management ideology of countries and common career patterns that influence the role of management in a country (Aguilera and Jackson, 2003).
Therefore, to properly understand how top management deals with uncertainties from operating in international markets, this study focuses on the characteristics of the board and CEO. More specific, this research is based on a primary dataset of the 100 largest European firms, which makes it possible to connect some of the top characteristics of the board and CEO to international firm diversification.
The aim of the study is to examine top management characteristics by using five key variables.
These key variables appear in the ongoing corporate governance debate and will help to
improve the understanding of top-level management characteristics. This study will be able to
analyze the direct effects of international diversification as this can be examined by the
geographical footprint of the firm (Bruno and Song Shin, 2014). Consequently, this study will
examine how board and CEO characteristics influence the international diversification process
of the firm to delineate the empirical gap and understandings. The examination of this study
makes use of a primary dataset of a sample of the top 100 European firms
.Examining this
sample will lead to understanding the wide variety of industries located in a large variety of
countries on the European continent. Assessing this sample thus implies that understanding the
leading European firms’ top management characteristics and how they relate to the international
diversification strategy is possible. The creation of a new database could result in the possibility
to measure relevant aspects and could thus lead to interesting additional insights. More specific,
sourcing of primary data could generate a possible exploration of annual reports in a more
profound way and could detect differences between firms and their corporate governance characteristics (Aguilera and Jackson, 2003).
In this study, examination of the dependent variable will be completed using the direct approach; the number of different host countries measured by the geographical footprint of the firm. Consequently, it would be able to visualize a more direct effect of international diversification as a result of board and CEO characteristics. Since most countries have high international trade rates, there is a high potential of international diversification which could skew the sample towards the high-end of international diversification.
This research will make an attempt to examine which board and CEO characteristics matter in influencing the global expansion through international diversification in the international business strategy. Therefore, using board and CEO characteristics narrows down the influences of the predictor variables to a more unique examination to understand the antecedents of successful international diversification. As a result, this study could eventually assist firms planning to internationalize or trying to keep their competitive position. Furthermore, this study will give insights in selecting and composing the right CEO and board members and will function as support for human resource management in the recruiting process. Lastly, outcomes of this study will try to answer the following research question: What is the effect of board and CEO characteristics on the international diversification of European firms?
The outline of this paper is as follows. Firstly, the introduction and research question will be presented to explain and underline several influential conditions concerning international diversification. Secondly, the literature will be presented which explains the deeper understanding of the central issues. Besides, these and other key subjects will introduce the hypotheses and the conceptual model. Thirdly, the variables will be introduced which will address the insights related to the measurement and examination of the variables followed by an estimation of the methods used in this study. Fourthly, the research measure and data section aims to create and identify a deeper explanation of the research methods and dataset in this study. After this section, the focus will be on the multiple regression assumptions that need to be confirmed in order to perform the multiple regression analysis. After the confirmation, the multiple regression analysis is executed and performed in which the hypotheses are tested.
Finally, the study will be concluded by an appraisal in which the multiple regression analyses
will be analyzed, the hypotheses will be assessed and the conclusion will be given to explain
the theories used and their empirical results. The last section will cover the limitations of the examination and its outcomes. Besides, this final section will reflect on the research and provides several directions and implications and suggestions for further research.
2. Theory
To illustrate the importance of board and CEO characteristics, the following section will introduce the Upper Echelon Theory which will explain this matter. Accordingly, the function of the board and CEO will be analyzed, followed a demonstration of the significance of international diversification of the firm. Consequently, the hypotheses section will combine these noteworthy concepts with major models to help construct the hypotheses.
2.1 Upper Echelon Theory
The main principle of the Upper Echelon theory is the creation of a specific strategic choice based on the executives’ experiences and the strategic situation of the firm. Therefore, the Upper Echelon theory (Hambrick and Mason, 1984) suggests that executives’ experiences, values and personalities will greatly influence the interpretation of situations in strategic actions. As result of these past experiences they focus on the international characteristics that can significantly impact strategic firm performance. Consequently, Finkelstein, Hambrick and Cannella (2008) indicate the importance of the Upper Echelon theory to boards and how demographic characteristics and experiences could result in organizational outcomes.
Moreover, the central idea of the Upper Echelon theory demonstrates that executives’
experiences, values and personality highly influence their interpretation of different situations and affects their choices (Hambrick, 2007). Besides, Hambrick (2007) argues that demographic top management characteristics influence the executives’ actions in organizational strategies.
Above all, international experiences will function as a competitive, or even firm specific advantage (FSA).
2.2 Function of the board and CEO
The literature presents a dominant theory explaining the functions of the board; the agency theory. According to the agency theory, interests of the manager are not in line with shareholder interests as a result of separation of control and ownership (Fama and Jensen, 1983).
Contrastingly, according to Eisenhardt (1989) the agency theory comprehends the board of
directors as a mechanism of internal control that focuses on safeguarding shareholder interest
from managerial opportunism. While most researchers tend to group the CEO, boards and top management teams as a whole, this study will thus examine the Chief Executive Officer (CEO) separately. A reason for this separation is provided by Carpenter and Wade (2002), who state that both of these parties possess differences in power and status.
2.3 International firm diversification
International diversification in firms is important for their growth strategy; especially for firms where the home market offers limited growth potential (Hsu, Chen and Cheng, 2013). As a consequence, there have been many studies on international diversification and firm performance (Kircka et al., 2011; Wiersema and Bowen, 2011; Yang and Driffield, 2012). As the success of companies is a result of mastering the risks of international diversification, firms should develop appropriate business strategies. Consequently, international diversification strategies of firms will have a risk reducing effect. Additionally, market diversification creates stability in the revenue streams, and firm performance is less dependent and vulnerable to country-specific shocks (Hitt et al.,1997). Moreover, several researchers argue that higher international diversification of firms leads to a minimization of cost as firms can shift activities to the lowest-cost location and search for the lowest labor and material cost (Click and Harrizon, 2000; Thomas and Eden, 2004; Denis, Denis and Yost 2002). To examine international diversification this study works with the firm’s geographical footprint as a proxy of measurement (Bruno and Song Shin, 2014).
3. Hypotheses
In the previous section the Upper Echelon theory explained the importance of former characteristics of the board and CEO. Besides, it indicated how the board and CEO function in an organization and why international diversification is important for the firm. The following paragraphs will elaborate on some of the most important top management characteristics (Aguilera and Jackson, 2003) and lead to the hypotheses.
3.1 International board diversity
In a recent article, Hambrick (2007) describes the importance of the international characteristics
of top management. Similarly, using international indicators, researchers found significant
evidence that these profiles are highly related to strategy and performance outcomes (Boeker,
1997; D’Aveni, 1990). Moreover, they tested that these psychological and social processes of internationalization lead towards international orientated behavior. Consequently, these positive results emphasize a firms’ potential when internationalizing their top teams in the international diversification strategy (Click and Harrizon, 2000, Thomas and Eden, 2004;
Denis, Denis and Yost 2002). To illustrate the advantages of a firm’s internationalization, Chin (2013) explains how different international experiences and backgrounds can maximize success and create opportunities in the international business strategies. This underlines the importance of selecting the right people for these tasks (Chin, 2013). From a resource- and knowledge- based perspective, foreigners in a firm’s board can be a source of valuable knowledge for new geographical market entry. Consequently, they increase the quality of decision making and provide for further development within this scope (Zahra and Filatoctchev, 2004). Also, Gomez-Meijia and Palich (1997) and De Jong and Van Houten (2014) declare that performance studies on MNE globalization should not only focus on the extent of international diversification. Instead, they should take the impact of cultural diversity within top management into account as this improves international orientated behavior. Consequently, the impact of international diverse boards leads to improved and increased international diversification of the firm, which results in the following hypothesis:
H1a: The higher the share of non-nationals on the board, the higher international diversification
H1b: The higher number of different nationalities on the board, the higher international diversification
3.2 Board size
As explained in the Upper Echelon theory, top management backgrounds are of importance for the firm strategy (Hendersons et al, 2006). Taking this into account, the relation between board composition and international diversification of the firm can be measured. To illustrate, larger boards benefit from being composed out of a higher number different inputs of human capital.
This will positively influence and affect the international diversification and could possibly
create synergies (Sanders and Carpenter, 1998; Carpenter and Fredrickson, 2001). In addition,
according to Simons et al. (1999) large teams have greater opportunity and more confidence in
adopting riskier strategies in the international business strategy. Similarly, Hambrick and
D’Aveni (1992) argue that teams with more human resources are more likely to achieve
significant international diversification levels as they have a larger variety of human resources and are therefore more capable to deal with foreign and turbulent markets. Finally, Levy (2005) initiates that larger boards have a better fit with uncertain situations and uncontrollable factors that appear in the international market setting.
However, as in any collaboration within a large group, there is a point in which a turn in direction can be expected (Smith et al., 1994). A board containing too many members could result in interpersonal distrust, problems in communication and potential conflicts among the members, which can negatively influence the international diversification. It is thus stated that the board size has an Inverted-U shape relation to the international diversification process of the firm.
H2: Board size has an Inverted-U shape relation to international diversification
3.3 CEO tenure
As boards function to guide the organization, the CEO holds the management duties. Therefore, Finkelstein and Hambrick (1996) state that executives will perceive strategic choices selectively based on their experiential history and background. Tenure is the is the positioned time of an executive at a function in top management (Hsu, et al., 2013). Along the same line of reasoning, longer tenure results in more organizational familiarity, and well-established personal relations with board members, managers and other executives. In addition, market knowledge, knowledge on institutions, processes and cultures could be central issues in helping to understand the unknown and new environments (Westhead, et al., 2001). Consequently, these experiences are beneficial for the international diversification process of the firm. For example, because lengthier tenure of executives could create social cohesion and jointly shared cognitive structures, this results in potential higher international diversification. Besides, organizational familiarity enhances CEO’s interpersonal relations and greater familiarity with the firm; its internal and external environment can be critical for the firm its international diversification process (Tihanyi, et al., 2000). Additionally, the time dimensions CEO tenure, should effect the international diversification positively (Hsu, et al., 2013). Hence, this results in the following hypothesis:
H3: Higher CEO tenure leads to higher international diversification.
3.4 CEO duality
CEO duality appears when both the CEO position and the chairman of the board of directors are covered by the same person (Bordean and Borza, 2009). According to the agency theory, separation of roles within the top management of the company is in favor of the quality of decision making, concerning the international business strategy. According to Ruigrok et al.
(2006), there appears to be a negative association amongst CEO duality and the strategic orientation of the firm. They argue that CEO duality could lead to conflicts of interest as a result of a claim on two positions by the CEO. Consequently, this results in limited international orientated behavior and a more passive attitude towards foreign expansion. As a result, separation of CEO and the position of board chairman is argued to be more successful and influential for international diversification (Adams et al., 2005). Consequently, Adams et al., (2005) conclude a negative effect of CEO duality on the firm’s strategies. Therefore, the hypothesis is as follows:
H4: There is a negative effect of CEO duality on international diversification.
4. Conceptual model
Regarding hypothesis one A and one B this research claims international board diversity has a positive effect on international diversification. This in the form of the share of non-nationals on the board and the number of different nationalities on the board. Hypothesis two, board size has first a positive effect but negatively beyond a certain point, thus hypothesizing an inverted U- shape relation exists between board size and international firm diversification. Accordingly, this study claims that CEO tenure has a positive effect on international firm diversification.
Regarding hypothesis four, CEO duality has a negative effect on international firm diversification.
Figure 1: Conceptual Model
5. Methods
The previous section discussed the introduction of the theory and identified the questions concerning the Upper Echelon theory, the importance of the CEO and board and a description of the facets of international diversification. Hereafter, the hypotheses highlighted the expectations, including the conceptual model that explained the relation between the five independent variables and the dependent variable. This section examines the development and understandings of the dataset, introduction of the variables and the multiple regression assumptions. Finally, the empirical results obtained throughout this research will answer the hypotheses. These will lead into the discussion of the conclusion and the implications.
This research’s primary intention is to measure and examine the direct effects of the independent variables on the dependent variable and the related insights concerning these variables. As this connection will demonstrate which board and CEO characteristics are important or should be combined to positively influence international firm diversification. In this research, the direct effects of the international diversification will be measured using the geographical footprint. The aim was therefore to use several control variables to check for appearances in certain industries or regions within Europe. Moreover, the control variable is a constant variable in the experiment and will test the relative impact of the independent variables as this variable clearly identifies the existing relation. The use of control variables makes it possible to categorize the dataset based on industry and region which could uncover significant differences between the sample items. Unfortunately, since the sample is rather small, categorizing the data creates sub-samples which are too small to include and examine them in this analysis. These data will thus not be included in this research, but are suggested for future research. In the following part, the dataset and timeframe will be explained. Subsequently, the variables and their relation to the data measures will be explained, whereupon the data measure will assess the dataset.
5.1 Dataset
This paper will examine how certain demographic and corporate governance structural aspects
and characteristics of top management relate to international diversification in international
business strategy. These firms should therefore be identified as our population of interests and
their boards and CEOs international experience should be examined based on the formulated
hypotheses. Consequently, it was decided to select these firms as their data on boards and
CEOs, international experience and the firms’ geographical footprint were better accessible.
This study requires a number of publicly available variables which are more transparent to test.
The European top 100 firms would therefore be appropriate sample for this. While focusing on these firms, a large variety of industries will be covered; all in terms of international diversification levels. The firm’s industry characteristics and region within Europe matter in studying the data set, since corporate governance regulations and backgrounds based on the Upper Echelon theory differ across the continent. Unfortunately, it will not be possible to test these differences as a data set of 100 samples is too small. Subsequently, corporate governance literature has been used to give an explanation in the result and conclusion section.
Respectively, this research makes use of a primary data set collected through available secondary data on the top management characteristics of the top 100 European firms. As the sample focused on the top 100 European firms, a sample of 101 firms was created based on the Fortune top 100 list of European firms based on their revenue figures. The data have been extracted from the Bureau van Dijk (BVD) and the ORBIS databases, which are both available for students at the University of Groningen. Moreover, these databases provide accurate information from international and domestic company databases and extensive information on corporations. If this, for instance, would not cover all the needed information, the availability of corporate websites and annual reports made it possible to complete the dataset. Therefore, these sources will cover the total need for information. The process of data sourcing made it possible to generate an individual primary dataset. A significant factor of a primary dataset is possibility to get the right and accurate data which is needed and creates the possibility to measure the intended variables. Consequently, it is possible to generate a more precise answer on the research question. Furthermore, a primary dataset would not create any doubts on the collected data and their additional insights. Another reason for this method is these firm’s obligation to share their company profile and financial information. Despite all these positive aspects, it is important to note that a primary dataset could also bring about several disadvantages. For instance, decision making on what should be measured, ensuring quality and the costs and time it takes to develop a primary dataset. As explained before, the focus was on the top 100 European firms which led to a collection process without any constraints.
5.2 Timeframe
As visualized in the conceptual model, the ways in which the independent variables of the
international board experiences, CEO duality board size and CEO tenure have their influence
on international diversification is clearly stated. Consequently, the firms that function as a sample in this research are considered to have similarities in background, history, and processing strategies. Therefore, this research will take a time dimension into account to measure, the international diversification. Therefore, the measure will compare and examine a time frame of five years; from 2009 until 2014, as 2014 is the most recent year available. Bao et al. (2014) identified that the average board tenure was 10.01 and top management tenure was 9.62 of the Fortune 1000 firms with a revenue of at least $1.5 billion. As these firms are comparable to the top 100 European firms, the same time dimension will be used. To create a timeframe within this research a ten-year average is used. However, as the policy implication will be more rapid than the average tenure, this study focuses on top management characteristics of 2009 and how these affect the international diversification results in 2014. It will thus take a five-time period in which the CEO or a board member will have influence in management.
Therefore, 50 percent of the average top management tenure will function as implementation time of firm policies.
5.3 Variables
After introducing the theory and the hypotheses, this section will explain the collected variables to complete the dataset. Besides, the data will be identified and a description of the data measuring will be covered.
Independent measure: top management characteristics
The variables used for predicting the dependent variable are: international board experiences, number of different nationalities, board size (number of members) and CEO tenure. These predicting variables will be measured to master the theory covering the board and CEO characteristics. It is therefore essential to use a ratio, in this case interval data, and dichotomous categorical data. Therefore, international board experiences and board size are measured as interval data. CEO tenure is measured in years and is interval data. CEO duality is measured as a dichotomous categorical data that takes the value of (1) if the CEO and chairman are the same person and (0) if this is not the case.
Dependent measure: international diversification
A direct measure of international diversification was used to measure the dependent variable.
This method particularly measures the geographical footprint of the firm. Consequently, the
international diversification focuses on risk taking as a direct effect of the firm’s geographical
footprint, international scope and dispersion that appears with the diversification strategy (Sanders and Carpenter, 1998; Hitt et al. 1997). More specifically, the measure used in this research is the firm’s number of host countries in exact and continuous numbers, consequently categorized as interval data.
6. Research measure
Using a multiple regression analysis creates the possibility to do a causal analysis and forecast and predict the effects and trends as they appear in the dataset. Consequently, the causal relation appears when studying one continuous dependent variable, predicted or influenced by two or more independent variables. This causal relation determines if an independent variable really affects the dependent variable and explains the phenomenon that appears in the relation.
Accordingly, the dependent variable is measured by interval data and the independent variables are measured by interval and dichotomous categorical data. Multiple regression suggests that there should at least be three variables that contain at least 20 cases per independent variable.
The dependent variable measures international diversification by the geographical footprint, assessing the number of host countries in 2014 as a result of the top management characteristic from 2009. The multiple regression analysis will predict the Y value by the given X1, X2, X3, X4 and X5 values.
First, an Enter-method multiple regression analysis was accomplished in which all data were put equally in the regression. Due to insignificant findings, the Stepwise-method was used.
Despite the insignificant outcomes another method can be used to assess the dataset. This was done using the Stepwise-method: this software assesses all predictor variables and starts with the one that predicts the highest effect on the dependent variable. Repeating this process with the variable resulted in predicting the highest value on the dependent variable. This procedure continues until all independent variables were added. The Stepwise-method starts with the independent variable with the highest loading in the Enter-method and the independent variable with the least weight at the end of the method. Using this method, the independent variables are tested separately as a result of categorizing them according to their weight in the Enter-method.
Consequently, this results in a more precise measure where independent variables will predict the dependent variable. However, this method produces the alleged suppressor effects.
Moreover, the result of these effects occur when independent variables are only significant
when another predicting variable is held constant. Conversely there are two major deficiencies
that can appear in Stepwise regression. First, the selected model was created out of the many possibilities that could be considered to better include the data. Second, the evaluation by order could lead to undervaluing possible combinations of variables.
7. Evaluation of method assumptions
Assessing the multiple regression analysis can be done after checking and correcting the primary dataset on the top 100 European firms, along the lines of assumptions for multiple linear regression (Pallant, 2010; Field 2013). The examination of the multiple regression assumptions implies that the dataset can be examined and tested by the multiple regression analyses. More specific, these multiple regression assumptions are as follows: linearity, normality, multicollinearity, auto-correlation/independence of observations and homoscedasticity. In case it does not pass any of the assumptions the dataset will be statistically corrected. In the following section, this new dataset will be used for the multiple regression analysis.
7.1 Test for linearity
Linearity examines if the relation between the variables is linear, this can be shown by a scatterplot. When noticing a straight line rather than a curve it can be assumed that there is linearity. In order to check linearity a scatterplot was created. Considering this scatterplot, linearity can be assumed. Following this indication, using a scatterplot to check the distribution of the variables, the outliers were tested. In this test, 3 outliers were detected and deleted from the sample which currently contains 98 firms. Moreover, deleting the outliers was done along the lines of the outlier labeling rule. According to Tukey (1977), in the usage of the test, a more appropriate multiplier was used: 2.2 instead of 1.5. Hoaglin et al. (1986, 1987) even argue that examination along this measure can be seen as the most valid way to detect outliers in data sets (Hoaglin et al., 1986 and Hoaglin et al., 1987). Besides, Cook’s distance was completed to create a scatterplot and check for any outliers (appendix B). As a result of the aforementioned analyses, two more firms have been deleted from the sample which leads to a last total of 96 firms.
7.2 Test for normality
In addition, a normality test is needed to estimate and evaluate whether there is normal
distribution. By using visual assessment of normality, it will be possible to check the histogram
and the fitted normal curve, subsequently a normal P-P plot and lastly a test for normality was performed (Appendix D). When checking and testing the dataset using the Kolmogrorov- Smirnov test and Shapiro-Wilk test, it could be concluded that there were no non-normal results. Both methods tested the variables and identified the dataset as individual variables. To illustrate, the Shapiro-Wilk test was used to test small samples, below 2000, which is the case in this study (Shapiro and Wilk, 1965). As a result of high p-values, the alternative hypothesis can be rejected. It could therefore be concluded that the data is normally distributed, which means that there is no need to transform the dataset.
7.3 Test for multicollinearity
A multicollinearity test highlights the conditions to ensure and confirm that the function of the independent variable is a predicting variable of the dependent variable. Therefore, little or no multicollinearity in the data set is the third assumption of the multiple regression analysis. This will occur when the independent variables are not independent from one another. Thus an important related assumption is that the error mean is uncorrelated. When testing for multicollinearity a Pearson correlation matrix can be checked (appendix E). Furthermore, another test for multicollinearity should be used, this is the variance inflation factor (VIF).
Consequently, the VIF-factor and the tolerance score can be used to check for multicollinearity.
According to the literature, there are several acceptable levels concerning the VIF levels. Even though it is mostly stated that the maximum value should not be higher than or exceed 10 (Kennedy, 1992 and Neter et al., 1989), it is sometimes argued that a maximum value of 5 (Rogerson, 2001) or 4 (Pan and Jackson, 2008) is more specific.
As VIF levels around 3.5 assume ‘something’, a score of 5 assumes ‘little’ and a score of 10
assumes a definite case of multicollinearity (Neter et al., 1989), this study assumes that a VIF
level below 3.5 and preferable below 2 account as sufficient and indicate that multicollinearity
is not a problem. Even though a greater chance of high multicollinearity is a valid result of
using multiple independent variables, the samples in this study did show scores above 3.5 which
should be corrected. This was necessary for tolerance factors below 0.3 and the related outcome
of the correlation matrix gave an indication of multicollinearity. Therefore, the tolerance factor
of one variable was below 0.3 and was mean centered. Besides, the same variable demonstrated
a high score in the correlation matrix appeared. Consequently, mean centering this variable
(number of different nationalities) reduced the VIF-factor and created a better correlation and
VIF score (Appendix G). Therefore, the dataset met the expectation of multicollinearity.
7.4 Test for autocorrelation – independence of observations
To check autocorrelation, a scatterplot could be helpful while using the Durbin Watson test.
The Durbin Watson test is an indication for independence of observations among the dataset.
The dataset needs little or no autocorrelation for multiple regression analyses. To illustrate, scores around 2 indicate no autocorrelation and, as a rule of thumb, values between 1.5 and 2.5 indicate no auto-correlation. This data set came out to be 1.867 and it can be concluded that there is little to no autocorrelation (Table 2 and Appendix F).
7.5 Test for homoscedasticity
Homoscedasticity illustrates that error terms among the regression are equal or constant. When this is not the case, there is an indication of heteroscedasticity. This results in biased estimates of the standard errors and functions as a source of bias in the test statistics (Hill et al., 2009).
While a scatterplot can be used to check for homoscedasticity, the Breaush-Pagan and Koenker test examines the dataset for homoscedasticity. Therefore, to test for homoscedasticity, both the Breusch-Pagan test and a Koenker test will be used. The analysis of the plot appears to be without any small disturbances. However, as this is a small sample without examination of the residual histogram, it cannot be judged whether this is statistical significant or not. Likewise, the P-P plot has data points close to the least square fit line, with only small deviations appearing. However, checking the plot with the standardized residuals against the standardized predictive value, there are many homoscedastic data points while only a few are heteroscedastic. Therefore, using the Breusch-Pagan test, the outcomes indicates, a BP of 4,751 and a significance of 0.447 which is higher than 0.05. This implies that the H0 for homoscedasticity cannot be rejected. A similar situation occurs with regards to the Koenker test which is more appropriate for smaller samples. In this case, the K is 4.334 with a significance of 0.502. This implies that it is not possible to reject H0 for homoscedasticity (Koenker, 1981).
The result matrix can be found in appendix C.
Ultimately, all assumptions of the multiple regression analysis are satisfied and all disturbing
data points have been corrected by statistical procedures using SPSS.
8. Empirical results
The multiple regression analysis will be able to explain the relation between the several independent variables and the continuous dependent variable. At the center of the multiple regression analysis, plotting a single line through the scatter plot is the main focus and developing the regression formula. This can be done by first checking for correlation and directionality within the dataset. Secondly, the model can be estimated and, lastly, the validity and efficiency of the tested model can be evaluated. The application of the statistical methods was performed using the ideas and methods of Field (2013) and Pallant (2010).
8.1 Descriptive statistics
Table 1 provides an overview of some of the relevant descriptive statistics of the data. It should be taken into account that the new DV has been manipulated along the lines of the multiple regression assumptions, which results in a different weight of the real, uncorrected original mean of the dependent variable international expansion. The non-manipulated mean of the dependent variable is 56.9 countries to which a firm has expanded to, with a minimum of 4 countries and a maximum of 220 countries. The share of non-national has a minimum of 0 and a maximum of 14 with an average of 4.22 non-nationals in the board of directors. The number of different nationalities has an average of 4,7968. The board size has a minimum of 2 and a maximum of 28 with an average of 12,07. CEO tenure varies from 2 to 28 years with an average of 4,93 years. And at last, CEO duality emerges more often than it does not.
Descriptive Statistics
Minimum Maximum Mean Standard Deviation
Dependent variable International diversification
0,48 2,34 1,6210 0,36627
Independent variables
Share of non- nationals
0,00 14 4,2292 3,60549
Number of different nationalities
0,00 36,84 4,7968 5,90439
Board size 2 28 12,07 4,293
CEO tenure 0 25 4,93 4,016
CEO duality 1 2 1,42 0,496
Table 1: Descriptive Statistics
8.2 Regression results
To start with, the Enter method was used to test for multiple regression. In this model, all the independent variables are being arranged into one bundle and are tested against the dependent variable. As a result of an insignificant outcome of the model, the Stepwise model was used to test the effects of the variables independently. Applying this test, the highest predictor variables were plotted to the independent variable at first and lastly to the predictor variable which was, according to the Beta values as which appeared in the coefficient model, the least significant.
Thus, the main flaws of stepwise regression do not influence the results of this study and will help to find the possible predictor. There are several scores linked to terminologies that make the multiple regression understandable. Firstly, a description of the used terminologies will be presented. Using the beta value (B-value), it is possible to measure how effective a predictor variable can influence the dependent variable. The R functions as a measure and reflector between the observed value and the predicted value of the data set and their relation to the dependent variable. Whereas R Square (R2) describes the square measure of association that describes the percentile of overlap between the independent and dependent variable. The dependent variable is a statistical measure describing how close the data is plotted to the regression line.
Model Summary b
Change Statistics
Model R R
Square
Adjusted R Square
Std.
Error of the Estimate
R Square Change
F Change
df1 df2 Sig F Change
Durbin- Watson
1 0,322a 0,103 0,054 0,35632 0,103 2,076 5 90 ,076 1,867
a. Predictors: (Constant), share of non-nationals, number of different nationalities, board size, CEO tenure, and CEO duality b. Dependent Variable: International Diversification
Table 2: Model Summary Enter-Method
The R Square in this model is able to predict 0.103 (10.3%) of the variability of the effective
data around the mean. This appear to be a low number, though the more complicated a model
or data set appears to be, the better it will be able to explain any disturbance and fluctuations of
the outcome. The adjusted R2 value will adjust according to the number of predictors in a
dataset. With an adjusted R Square of 0.054 (5.4%) there is, for example, a very low
contribution of the independent variables on the dependent variable. As the R Square value
interpretation depends on the distribution of other values and predictors, it will not be possible
to conclude whether this is ‘right’ or ‘wrong’. The significance value was therefore checked,
which appeared to be larger than the value on which the significance of the model will be concluded: 0.076. Based on the model summary and the ANOVA table of the Enter-model (Appendix F) it can be concluded that the model is not significant. Consequently, the Stepwise model will be used to further test the data set. As a result of the non-significant outcome of the Enter-model, the Stepwise-model will be used to test the data set. This will demonstrate how the independent variables relate to the dependent variable, both separately and in smaller groups. This is in contrast with the Enter-model, where both variables are placed in one large group. Using the Standardized Coefficient Beta enables the creation of a ranking as input for the Stepwise-model. Firstly, international board characteristics (0.244), secondly number of different nationalities (0.151), thirdly CEO duality (0.147), fourthly board size (0.013) and lastly CEO tenure (-0.060).
The correlation model (Appendix E) demonstrates several correlated factors. The highest correlation is 0.348 and emerges between international board characteristics and board size.
Several other factors show some forms of correlation and others factors even demonstrate negative correlation. In the Stepwise-model ANOVA table (Appendix G) it can be found that four out of the five created independent variable groups are significant. Evidently, the model appears to be insignificant exclusively when adding CEO tenure to the complete group of all independent variables.
Model Summary b
Change Statistics
Model R R
Square
Adjusted R Square
Std.
Error of the Estimate
R Square Change
F Change
df1 df2 Sig. F Change
Durbin- Watson
1 ,250a ,063 ,053 ,35648 ,063 6,293 1 94 ,014
2 ,284b ,081 ,061 ,35496 ,018 1,807 1 93 ,182
3 ,316c ,100 ,071 ,35310 ,019 1,982 1 92 ,163
4 ,316c ,100 ,060 ,35503 ,000 ,000 1 91 ,982
5 ,322d ,103 ,054 ,35632 ,033 ,341 1 90 ,561 1,867
a. Dependent Variable: international firm diversification b. Predictors: (Constant), share of non-nationals
c. Predictors: (Constant), share of non-nationals, and number of different nationalities
d. Predictors: (Constant), share of non-nationals, number of different nationalities, and CEO duality
e. Predictors: (Constant), share of non-nationals, number of different nationalities, CEO duality, and board size
f. Predictors: (Constant), share of non-nationals, number of different nationalities, board size, CEO duality, board size, and CEO tenure
Table 3: Model Summary Stepwise-Method