Project topics and materials
Get Complete Project Material File(s) Now! »
METHODOLOGY
The objective of this chapter is to state the methodology used in this thesis. There are six sections in this chapter: In the first section research philosophy and approach is addressed, followed by research strategy and method in the second section. Later, the process of sample selection and data collection is outlined in the third section, followed by the variables and measures in the forth section, and control variables in the fifth section. The final section is to discuss reliability, validity, and common method bias.
Research Philosophy and Approach
Before arriving at a research approach, it is important to reflect on research philosophy “which we define as the basic belief system or world view that guides the investigation, not only in choices of method but in ontologically and epistemologically fundamental ways” (Guba and Lincoln, 1994). Ontology, epistemology and axiology are three major schools of research philosophies (Saunders, Lewis and Thornhill, 2009): ontology concerns the nature of reality, epistemology concerns the nature of knowledge and what constitutes acceptable knowledge in a specific field, and axiology concerns the judgments about the role of values in research. The research philosophy we adopt underpins our research paradigm and thus affects the way we design the research process.
In social science research, the term of paradigm is often used to define the way of “examining social phenomena from which particular understandings of these phenomena can be gained and explanations attempted” (Saunders et al., 2009). There are four main research paradigms: positivism, realism, interpretivism and pragmatism.
Both positivism and realism adopt a philosophic stance on natural science, essentially applying a natural science approach to social science. Ontologically, these two paradigms share the view that the reality is external, objective and independent; epistemologically, both paradigms rely on observable phenomena providing credible data; and axiologically, they advocate the separation of the researcher from what is being researched by taking an objective stance (Saunders et al., 2009; Wahyuni, 2012). The distinction between positivism and realism is that positivist researchers seek to obtain a law-like generalization by conducting value-free research and they believe in the existence of universal generalizability, which can be applied across contexts. Realist researchers on the other hand challenge the belief of universal truth, especially critical realist researchers who consider knowledge as a result of social conditioning and thus they focus on explaining phenomena within a context or contexts. In terms of methodological choice, it is common for both positivists and realists to adopt a quantitative approach in their research.
Interpretivism emphasizes the difference between humans in their roles as social actors. In social science research, interpretivists reject objectivism and understand the social world from the subjective meanings that people attach to it. Interpretivists favor qualitative data that provide the details of social constructs, and prefer a narrative form of analysis to describe a particular social reality studied. Axiologically, interpretivist researchers consider they are part of what is being researched, thus the experiences and values of both participants and researchers influence data collection and data analysis (Wahyuni, 2012; Saunders et al., 2009).
Pragmatism is one branch of research paradigm, which argues that the research question is the most important determinant of the research philosophy. Pragmatists believe that objectivism and subjectivism are not mutually exclusive, thus it is possible to work with both positivist and interpretivist positions. In terms of methodological choice, pragmatists favor mixed or multiple method design, which integrates both quantitative and qualitative research (Saunders et al., 2009).
In this study, the paradigm of pragmatism prevails, which advocates the importance of research objective rather than a specific research philosophy to be adopted. It is intended to examine the degree of SEW and the effect of SEW on the internationalization of family-owned MNCs. Moreover, it is expected to offer some generalizable research results, which could be applied to other family-owned MNCs. To achieve the objectives, a deductive approach is adopted in which theoretical hypotheses are developed from the theories of international business and family business, and a quantitative research method is designed to test the hypotheses. In this study, the research data are collected and analyzed quantitatively, which owes more to a positivist position; however, a critical realist position is taken into account because the theoretical hypotheses are developed and tested within the context of European family-owned Hidden Champions; moreover, an interpretivist position is also involved because some of data collected in this study are qualitative in nature and the experience of the author also influences data collection and analysis.
The deductive approach adopted in this study follows Robson (2002)’s five sequential steps as listed below:
(1) developing hypotheses (testable propositions about the relationship between SEW’s five dimensions and internationalization performance) from the theories of SEW and internationalization;
(2) expressing the hypotheses in operational terms (that is, indicating exactly how the variables are to be measured), which propose the relationship between SEW, international strategies, and internationalization performance;
(3) testing the operational hypotheses;
(4) examining the specific outcome of the inquiry (it will either tend to confirm the new theory or indicate the need for its modification);
(5) if necessary, modifying the new theory in the light of the findings.
Research Strategy
The nature of research can be either exploratory, descriptive, explanatory, or a combination of all. Exploratory research is a means of finding out what is happening and further seeking new insights from the findings; descriptive research is a means of portraying an accurate profile of what is being studied; and explanatory research is a means of establishing causal relationships between variables, or explaining reasons (Saunders et al., 2009). In this study, a combination of descriptive and explanatory research is employed, which is also defined as a descripto-explanatory study. Descriptive research is undertaken firstly to have psychometric measurement on SEW as a distinction of family firms; and then explanatory research is undertaken to explain the effect of SEW on the internationalization of family firms.
There are five major research strategies in social science research: experiment, survey, case study, action research, grounded theory, ethnography, and archival research. Experiment and survey strategies are principally associated with quantitative study. Compared to experiment strategy, survey strategy is strong in realism, practical significance and normative quality (Slater and Atuahene-Gima, 2004). With the survey strategy, the primary data are collected specifically to address the research question; and when sampling is used, it is possible to generate findings that are representing the whole population (Saunders et al., 2009). In terms of data collection techniques, questionnaire, structured observation and structured interviews all fall into survey strategy. Questionnaires tend to be used for descriptive or explanatory research, and they enable researchers to “identify and describe the variability in different phenomena” and to “examine and explain relationships between variables, in particular cause-and-effect relationships” (Saunders et al., 2009). In this study, besides secondary data research, the online survey strategy is employed and an online questionnaire is used for primary data collection, and a statistical analysis is conducted to examine the construction, reliability and validity of the five-dimension scale of SEW, as well as the relationships of SEW and its five dimensions in the internationalization of family-owned MNCs.
Online surveys have the advantages of producing more accurate data, faster data collection, and at lower cost, thus there is an increasing trend to online surveys. There are three forms of online surveys: email surveys, HTML form surveys, and downloadable interactive surveys (Slater and Atuahene-Gima, 2004). Email surveys are conducted by sending respondents email questionnaires; in HTML form surveys, respondents are typically invited by emails to participate in the online survey and a web link is provided from which respondents can assess and complete the questionnaires. In terms of the invitation email, Slater and Atuahene-Gima (2004) list some effective components as shown below:
– Personalized email contact;
– A subject that indicates the topic;
– Where the email addresses were found;
– Who is conducting this survey;
– What the survey is used for;
– A brief introduction of the topic;
– The approximate time required to complete the questionnaire.
In this study, a mixed online survey is used to collect primary data. An email with the link to the online questionnaire is sent to intended respondents individually, and an attached copy of questionnaire is provided alongside. Therefore, respondents could choose to either fill out the online questionnaire or use the attached email questionnaire to participate in the survey. Out of the total returned questionnaires, 95% respondents have used the online questionnaire to participate in the survey. As for the invitation email, besides the items listed above, the intended respondents are also informed that the data collected would be handled strictly confidentially and anonymously, and that a summary report would be provided to those who participate in the survey upon completion of the study.
A cross-sectional study is a study that refers to a study which takes a “snapshot” of a particular phenomenon (or phenomena) at a particular time (Saunders et al., 2009). It is common for cross-sectional studies to employ a survey strategy (Robson, 2002), and the data collected can be used to explain how pre-defined factors are related to one another in different organizations. For this thesis, a cross-sectional study is conducted in order to examine the varying degrees of SEW in different family firms and investigate whether there is a common relationship between SEW and the internationalization of family firms.
Sample Selection and Data Collection
Sampling plays an important role in quantitative research. A well-designed sampling plan allows a generalization of relationships found in the samples to the entire population.
Sampling can be categorized into two types: probability and non-probability (Saunders et al., 2009). In probability sampling, the samples are selected from the population on statistical grounds; and with non-probability sampling, the probability of each sample being selected from the total population is not known. Saunders et al. (2009) discuss different probability sampling techniques and non-probability sampling techniques separately, and in the meantime they also point out that for some research projects it is necessary to use both probability and non-probability sampling techniques. In this study, both probability and non-probability sampling techniques are considered when the author develops the list of the companies to be surveyed.
The data presented by Simon (2013) show a total number of 1,977 Hidden Champions in Europe, approximately 70% of them being family-owned (Simon, 2009), thus the total number of European family-owned Hidden Champions is estimated at around 1,380 whereas the total number of European family-owned Hidden Champions which have foreign subsidiaries must be less than 1,380. Because there is no existing database of European family-owned Hidden Champions, in this study the sampling frame is generated mainly via secondary research on the company lists provided in Hidden Champions of the 21st Century (Simon, 2009 and 2013). In order to enhance the population, more European family-owned Hidden Champions with foreign subsidiaries are added after screening the data of Hidden Champions provided by Biesalski & Company (2013) and 21st-Austria (2013). Resources for secondary data research in this study include books, business journals, and Internet resources such as company websites, an online company database (Amadeus) and other relevant websites accessed via Google search. In the secondary research, the criteria of (1) Hidden Champion, (2) European firm, (3) family firm, and (4) MNC are used to develop the sampling frame for the survey. The final sampling frame includes a total of 317 European family-owned Hidden Champions, which have subsidiaries in foreign countries.
With the secondary research, a database of family-owned Hidden Champions for survey is built up, which includes information regarding company name, company website, country of origin, ownership, name and email address of CEO (the term CEO used in this study refers to the head of top management, e.g. CEO, managing partner or chairman of management board). Other data are also collected if they were available, including the founding year of the firm, the number of family generation(s) involved, CEO duality or not, the number of foreign subsidiaries, industry engaged in, and revenue in 2012.
The online survey for data collection was conducted between April and middle of July, 2014, and emails were sent to the whole population of 317 companies. The CEO of each surveyed company was chosen as the intended respondent, because he or she has a good knowledge of the company’s international activities and performance and of the involvement of the owning family in the business. Only the input from the CEOs of these companies could ensure the quality of data collection. In this study, four methods are administered to enhance the response rate: first, the number of questions for the survey is limited to 25; second, the questionnaire starts with a brief introduction of this study and the rules applied in this survey; third, it is promised to provide a summary report of the study to the respondents who participate in this survey; fourth, a personalized email is sent to the CEO of each targeted company for the survey invitation. Compared to using an email invitation system of survey softwares (e.g. Qualtrics, Google Form), it is a time-consuming approach to send emails to each intended respondent personally for a survey invitation and the later follow-up. However it is a good way to enhance the response rate (Saunders et al., 2009), particularly in studies whose population is relatively small and the response rate is usually low, according to the literature.
In this study, a three-wave email is carried out for primary data collection, including an initial survey email and two reminder emails sent to those who have not responded to earier inquiries. Among the total 317 firms surveyed, 34 intended respondents could not be reached due to wrong email addresses, and responses have been obtained from 103 firms. Of the total 103 responses, 8 firms are not in the scope of the study due to being non-family firms, not having international subsidiaries, or having revenue higher than 3 billion euros in 2012. 21 firms have declined to participate in the survey due to various reasons, for example firm rule for no participation in external surveys, confidentiality of internal firm data, or lack of time. The final sample is consisted of 74 European family-owned Hidden Champions, which had invested in foreign countries, representing a response rate of 27%, as calculated according to the formula given below. This response rate is higher than the response rates of similar studies which are usually around 10-15% due to the collection of data that respondents may consider to be intrusive and highly confidential (Simon, 2009; Zellweger et al., 2012).
Of the total 74 valid questionnaires, 64 respondents have filled out their email address information for requiring a copy of the summary report of this study, indicating that the research topic is relevant and interesting to the respondents. For the 64 responses which provided email information, further secondary research is conducted to collect firm data which includes country of origin, founding year of the firm, family generation(s) involved, CEO duality or not, the number of foreign subsidiaries, industry engaged in, and revenue in 2012.
As statistical data of European family-owned Hidden Champions with foreign subsidiaries is not available, the statistical data of Hidden Champions in the German language area (Simon, 2009) is used to analyze the representativeness of the survey respondents. A comparison of descriptive statistics between the family-owned Hidden Champions surveyed in this study and the Hidden Champions surveyed in Simon (2009) is provided in Table 3.1. The data show that there is a good match between the survey respondents in this study and the Hidden Champions surveyed in Simon (2009), in terms of the diversity of founding year, industry sector, and revenue. The data show that the average age of the family-owned Hidden Champions surveyed in this study is longer than that of the Hidden Champions surveyed in Simon (2009), which could be explained by the nature of family firms’ endurance. In terms of the significant difference in average revenue, it might be caused by the growth of Hidden Champions over these years because Simon’s survey data of Hidden Champion was published in 2009 (Simon, 2009). The comparison analysis, together with the high response rate in this study compared to other similar studies, supports that the findings generalized from the survey respondents in this study could represent all of the European family-owned Hidden Champions, which have made investments in foreign countries.
Variables and Measures
SEW
SEW is independent in this study. A 15-item measure constructed from five dimensions is used to test the degree of SEW. More detailed information about the 15 items is provided in the following parts on the five dimensions. This 15-item scale has Cronbach’s Alpha at 0.84. Degree of SEW is the mean value of the 15 items.
Family Control and Influence (FCI)
FCI is independent in this study. A 5-item measure is used to test FCI. Four items are taken
from Berrone et al. (2012), and one new item is added which is “In my family business, family members have intensive interaction with managers in foreign subsidiaries, including constant international trips”. The five-item scale demonstrates a strong degree of internal reliability (a = 0.83). FCI is the mean value of the five items.
Identification of Family Members with the Firm (IFMF)
IFMF is independent in this study. A four-item measure is used to test IFMF. Three items are taken from Berrone et al. (2012), and one new item is added as discussed above. Among the three items proposed by Berrone et al (2012), only one is proposed originally for IFMF. One item proposed for BST and one item proposed for EA are re-allocated to measure IFMF according to the results of an Explorative Factor Analysis. This four-item scale has Cronbach’s Alpha at 0.72. IFMF is the mean value of the four items.
Binding Social Ties (BST)
BST is independent in this study. A two-item measure is used to test BST. Three items have been designed to test BST. After the Explorative Factor Analysis, two items are left to measure BST. The item of “In my family business, non-family employees are treated as part of the family” is allocated to measure IFMF. Indeed this re-allocation is consistent with the study of Naldi et al (2013), which considers the building of trust relationships with employees to be motivated by family identification. In addition, the revised two-item measure increases Cronbach’s Alpha from 0.53 to 0.57. BST is mean value of the two items.
Emotional Attachment (EA)
EA is independent in this study. A two-item measure is used to test EA. Three items have been designed to test EA. After the Explorative Factor Analysis, only two items are left to measure EA. The item of “In my family business, strong emotional ties among family members help us maintain a positive self-concept” is allocated to measure IFMF. The revised two-item measure increases Cronbach’s Alpha from 0.65 to 0.75. EA is mean value of the two items.
Renewal of Family Bonds to the Firm through Dynastic Succession (RFB)
RFB is independent in this study. A two-item measure is used to test EA. The item of “Family owners are less likely to evaluate their investment on a short-term basis” is dropped after the reliability test and the Explorative Factor Analysis. The revised two-item measure increases Cronbach’s Alpha from 0.44 to 0.62. EA is mean value of the two items.
Degree of Internationalization
Degree of internationalization is a dependent variable in this study. Because the majority of the sample are private firms and there are no sufficient secondary data sources to gain objective measures of their degree of internationalization, the self-reported approach is used to measure the degree of internationalization. This study uses a three-item measure to access the degree of internationalization which is TASi. This three-item scale has Cronbach’s Alpha at 0.67. As discussed above, TASi is the mean value of ASi and SSi which are calculated according to the formulas below.
ASi = Assets Spread index = FATA x NFC
SSi = Sales Spread index = FSTS x NFC
NFC ranged from: 1=0; 2=1-15; 3=16-30, 4=31-45; 5=45 above
FATA ranged from: 1=25% less; 2=25%-50%; 3= 51%-75%; 4=75%
FSTS ranged from: 1=25% less; 2=25%-50%; 3= 51%-75%; 4=75%.
Control Variables
Three variables are controlled in this study, which are country of origin, firm age, and industrial sector. In this study, 59% of surveyed companies are German companies, and the remainder are other European companies, thus Country of Origin is controlled as a dummy variable. Firm Age is chosen because older firms could have increased cumulative experience that could facilitate internationalization, and in this study it is assessed in terms of number of years from establishment to 2013. Because some industrial structures may encourage internationalization more than others, the industrial sector is chosen to be one control variable. In this study, the three types of sectors are identified: industrial goods sector, consumer goods sector and services sector. They are dummy coded using the industrial goods sector as reference.
Firm size is not chosen to be a control variable in this study. In the relevant literature it is common to assess firm size on the basis of employment figures or assets. Because a cross-sectional method is used in this study and the benchmarks of employment or assets for firms are very different in different industries, it is not proper to choose employment or assets to measure firm size in this study. Revenue in 2012 was initially planned to serve as control variable on firm size, but was dropped during the analysis because this variable affects the presentation of SEW’s five dimensions and their effects on the degree of internationalization. The relationship between firm revenue and degree of internationalization is complex. On the one hand, a firm with higher revenue can engage more in overseas markets, thus firm revenue positively influences a firm’s degree of internationalization; on the other hand, a firm with higher degree of internationalization could have higher firm revenue, thus firm revenue is influenced by a firm’s degree of internationalization.
Moreover, ownership has not been chosen to be a controlled variable in this study, because owning families are the sole shareholder or the largest shareholder in almost all surveyed companies. It has been reported that internationalization is encouraged by the second and subsequent generations in family firms (e.g., Fernandez and Nieto, 2005). As over 90% of surveyed companies in this study are run by the second or subsequent generation, the generation variable is not controlled in data analysis.
Of the total 74 companies that have participated in the survey, company demographic details are provided in Table 3.2.
Reliability, Validity, and Common Method Bias
Reliability
Reliability refers to the extent to which the data collection techniques and analytic procedure will generate consistent findings (Saunders et al., 2009). It represents the repeatability of the measurement. Internal consistency is commonly used to estimate the reliability of the measurement by running a correlation among indicators designed to measure a concept, and Cronbach’s Alpha is the common correlation value used in computing to represent the estimate of reliability. A rule of thumb employed by many researchers is to accept Cronbach’s Alpha at 0.7 or more (Salter and Atuahene-Gima, 2004). In strategy studies, many researchers use scales with a Cronbach’s Alpha of less than 0.7 and sometimes even less than 0.6 in their estimated models (Fornell, Lorange and Rios, 1990; Birkingshaw, Morrison and Hulland, 1995; Hulland, 1999; Wijbenga et al., 2007). According to Carmines and Zeller (1979), reliability should be measured for any multi-item scale to assess the quality of the scale.
Validity
Validity refers to the extent to which the measurement measures what it is intended to measure, and the findings are really about what they appear to be about (Saunders et al., 2009). There are three most basic types of validity, which are content validity, criterion-related validity and construct validity (Carmines and Zeller, 1979). Construct validity focuses on the extent to which the performance of a particular measure is consistent with theoretical expectations. In contrast to content validity and criterion-related validity, construct validity can be assessed in social science research. If there is inconsistency between the theoretical prediction and the empirical findings, it is drawn that the measure lacks construct validity (Carmines and Zeller, 1979).
In addition, generalizability, sometimes referred to as external validity, is considered a major criterion for evaluating the quality of a study. Generalizability is concerned with the extent to which the research findings are generalizable and can be applied to other relevant contexts (Saunders et al., 2009). There are three types of generalizability: statistical generalization, analytic generalization and transferability (Polit and Beck, 2010). In this study, as discussed in the part of sample selection and data collection, the usually high response rate and the comparison analysis support the statistical generalization of the survey in this study. Moreover, a detailed explanation of sample selection and data collection and the questionnaire in Appendix II are in place to enhance the transferability of the study.
Common Method Bias
Common method bias has the potential to generate spurious results. Salter and Atuahene-Gima (2004) summarize the following steps, which may help to reduce common method bias:
“- avoid any implication that there is preferred response;
– make responses to all items of equal effort;
– pay close attention to details of item wording;
– use items that are less subject to bias;
– keep the questionnaire as short as possible, without impacting research objectives, to minimize respondent fatigue;
– provide clear instructions;
– randomize the ordering of scale items; and
– reverse code some items so that the same end of a Likert-type response format is not always the positive end.”
(Salter and Atuahene-Gima, 2004)
Among various common method biases, non-response bias is a concern in many studies. To evaluate the non-response bias in this study, following Chrisman, Chua, Pearson and Barnett (2012), a comparative analysis is conducted between the respondents to the first round of emails and the respondents to the third round of emails. The respondents to the third round of emails are non-respondents compared to respondents to the first round of emails, and are considered to be similar to those who have no response in the survey (e.g., Chrisman et al.,2012; Schilke and Goerzen, 2010; Kellermanns and Eddleston, 2007). ANAVO test is conducted to test Country of Origin, Firm Age, Industrial Sector, SEW and TASi, and the results have shown that there is no significant difference between data from the first round respondents and the third round respondents.
Missing data bias is a concern in empirical studies, especially the studies of family firms that rely on the collection of primary data that respondents would consider to be highly confidential (Zellweger et al, 2012). In this study, Little’s MCAR test under SPSS is conducted to examine whether the missing data are randomly distributed, and the results show the missing data are distributed at completely random way (Chi-Square = 207.512, DF = 246, Sig.
= 0.964). In the data analysis of this study, missing values are replaced by the mean of the item. A summary report of the missing data is provided in Table 3.3.
In this study, 10 respondents filled out the questionnaire anonymously, indicating these respondents considered some information provided confidential and would therefore like to provide the information anonymously. This concern about confidentiality is also confirmed by another phenomenon that some respondents did not answer the question on percentage of foreign assets to total assets, and told the author that it was confidential when they were further contacted about the information.
Table of Contents
1 Introduction
1.1 Background
1.2 Problem Discussion
1.3 Resea rch Purpose
1.4 Definitions
2 Theoretical Framework
2.1 Socioemotional Wealth
2.2 Internationalization
2.3 SEW and Internationalization
2.4 Hypothese s Development
3 Methodology
3.1 Research Philosophy and Research Approach
3.2 Research Strategy and Research Method
3.3 Sample Selection and Data Collection
3.4 Variables and M easure s
3.5 Control variables
3.6 Reliability, Validity and Common M eth od B ias 3 8
4 Analysis and Results
4.1 Operationalization of Five-dimension Scale of SEW
4.2 SEW and Internationalization
5 Discussion
5.1 Study Findings
5.2 Contribution
5.3 Limitations
5.4 Further Research
6 Conclusion
References
GET THE COMPLETE PROJECT
