The impact of societal and individual level factors
on emergence of new entrepreneurship:
A cross-national analysis
Dissertation/Thesis
Advanced International Business Management Dual Award
Jan Skrabka
S3191710, University of Groningen 160658949, Newcastle University
Supervisors: Dr. Rian Drogendijk Dr. Harsh Kumar Jha
Abstract: This research examines the effects of societal level variables and individual level variables on the emergence of new entrepreneurship. The research is based on quantitative analysis of the data describing the variables in 104 countries around the world between 2007 and 2016. The innovative research design aims to study and combine the effects of individual level variables, such as fear of failure and perceived business capabilities with societal level variables, such as freedom from corruption or financing for entrepreneurs. Contrary to previous studies, which have often analysed only small samples, the author analyses the sample of countries during 10 years. The author found out, based on this large dataset, that despite the results some of other research studies, some of the variables do not have a big impact on the emergence of new entrepreneurship. The regression analysis shows that only Freedom from corruption and Perceived Business Capabilities have a statistically significant positive effect on the emergence of new entrepreneurship. The effect of Perceive Business Capabilities on emergence of entrepreneurship is positively affected by Primary and secondary education entrepreneurial activities, whereas the other hypotheses are not supported.
Acknowledgements
I would like to thank my supervisors Dr. Rian Drogendijk and Dr. Harsh Kumar Jha for their support, patience and guidance.
I would also like to express my thanks to my parents, who have taught me to value education and supported me during my studies.
3.4.4 Control variable ... 30
3.4.5 Summary of variables ... 31
3.5 COUNTRIES USED IN THE RESEARCH ... 32
3.6 ETHICAL ISSUES & RISKS OF THE RESEARCH ... 33
1 Introduction
“All that is valuable in human society depends upon the opportunity for development accorded to the individual.”
Albert Einstein
Entrepreneurship is considered as an important driver of economic prosperity and growth of countries, higher social mobility as well as the higher employment (Wong, Ho, & Autio, 2005; Lumpkin & Dess, 1996; Acs et al., 2012). The support of entrepreneurship from the state should, therefore, lead to the improvement in the economic growth of countries. We should not also forget the impact of entrepreneurship on local communities and individuals, as entrepreneurship can be the only option how to earn money for some people, especially for those in developing countries with the lack of job positions (Amorós & Bosma, 2014).
There are massive differences between the levels of entrepreneurial activity in various countries (van Stel et al., 2007). For example, only 2,4 % of the population in the productive age in Italy have started their own business, whereas this rate in Estonia amounts 8.8 % (Amorós & Bosma, 2014).
Similarly, there are also massive differences in the view of entrepreneurship and popularity of entrepreneurship in various countries. For example, 82 % of people in Turkey prefer entrepreneurship to employment, whereas only 22 % of people in Sweden have such preferences (European Commission, 2012).
Because of the importance of entrepreneurship for various actors, such as policymakers, companies, consultants, advisors, as well as the whole states and the EU, as well as scholars and students, entrepreneurship becomes widely discussed and researched topic in academic literature (Bruton et al., 2008). However, the practical implications of the research in this area are not insignificant, as the awareness of the entrepreneurship drivers helps the states and policymakers to prepare better policies to support entrepreneurship, and therefore also indirectly the economic growth (Lumpkin & Dess, 1996). The literature also describes the positive effects of such growth on the life of people and increasing common good of the society (Mlčoch, 2006; Trojan, 2012). Despite that there are numerous authors interested in finding the determinants of entrepreneurship, such as Thurik et al., 2003, Díaz-Casero et al., 2012, Nyström, 2008, Bjørnskov a Foss, 2008, there is one common problem related to the methodology of the big part of recent academic works in this area. The authors of such studies often use only a limited datasets with less than a hundred observations. Moreover, the authors often analyse only cross-sectional data, which can lead to biased results, because the values of the variables in some countries could be unusually high or low for such year. Therefore, if the dataset contains the observations for more years (a time component), the analysis can eliminate the impact of the extreme values in some years and bring more reliable results. However, the missing time component in the datasets of former research studies in this area could be explained by the fact that the large datasets related to entrepreneurship have become accessible only in recent years (such as Global Entrepreneurship Monitor (GEM), 2017).
Therefore the econometric methods used in this research together with large panel data are able to provide more reliable results, which can be subsequently reflected in the right public policy recommendations that really support the entrepreneurship and economic growth.
However, the research conducted on the factors that influence new entrepreneurship does not cover all the factors that may influence the entrepreneurship, analyses only a small number of influencing factors or even considers only the factors of the same type (such as only financial factors). The impact of societal and individual level factors on the emergence of new business activities is often analysed for these two types of factors separately, which does not allow understanding the mutual effects (Scott, 2002).
Moreover, there is only a very limited literature about the interaction effects between the factors influencing the entrepreneurship.
In order to bring a complex approach and present a more reliable analysis that can answer the research question, this research combines the societal level explanations and individual level explanations by analysing the impact of both types of factors together, as suggested by Scott, 2002.
The research is based on a quantitative analysis of the data describing various factors that are supposed to have an impact on entrepreneurship. In order to prepare sound results that can be generalized not only for the analysed countries but also on the whole world, the study analyses 104 countries around the world. A time component for eliminating the impact of the extremes in some years is used, therefore the data for all countries are analysed for 10 years between 2007 and 2016.
1.1 Research question
Therefore, the research is designed to answer the following research question:
1.2 Structure of the thesis
The thesis is structured in the following way:
First, the outline of the overall research setting is presented, followed by setting the research question. An extensive literature review of academic papers and other literature allows me to present the theoretical background and explanations of the researched phenomenon. In order to justify my choices, hypotheses development is carried out separately for each hypothesis, followed by conceptual model of the study.
methodology together with justification of econometrics methods used in this research is followed by descriptive analysis and regression analysis.
2 Literature review
2.1 Entrepreneurship
2.1.1 Definitions and explanation of entrepreneurship
In order to analyse the effects of various factors on entrepreneurship, entrepreneurship has to be clearly defined first. There exist many different definitions, such as:
“Entrepreneurship is a context dependent social process through which individuals and teams create wealth by bringing together unique packages of resources to exploit marketplace opportunities.” (Ireland, Hitt, & Sirmon, 2003) or “Entrepreneurship is the mindset and process to create and develop economic activity by blending risk-taking, creativity and/or innovation with sound management, within a new or an existing organisation.” (Commission of the European Communities, 2003) or “Entrepreneurship is the act of innovation involving endowing existing resources with new wealth-producing capacity,” (Drucker, 1985).
There is also another definition of entrepreneurship, which tries to suggest also some of the possible options how could the entrepreneurship be practically organised: "Any attempt at new business or new venture creation, such as self-employment, a new business organization, or the expansion of an existing business, by an individual, a team of individuals, or an established business" (Global Entrepreneurship Monitor (GEM), 2017).
used in various definitions of entrepreneurship according to OECD literature overview of the definitions of entrepreneurship are shown in Figure 1.
This short analysis of the concepts, which are used as a basis for various definitions of entrepreneurship, therefore leads to the conclusion that the primary purpose of entrepreneurship is to create an added value (Filion, 2011). Despite that, the importance of the entrepreneurship is not clear from the definitions and conclusion, therefore needs to be discussed separately.
2.1.2 Importance of Entrepreneurship
The importance of entrepreneurship is not a new phenomenon studied only in recent years. In spite of that, we can find out that the famous Harvard professor J. A. Schumpeter described the positive impact of entrepreneurship on the economic
Source: Filion, 2011
development of countries and on innovations at the beginning of 20th century (Schumpeter, 1934). He has also developed two theories of innovations. First of them, so-called Mark I, he has concluded that the innovations and positive changes in technology are driven by entrepreneurs (referred as Unternehmergeist). The key according to Schumpeter lies in the fact that the entrepreneurs are "doing of new things or doing the things that are already being done in a new way" (Schumpeter, 1947). In his later theory, Mark II, he has proposed that the driving factor of innovations and development is rather the impact of huge corporations rather than small entrepreneurs. Later on, there were a number of scholars examining entrepreneurship and its importance from different perspectives.
The aspects of economic growth and higher social mobility, as well as lower unemployment stimulated by entrepreneurship, have been studied for example by Acs et al., 2012, Lumpkin & Dess, 1996 or Wong, Ho & Autio, 2005.
Especially in developing countries, one should carefully consider the impact of entrepreneurship not only on a country as a whole but also on local communities and individuals. The recent research in this field has confirmed that entrepreneurship plays a crucial role in providing the necessary income particularly for poorer people in certain low-income countries, as the other job opportunities are not always present (Amorós & Bosma, 2014).
Despite that the impact on businesses draws the primary attention of policymakers during the preparation of new regulation, there are also some other levels that shall be taken into account. These other non-business levels could be more philosophical in nature, such as a creation of common good. The positive effect of entrepreneurship on the creation of common good is examined and discussed by Mlčoch, 2006.
2.2 Factors explaining the entrepreneurship
which is in line with the research question, which tries to identify and analyse the factors influencing the emergence of new entrepreneurship.
In order to analyse the literature describing the influence of individual factors in detail, the discussion of literature about each factor in included in the following sub-chapters. For the purpose of this research, the factors influencing the entrepreneurship are divided into two groups, but I am aware that this division of factors can be done also based on different criteria, and therefore the groups could also be different.
The first group consists of the factors – societal level factors – which are supposed to have an impact on entrepreneurship, based on the literature discussed in the respective subchapters. The second one describes the individual level factors and their impact on entrepreneurship.
The division of the factors influencing the entrepreneurship on societal level and individual level factors is not a new concept, but it is rather widely discussed in the academic literature related to entrepreneurship (e.g. Stephan & Uhlaner, 2010; Sternberg & Litzenberger, 2004 or Thornton, 1999). The distinction between these two groups of factors is that the societal level factors drive the demand for entrepreneurs, whereas individual level factors drive the supply for entrepreneurs. Therefore are these two groups also referred to as demand side and supply side factors (Thornton, 1999; Stephan & Uhlaner, 2010).
2.3 Societal level factors and their impact on entrepreneurship
regulating the requirements for any business activity can be seen as determinants of the societal level of factors, which affect the entrepreneurship.
Broadman’s extensive qualitative and quantitative research has shown, that the low level of certain societal variables in some countries (for example the presence of corruption, lack of transparency of businesses) has a negative impact on entrepreneurship (Broadman, 2004). Many other scholars support such view in their research papers (Salimath & Cullen, 2010; Co, 2004; Tonoyan, 2011; McMullen, Bagby, & Palich, 2008).
2.3.1 Freedom from corruption
The corruption is defined as: “Corruption is the abuse of entrusted power for private gain. It can be classified as grand, petty and political, depending on the amounts of money lost and the sector where it occurs” (Transparency, 2017).
The effects of corruption on the entrepreneurship were not always clear. For example, the purpose of the research by Taslim in 1994 was to persuade some of the economic professionals in Bangladesh that corruption has a negative effect on entrepreneurship, and it is not just a reallocation of resources (Taslim, 1994).
The effects of corruption have also been studied by Avnimelec, his paper“proposes worldwide empirical evidence that countries with high levels of corruption usually face low levels of productive entrepreneurship.” (Avnimelech et al., 2014).
The corruption is also perceived as a strongly unethical and unlawful behavior of individuals or companies, or even the states (Trojan, 2012; Mlčoch, 2006).
The presence of corruption has also various impacts: trust in contractual relationships is destroyed by corruption, institutions are not reliable, and the rule of law is substantially harmed (McMullen, Bagby, & Palich, 2008).
a positive impact on entrepreneurship in such country. The research by Avnimelech and Zelekha, 2011, leads to the same conclusion.
This adverse effect of corruption can be also seen in the daily life of countries with high corruption (usually in developing countries), where it is often easier to pay a bribe to the government officials than to deal with the complicated state permission or licencing process. However, such situation discourages some prospective entrepreneurs from starting the business, especially when they try to avoid the corruption, which is illegal and immoral (Baláž, 2009; Mlčoch, 2006).
Based on the literature suggested above, the following hypothesis is formulated:
H1: The higher level of Freedom from corruption is positively associated with the emergence of new entrepreneurship.
2.3.2 Financing for entrepreneurs
The access to finance is a crucial for every business activity, especially for a new entrepreneurship. Van Auken confirmed by his research, that the lack of finance for entrepreneurs demotivates people to start the entrepreneurship and the capital serves as a barrier to entrepreneurship (Van Auken, 1999). The other research has revealed, that the access to financing determines not only the market entry but also the performance of businesses on the market after entering it (Ho & Wong, 2007).
Miller et al, 2014, concludes that financing for entrepreneurs and its impact on entrepreneurship is also connected to developed and functional financial markets, that allow the entrepreneurs to get the financing of their needs on the market.
In line with the theoretical and practical explanation, the research hypothesis has been formulated:
H2: The higher level of Financing for entrepreneurs is positively associated with the emergence of new entrepreneurship.
2.3.3 Governmental support and policies
The support and policies of the government for entrepreneurship are aimed to stimulate the economic growth and employment by positive influence on the number of entrepreneurs in the country (Minniti, 2008). Samila & Sorenson discussed the effects of governmental support and they considered even the possibility that the effect of this support on entrepreneurship could be negative (Samila & Sorenson, 2011). However, both of the studies have been conducted in the US.
From the practical point of view, governmental support of entrepreneurship or new state policies, which are in favour of entrepreneurship, can result in a decrease of the length of the process leading to the registration of a company, decrease of taxes for entrepreneurs or it can speed up the process leading to issuance of a building permit, which stimulates the economy and entrepreneurship.
The effects of governmental policies have been studied also by Da Rin et al., 2006, but his research covers especially the innovative hi-tech companies and venture capital. As the current research seems do not cover the effect of governmental support and policies on the entrepreneurship, but only certain markets or aspects, the research hypothesis has been formulated to examine the effects of governmental support and policies in general:
2.4 Individual level factors and their impact on entrepreneurship
This research analyses the impact of following individual level factors: fear of failure and perceived business capabilities on TEA rate. Korunka et al., 2003, suggested that there are several factors related to individual – personal level, that affect the entrepreneurship. The motivation for choosing the following variables is based on the academic literature and GEM databases, as described below.
2.4.1 Fear of failure
The impact of the fear of failure that prevents people from starting their business on the emergence of new entrepreneurship is not really clear because of the lack of research in this area. However, it has been proven that fear of failure “does not show to be significantly different for sustainable and traditional entrepreneurs” (Vonck, 2013). Amorós & Bosma, 2014, proposed a finding that self-perception and attitudes towards entrepreneurship have in fact the real effect on the real entrepreneurship. This is, however, a general statement, which leads formulation of the hypothesis that can explain the effect more specifically. There are just a few studies focused on the impact of fear of failure. Cacciotti and Hayton, 2014, observed, that the impact of fear of failure on the entrepreneurship is not a well-described topic, however, they concluded that it presents "a barrier to entrepreneurial behavior". researched the impact of fear of failure in China and Taiwan, but a comprehensive worldwide study is still missing. The impact of fear of failure on new entrepreneurship is a classical example of uncertainty and risk avoidance, which is to some extent a natural behavior of the majority of people. If the fear of failure is big, such person prefers employment to entrepreneurship (Baláž, 2009).
H4: The higher level of Fear of failure is negatively associated with the emergence of new entrepreneurship.
2.4.2 Perceived business capabilities
Entrepreneurial attitudes, including the perceived business capabilities, have been studied in comparative perspective between men and women in Iran and neighboring countries (Sarfaraz, 2017).
The GEM Consortium has prepared the basic analysis of their own data, which shows, that there might be some correlation between perceived business capabilities and TEA (GEM 2014 Global report, 2015), but the academic literature on this topic is almost missing, except the broad statements arising from the research of Amorós & Bosma, 2014, which confirm the effect of perceived business capabilities on entrepreneurship. Hatakka, 2015, examines and confirms the relationship between business capabilities and entrepreneurship only in the comparative study of Sweden and India.
However, the findings from managerial psychology indicate, that people are more open to start a business activity if they feel they have necessary skills and knowledge (Lukeš & Nový, 2005).
In order to bring a generally valid statement, this research tests the relationship, which is already examined and partially confirmed in some countries, on a cross-national data. Therefore, a specific hypothesis has been formulated:
H5: The higher level of Perceived business capabilities is positively associated with the emergence of new entrepreneurship.
2.4.3 Primary and secondary education entrepreneurial activities / Higher Education entrepreneurial activities
entrepreneurial activity, such as Dabic et al., 2011, but they focused only on the case of Slovenia. Similarly, the other study, made by Piña et al., 2014, focused only on Spain. However, I aim to test those effects not only on specific countries but rather on the worldwide data. The indirect effect of education entrepreneurial activities instead of a direct one has been selected for testing. It can be reasonably assumed that education entrepreneurial activities have an impact on perceived business capabilities because the people engaged in education entrepreneurial activities had some kind of training which increased their real business capabilities (and thus also the perceived one).
There exists also the research conducted by O’Connor, which discusses and confirms the impact of entrepreneurial activities and education on entrepreneurship, and even formulates the practical recommendation for policymakers: “To influence economic
growth, policy-makers should support and encourage the provision of entrepreneurship education as a means to connect new ideas, technologies and new applications of knowledge to business formation and expansion.” (O’Connor, 2013)
Moreover, there is another research by Johansen, which examines the willingness of young people with an entrepreneurial education to become entrepreneurs and finds a positive impact of such education (Johansen, 2010)
Raposo & Paço also examined the impact of entrepreneurial education and entrepreneurship activity and they have found that “education provides individuals with a sense of autonomy independence and self-confidence”. In terms of this research, this statement supports the idea that Perceived business capabilities are supported by presence of entrepreneurial education (Raposo & Paço, 2011). Considering these findings, the research hypotheses have been formulated:
H6: Primary and secondary education entrepreneurial activities positively affect the relationship between Perceived business capabilities on the emergence of new entrepreneurship.
2.5 Conceptual model
Societal level independent variable: Freedom from corruption (+) Financing for entrepreneurs (+) Governmental support and policies (+)
Individual level independent variable: Fear of failure (-)
Individual level independent variable: Perceived business capabilities (+)
Dependent variable: Total Early-Stage
Entrepreneurial Activity (TEA)
Moderating variables:
Primary and secondary education entrepreneurial activities (+) Higher education entrepreneurial activities (+)
Figure 2 – Conceptual model
3 Methodology
3.1 Research philosophy and approach
3.1.1 Positivist research philosophy
In line with the research question and quantitative nature of this research, positivist research philosophy is used (Neuman, 2010). The positivist research philosophy is characterized by the fact that the researcher does not interfere with the phenomenon and reality that is being studied and use value free observations (Neuman, 2010). This, however, does not mean that the data are not analysed, evaluated and discussed, but it rather describes the situation when the researcher does not play an active role in the data creation (Saunders et al., 2016). Using the positivist research philosophy is usually common for quantitative research, whereas interpretivist research philosophy is usually used for qualitative research (Saunders et al., 2016).
3.1.2 Deductive approach
There are two dominant research approaches that are used in business-related disciplines (Neuman, 2010). The inductive approach starts with collecting and analysing the data and builds the theory on the results of the analysis, whereas the deductive approach builds on the existing theoretical concepts, on which the hypotheses are formed and verified based on the data (Saunders et al., 2016).
3.2 Data collection and data sources
There are generally two main categories of data used in research; either primary data, collected by the researcher to answer the research question or secondary data collected by some other researchers (Saunders et al., 2016).
This research is based on the secondary data, as it is not really possible to collect all the primary current and past data across the countries from all continents for this research project.
The research uses credible databases in order to test the hypotheses outlined above. Nowadays, there are numerous databases available, such as a sound OECD database, which however includes only the data about OECD members (it has 35 members). The data about a larger pool of countries are collected in the sound and reliable Global Entrepreneurship Monitor (abbreviated as GEM) database, internally divided into two thematic pages for the same countries (GEM Adult Population Survey (APS), 2017 and GEM National Expert Survey (NES), 2017).
Moreover the GEM database states that it has the following three aims, which perfectly fits to the purpose of this research: “To measure differences in the level of early stage entrepreneurial activity between countries, To uncover the factors determining the levels of entrepreneurial activity, To identify policies that may enhance the level of entrepreneurial activity” (Global Entrepreneurship Monitor (GEM), 2017).
Gohmann (2012) suggest that Corruption perception index database prepared by Transparency International is a reliable database that can be used for research in international comparisons about the role of institutions on entrepreneurship, which is aligned with the topic and design of this research, therefore this database is used (Corruption Perception Index, 2017).
3.3 Methods used
As the data in the three databases used for this research do not include all countries around the world, the research is based on the sample of countries used by these three databases. The (relatively) smallest sample of countries is used in the Global Entrepreneurship Monitor (GEM Adult Population Survey (APS), 2017 and GEM National Expert Survey (NES), 2017), the Worldbank database and Corruption perception index database includes all the countries included in GEM database.
In order to have the sample enough to make a meaningful and reliable analysis, that allows to verify the hypotheses and to answer the research question, this research uses a high number of observations. Due to the limitations of the GEM project and database, some of the variables are collected by GEM project only since 2007. Therefore the data from 2007 until 2016 are analysed.
The dataset prepared by the author by combining the databases as specified above has 523 “full observations” in total, which allows me to use a regression model even for quite many variables. I use the term “full observation” to describe a “line” for one country for one year, where all the values of 9 variables are present. From the statistical point of view, each of these 523 “full observations” includes the value of all the dependent, independent, moderating and control variables. This dataset is called a panel data because it describes the variables for individual countries (cross-sectional) and across time in the interval of one year.
between balanced and unbalanced panel is crucial for this research; the definition is as follows: “A panel is said to be balanced if we have the same time periods, t = 1,..,T, for each cross section observation. For an unbalanced panel, the time dimension, denoted Ti, is specific to each individual,” (Hurlin, 2010).
I considered two types of treatment of unbalanced panel, which results in balanced panel, which allows performing the analysis much more easily.
First, it may be possible to compute or obtain the data about the variables for such countries that are not present in a given year in the GEM database. However, the values of all the variables from the GEM dataset are collected and computed by specific GEM methodology that cannot be replicated by the researcher for the countries that are not included in the dataset at all, neither to obtain the values of all variables for the countries, which are not included in a specific year.
Therefore the statistical methods that help me to obtain the necessary data are used. The specific method used in this research is called “imputing” (Enders, 2010). There are several types of imputing techniques, but the principle of imputing remains the same – the missing data are filled with other records (Gelman & Hill, 2006). I considered 3 types of imputing for the purpose of this research and selected the most appropriate one. The disadvantage of the imputing technique “Hot-deck - Last observation carried forward” is that it the bias and risk of invalid conclusions is increased (Molnar, 2008). The Cold-desk technique of imputation requires also another dataset, from which the missing data are taken (Enders, 2010). Finally, the technique of mean imputing seems appropriate for the purpose of this research, because it does not require additional datasets and it also does not change the important statistical parameters variable in the dataset (Enders, 2010; Gelman & Hill, 2006).
data for exact values of GDP for all countries between the years 2007 and 2016 and therefore the exact values of GDP and log(GDP) are used in the research in the whole period of 10 years. The computed mean value for all the other variables for a particular country is used for the years and countries for which the exact values have not been measured by GEM (Gelman & Hill, 2006).
By using the mean imputing method and direct import of the data about GDP for all years, I create a full dataset containing the values of all the variables for 104 countries between the years 2007 and 2016.
The dataset obtained by using imputing is now a referred as a balanced panel (Baltagi, 2008). This means that there is the same amount of observations for all countries used in this research, which all have the same weight in the analysis. Therefore the analysis is not negatively influenced more by the countries with more observations. The total amount of “full observations” is now 1040, every “full observation” still includes the values of 9 variables.
The full dataset including also all imputed variables cannot be analysed with a fixed effects model because this model would treat the countries separately while computation the results (Wooldridge, 2001). Therefore I choose the analysis with using a between estimator, which is a proper way how to analyse such data. Between estimator uses time-averages and than runs a cross-sectional regression. Grouping variables for the regression test with between estimator are individual countries.
The specification of the model is therefore as follows: 𝑇𝐸𝐴! = 𝛼 + 𝒙𝒊𝜷 + 𝜈!+ 𝜖!, where:
• 𝛼 means the intercept,
• TEA means Total Early-Stage Entrepreneurial Activity,
• Vector 𝒙𝒊 means vector of independent variables (Freedom from corruption,
“PC_BSEE”, Fear of failure rate, Governmental support and policies, Financing for entrepreneurs, log(GDP)),
• 𝜈! means the heterogeneity within the countries,
• 𝜖! means the error.
As the regression is used, several assumptions for a correct usage of regression have to be verified:
1. There are no significant outliers. The meaning of this assumption is to deal with observation with abnormal distance from other observations. This assumption is fulfilled, which can be seen in Table 3.
2. Multicollinearity of independent variables is not present. This assumption means that the independent variables are not dependent on each other (an independent variable is not a linear combination of the other variables). This assumption is verified by with using a multicollinearity test. The result of this test is attached as Annex 1.
3. Homoscedasticity of data. This assumption means that the variance in the error is constant. It is tested by using the Breusch-Pagan test. The result of the test is that heteroscedasticity is present in the data and this assumption is violated. Therefore the I use the regression with robust errors, which is used to deal with the cases which are in breach of homoscedasticity, as described in the chapter Findings.
bias. This assumption is presumed as valid in regression models without any tests.
5. Normality of the distribution of residuals. This assumption is fulfilled, as can be seen from result of the test of normality of residuals attached as Annex 2.
(Ramachandran & Tsokos, 2009; Rousseeuw & Leroy, 1987),
All the data analysis and data visualizations needed for this research are conducted with using advanced statistical software Stata, which allows me to perform non-standard regressions, as outlined above.
3.4 Variables
3.4.1 Dependent variable
The dependent variable used in all hypotheses of this research is “Total Early-Stage
Entrepreneurial Activity (TEA)” which is obtained from Global Entrepreneurship
Monitor database. The TEA rate is measured as a percentage of population between 18 and 64 years who are either nascent entrepreneur (“actively involved in setting up a business they will own or co-own; this business has not paid salaries, wages, or any other payments to the owners for more than three months”) or owner-manager of a new business (“owning and managing a running business that has paid salaries, wages, or any other payments to the owners for more than three months, but not more than 42 months”) (GEM Adult Population Survey (APS), 2017). TEA rate has been used also by similar studies, as it is a reliable proxy (Stephan & Uhlaner, 2010, Ho & Wong, 2007). 3.4.2 Independent variables
Freedom from corruption
Level of fear of failure
The level of fear of failure is defined as "Percentage of 18-64 population who indicate that fear of failure would prevent them from setting up a business” and it is extracted from GEM Adult Population Survey (APS), 2017. From the nature of percentage, it is a scale variable.
Level of Perceived business capabilities
The level of perceived business capabilities is defined as "Percentage of 18-64 population who believe they have the required skills and knowledge to start a business” and it is extracted from GEM Adult Population Survey (APS), 2017. From the nature of percentage, it is a scale variable.
Governmental support and policies
Governmental support and policies are defined as "The extent to which public policies support entrepreneurship" and it is extracted from GEM National Expert Survey (NES), 2017. It is a scale variable, “1” represents the lowest level Governmental support and policies, whereas “5” represents the highest level Governmental support and policies.
Financing for entrepreneurs
Financing for entrepreneurs is defined as "Availability of financial resources - equity of debt – for SMEs (including grants and subsidies)” and it is extracted from GEM National Expert Survey (NES), 2017. It is a scale variable, “1” represents the lowest level of Financing for entrepreneurs, whereas “5” represents the highest level of Financing of entrepreneurs.
3.4.3 Moderating variables
Primary and secondary education entrepreneurial activities and Higher education entrepreneurial activities are the moderators of the relationship between perceived
business capabilities on the emergence of new entrepreneurship. I suppose supposes that they positively moderate this relationship. They are defined as "the extent to which training in creating or managing SMEs is incorporated into the education and training". Both are obtained from GEM National Expert Survey (NES), 2017. It is a scale variable, “1” represents the lowest level of Primary and secondary education entrepreneurial activities or Higher education entrepreneurial activities, whereas “5” represents the highest level of Primary and secondary education entrepreneurial activities or Higher education entrepreneurial activities.
3.4.4 Control variable GDP per capita
The effects of economic development of countries on entrepreneurship have been already studied by various authors. GDP per capita is usually used as a variable that represents the economic development (Amorós & Bosma, 2014; Nyström, 2008 or Thurik et al., 2003).
The GEM data about entrepreneurship have been used by Amorós & Bosma; they have found out that the rise of GDP per capita is negatively correlated with entrepreneurship activity (Amorós & Bosma, 2014).
In order have a statistical model that better fits the data, Nyström proposes a small modification, as she uses rather the relationship between log of GDP per capita and entrepreneurship for her model - and she found this relationship statistically significant (Nyström, 2008).
that might be used (such as a number of technology start-ups), because these are highly correlated with the GDP, so they will not be able to explain the variance of the TEA. The gross domestic product of a particular country in a given year, expressed in USD per capita (Worldbank, 2017). It is a scale variable without a defined maximal value. 3.4.5 Summary of variables
This research uses 9 variables in total: one dependent variable, 5 independent variables, 2 moderating variables and one control variable. All the variables obtained from the data sources are shown in Table 1 below.
Table 1 – Summary of variables obtained from respective data sources
Type of the variable Name of the variable Data source Dependent variable Total Early-Stage Entrepreneurial Activity
(TEA)
GEM Adult Population Survey (APS), 2017
Independent variable Freedom from corruption Corruption Perception Index, 2017
Independent variable Fear of failure GEM Adult Population Survey (APS), 2017
Independent variable Perceived business capabilities
GEM Adult Population Survey (APS), 2017
Independent variable Governmental support and policies
GEM National Expert Survey (NES), 2017
Independent variable Financing for entrepreneurs
GEM National Expert Survey (NES), 2017
Moderating variable
Primary and secondary education entrepreneurial activities
GEM National Expert Survey (NES), 2017
Moderating variable Higher education
entrepreneurial activities
GEM National Expert Survey (NES), 2017
Control variable GDP per capita Worldbank, 2017
Source: Author
H6 and H7 and as a control variable in the regression model. These additional computed variables and their sources are shown in Table 2.
Table 2 - Summary of variables derived from the respective variables from Table 1
Name of the new
variable Original variables
PC_BSEE Perceived business capabilities, Primary and secondary education entrepreneurial activities
PC_PSEE Perceived business capabilities, Higher education entrepreneurial activities
log(GDP) GDP per capita
Source: Author
3.5 Countries used in the research
The research is based on a massive dataset, which includes the values of the variables (according to the Table 1 and Table 2) for 104 individual countries across the continents (GEM Adult Population Survey (APS), 2017; GEM National Expert Survey (NES), 2017). Such a big number of countries increase the informative value of the research because the regression and hypothesis tests are well supported by the data. The list of countries all countries is attached as Annex 3.
Therefore I obtained 1040 “full observations” after I used imputing, because this the final data includes 104 countries times 10 years of data.
Taking into account the total number of countries in the world, which is about 195, this research uses the data for more than a half of all existing independent countries in the world (Worldometers, 2017). It is highly unusual to have a statistical sample (n) – number of analysed countries – greater than the half of the population (N) – number of all countries in the world (Rousseeuw & Leroy, 1987).
3.6 Ethical issues & risks of the research
As the research is based on the analysis of secondary quantitative data, it does not involve or harm the human participants, animals or environment. No sensitive or controversial topics and questions are discussed and there is no conflict of interest of the researcher.
4 Findings
4.1 Descriptive statistics
In line with the research strategy outlined in the Methodology, the original data have been obtained from GEM Adult Population Survey (APS), 2017; GEM National Expert Survey (NES), 2017; Corruption Perception Index, 2017 and Worldbank, 2017. Such original data included 523 individual “full observations” (each “full observation” means the values of all variables for a single country in a given year). The results of descriptive statistics are included in Table 3 below.
Analysing the extreme values of each variable, substantial differences in the values can be observed.
The value of dependent variable – TEA – is just 2.10 % in Suriname in 2014 or 2.35 % in Italy in 2010. Contrary to that the TEA rate is 51.44 % in Vanuatu in 2012 or 41.46 % in Zambia in 2012. The mean of TEA is 12.28 % with the standard deviation of 8.00 %.
The values of independent variables are also in a big range:
The mean value of Perceived Business Capabilities is 51.44 with the minimum 8.65 in Russia in 2007 and 9.00 in Japan in 2010, whereas the maximal values around 89 are reported from African countries such as Senegal, Malawi, Nigeria or Uganda.
The other African country, Ghana, shows the lowest value (9.45) of Fear of Failure, whereas Kazakhstan shows the maximum of 75.42; the mean value is 34.34.
activities is just 1.15 in Egypt and 3.43 in Portugal. Egypt has also the lowest value for Higher education; the maximal values have been reported in the Philippines. However, these two moderating variables related to education are used also for computing the compound variables (as stated in Methodology) for the purpose of this research. Interaction effects of the compound variables and Perceived Business Capabilities are tested later in the regression, which can unveil that
The highest level of corruption (represented by the minimal value of freedom from corruption) was in Libya in 2013, followed by Venezuela and Angola; the least corrupt countries are Finland, Denmark and Finland.
The minimal level of financing for entrepreneurs was in Vanuatu; whereas the entrepreneurs in Israel have the best access to finance.
The weakest support of entrepreneurship by governmental support and policies can be found in Greece (1.59) and in Hungary (1.65); Tunisia with 4.55 stands on the other side.
The level of GDP was only about 333 USD per capita in Malawi in 2012, the maximal level of GDP per capita in the dataset is reported by Luxembourg (about 119 173 USD), followed by Norway. In line with the methodology section and respective literature, log(GDP) has been calculated and used in the research.
Table 3 – Descriptive statistics
Variable Mean Std. Dev. Min Max
Perceived Business Capabilities 51.44 15.21 8.65 89.48 Fear of Failure 34.34 9.45 10.43 75.42 Total Early-Stage Entrepreneurial Activity (TEA) 12.28 8 2.1 52.11 Primary and secondary education entrepreneurial activities 2.03 0.38 1.15 3.43 Higher education Entrepreneurial Education 2.84 0.34 1.79 3.83 Freedom from corruption 5.24 2.04 1.5 9.4 Financing for entrepreneurs 2.52 0.42 1.45 3.85 Governmental support and policies 2.55 0.48 1.59 4.55 GDP 22535.17 22507.62 332.92 119172.7 PC_PSEE 146.2 48.6 21.78 276.99 PC_BSEE 103.31 34.1 14.4 209.3 log(GDP) 9.48 1.14 5.81 11.69 Source: Author 4.2 Regression analysis
The dataset containing the balanced panel created by imputing the original data is analysed by using the between estimator, as described in the Methodology chapter. The significance level of 5 % is used, unless indicated otherwise.
4.2.1 Regression test of the impact of control variable on TEA
The regression test result shows that the p-value is 0.000, which is lower than the significance level of 0.1 % (0.001); therefore the null hypotheses of the regression can be rejected. Acceptance of alternative hypothesis of the regression indicates that the predictor – log(GDP) – is strongly associated with the level of TEA. The coefficient of the log(GDP), which is the output of the regression, is approximately -3.97, which means that log(GDP) affects the TEA negatively. The adjusted R-squared of this simple model is 0.31, which means that it explains 31 % of the variance of TEA; therefore a meaningful control variable is chosen for further tests.
Table 4 – regression for control variable
Linear regression Number of observations 1040 F (1,1038) 443.87 Prob. > F 0.000 R-squared 0.2995 Adj. R-squared 0.2989
Variable Coefficient Std. Err. P>t Sign. logGDP -3.881152 0.1842184 0.000 ***
_cons 49.18122 1.701845 0.000 ***
* - significant on 5 %, ** - significant on 1 %, *** - significant on 0.1 %
Source: Author
4.2.2 Regression test of the impact of independent and control variables on TEA Second, a similar test is performed on the model with independent and control variables and their impact on TEA rate, without any moderating variables. The test is performed on 1040 “full observations”.
The p-values of the tested variables indicate that only Perceived Business Capabilities, Freedom from corruption and log(GDP) are significant at 5 % significance level.
explains 59 % of the variance of TEA. The model with such a high R-squared is considered as a strong model (Gelman & Hill, 2006).
Table 5 – regression for control variable
Linear regression Number of observations 1040 F (1,1038) 230.88 Prob. > F 0.000 R-squared 0.5728 Adj. R-squared 0.5704
Variable Coefficient Std. Err. P>t Sign. PerceivedCapabilities 0.3962181 0.055951 0.000 *** FearOfFailureRate 0.1390461 0.0744021 0.065 Corruptionperceptionindex 1.25222 0.5126172 0.016 * financingforentrepreneurs -2.060708 2.034097 0.314 Governmentalsupportandpolicie -0.3564518 1.705039 0.836 logGDP -2.659875 0.7807099 0.001 *** _cons 11.7301 10.20282 0.253 * - significant on 5 %, ** - significant on 1 %, *** - significant on 0.1 % Source: Author 4.2.3 Regression test of the final model, including the moderating variables
activities is significant just about 5 % significance level, Financing for entrepreneurs is significant just about 10 % significance level, the other variables are not significant even at 10 % significance level. The value of R-squared of this model is 0.6265, which means that this model explains 62,7 % of the variance of TEA. The model with such a high R-squared is considered as a strong model (Gelman & Hill, 2006).
Table 6 – regression with using between estimator
Regression with between estimator, group variable: country Number of observations 1040 F (8,95) 19.92 Prob. > F 0.000 R-squared 0.6265
Variable Coefficient Std. Err. P>t Sign. PerceivedCapabilities 0.3165651 0.1253693 0.013 * PC_PSEE -0.0336102 0.0425965 0.432 PC_BSEE 0.0764557 0.0390078 0.053 FearOfFailureRate 0.1111006 0.0754548 0.144 Corruptionperceptionindex 1.161043 0.5144058 0.026 * financingforentrepreneurs -3.505459 2.144385 0.105 Governmentalsupportandpolicie -0.4678655 1.747154 0.789 logGDP -2.639557 0.783604 0.001 *** _cons 17.95542 10.84455 0.101 * - significant on 5 %, ** - significant on 1 %, *** - significant on 0.1 % Source: Author
account the regression (Stata, 2017). The other method, which is used in the research, uses a basic arithmetic (Cameron, 2006, Stata, 2017).
• when the respective coefficient obtained from the regression is greater than zero: § the p-value for one-sided research hypothesis proposing a negative influence on TEA is computed by subtracting the half of the p-value from 1
§ the p-value for one-sided research hypothesis proposing a positive influence on TEA is computed by dividing the p-value obtained from the regression
• when the respective coefficient obtained from the regression is lower than zero: § the p-value for one-sided research hypothesis proposing a negative
influence on TEA is computed by dividing the p-value obtained from the regression
§ the p-value for one-sided research hypothesis proposing a positive influence on TEA is computed by subtracting the half of the p-value from 1 (Stata, 2017).
Naturally, the significance of the specific variables in this chapter, therefore, differs from the significance computed in the regression, but the coefficient of all variables remains the same.
The significance of all research hypotheses is tested by this method:
H1: The higher level of Freedom from corruption is positively associated with the emergence of new entrepreneurship.
positive impact is accepted. Therefore the research hypothesis H1 is supported by the data.
H2: The higher level of Financing for entrepreneurs is positively associated with the emergence of new entrepreneurship.
The coefficient for Financing for entrepreneurship is equal -3.51, which is lower than zero. A negative value of the coefficient means that TEA is affected negatively by the Financing for entrepreneurship. As the one-sided p-value is equal 0.9475, which is greater than 0.05, the null hypothesis (negative impact) cannot be rejected. Therefore the research hypothesis H2 is not supported by the data.
H3: The higher level of Governmental support and policies is positively associated with the emergence of new entrepreneurship.
The coefficient for Governmental support and policies is equal -0.47, which is lower than zero. A negative value of the coefficient means that TEA is affected negatively by the Governmental support and policies. As the one-sided p-value is equal 0.6055, which is greater than 0.05, the one-sided null hypothesis of a negative impact cannot be rejected. Therefore the research hypothesis H3 is not supported by the data.
H4: The higher level of Fear of failure is negatively associated with the emergence of new entrepreneurship.
The coefficient for Fear of failure is equal 0.11, which is greater than zero. A positive value of the coefficient means that TEA is affected positively by the Fear of failure. As the one-sided p-value is equal 0.928, which is greater than 0.05, the one-sided null hypothesis of a positive impact cannot be rejected. Therefore the research hypothesis H4 is not supported by the data.
H5: The higher level of Perceived business capabilities is positively associated with the emergence of new entrepreneurship.
and secondary education entrepreneurial activities, which is used also in the hypothesis H6) and PC_PSEE (together with higher education entrepreneurial activities, which is also used in the hypothesis H7). Therefore it is not possible to easily isolate the effect of this variable from this model because the line “Perceived Business Capabilities” in the regression model in Table 6 explains the effect of the variable only partially.
Therefore the total effect of Perceived Business Capabilities is analysed from the previous regression model described in Table 2, which does not include the moderating variables.
The coefficient for Perceived Business Capabilities is equal 0.40, which is greater than zero. A positive value of the coefficient means that TEA is affected positively by the Perceived Business Capabilities. As the one-sided p-value is equal 0, which is lower than 0.001, the one-sided null hypothesis of a negative impact is rejected and alternative hypothesis of a positive impact is accepted even on 0.1 % significance level. Therefore the research hypothesis H5 is strongly supported by the data.
H6: Primary and secondary education entrepreneurial activities positively affect the relationship between Perceived business capabilities on the emergence of new entrepreneurship.
The coefficient for the moderating effect of Primary and secondary education entrepreneurial activities is equal 0.076, which is greater than zero. A negative value of the coefficient means that relationship between Perceived business capabilities and TEA is affected negatively by the Primary and secondary education entrepreneurial activities. As the one-sided p-value is equal 0.0265, which is lower than 0.05, the one-sided null hypothesis of a negative impact is rejected and alternative hypothesis of a positive impact is accepted. Therefore the research hypothesis H6 is supported by the data.
H7: Higher education entrepreneurial activities positively affect the relationship between Perceived business capabilities on the emergence of new entrepreneurship.
relationship between Perceived business capabilities and TEA is affected negatively by the Higher education entrepreneurial activities. As the one-sided p-value is equal 0.784, which is greater than 0.05, the one-sided null hypothesis of a negative impact cannot be rejected. Therefore the research hypothesis H7 is not supported by the data.
4.3 Further testing
After that, I need to deal with heterogeneity, which is present in the imputed panel data (with both time and cross-sectional dimension).
Therefore, I use the method of computation of means for all variables for individual countries and fit the OLS regression model on this mean data to check whether the results are the same as when applying panel data regression with between estimator. Despite that it may seem that such method ignores the time dimension of the data, the time dimension is reflected as the calculation of means, which takes into account the values of all variables in all the analysed years.
When we compare this method with the analysis of the variables for just one year, the mean data are much stable, as they handle with the extreme values of some variables that might be present during one or two years, but not during 10 years.
The regression of the modified data with the mean values shows that:
Table 7 – regression test without robust errors Number of observations 104 F (1,1038) 19.86 Prob. > F 0.000 R-squared 0.6258 Adj. R-squared 0.5943
Variable Coefficient Std. Err. P>t Sign.
PerceivedCapabilities 0.316871 0.1255555 0.013 * PC_PSEE -0.0334381 0.042691 0.435 PC_BSEE 0.0761666 0.0390451 0.054 FearOfFailureRate 0.111145 0.0755367 0.144 Corruptionperceptionindex 1.145658 0.5130764 0.028 * financingforentrepreneurs -3.521525 2.147138 0.104 Governmentalsupportandpolicie -0.4463023 1.747968 0.799 logGDP -2.624006 0.7846577 0.001 *** _cons 17.89165 10.87175 0.103 * - significant on 5 %, ** - significant on 1 %, *** - significant on 0.1 % Source: Author
In order to build a valid model, also needs to check the rest of assumptions or regression, which are stated in the Methodology chapter.
First, I have tested the modified dataset whether there is a normal distribution of residuals.
The result is that the null hypothesis of a normal distribution of residuals cannot be rejected at least at the 8% level, which is still acceptable for this model (Wooldridge, 2001). The graphical representation of the normal distribution of residuals is added as a histogram in Annex 2 together with the test of normality, which is included as Annex 6. Moreover, I also test whether heteroscedasticity is present in the analysed data with using Breusch-Pagan test, which is included as Annex 7.
in the data needs to be reflected because the assumption of homoscedasticity is violated (see the Methodology chapter for further details; Gelman & Hill, 2006).
Despite that that the assumption of homoscedasticity is violated, the regression model can still be used with a special method that handles the heteroscedasticity. This method is called as White robust errors (Rousseeuw & Leroy, 1987). These can be also known as heteroskedasticity-consistent standard error (HCSE) (Hayes & Cai, 2007).
This important, but not very well know econometrical method has some great advantages: “With this approach, the regression model is estimated using OLSs, but an alternative method of estimating the standard errors is employed that does not assume homoskedasticity. The appeal of this method lies in the fact that, unlike such methods as WLS, it requires neither knowledge about nor a model of the functional form of the heteroskedasticity. It does not require the use of an arbitrary transformation of Y, and no intensive computer simulation is necessary” (Hayes & Cai, 2007). Therefore this method is suitable, as the there is no knowledge about the heteroskedasticity, which is present in the data.
Moreover, further tests verify that multicollinearity is not present, as this is also a requirement for using a regression model. This allows the researcher to build one big multiple regression model by including all the variables (Rousseeuw & Leroy, 1987). The table with the results of the test for multicollinearity can be found in Annex 1. Therefore all the criteria and assumptions of the regression model with robust errors are fulfilled. The results of this final model are as follows:
Note that the significance of respective variables is lower, which is caused by the big robust errors, which are however necessary for a proper analysis of this data.
The adjusted R-squared (Coefficient of determination) of this final model is 0.59, which represents the fit of the data to the final regression line. Compared to R-squared, the adjusted one takes into account and penalizes for the number of non-significant variables.
Table 8 – regression for with robust errors
Number of observations 104 F (1,1038) 15.19 Prob. > F 0 R-squared 0.6258 Adj. R-squared 5.8397
Variable Coefficient Std. Err. P>t Sign. PerceivedCapabilities 0.316871 0.1509736 0.038 * PC_PSEE -0.0334381 0.0485447 0.493 PC_BSEE 0.0761666 0.0482013 0.117 FearOfFailureRate 0.111145 0.0940215 0.24 Corruptionperceptionindex 1.145658 0.4678781 0.016 * financingforentrepreneurs -3.521525 2.736052 0.201 Governmentalsupportandpolicie -0.4463023 1.62214 0.784 logGDP -2.624006 0.7041122 0.000 *** _cons 17.89165 10.41074 0.089 * - significant on 5 %, ** - significant on 1 %, *** - significant on 0.1 % Source: Author
5 Discussion
This chapter summarises the research project and compares the results obtained from the regression and hypothesis testing with the theory described at the beginning of this thesis. Despite that I have tried to understand and study the relevant literature, carefully collect the data and perform the statistical analysis in a right way, the research has definitely certain limitations, which are identified and discussed at the end of this chapter. The propositions for the following research are included at the very end of this chapter.
5.1 Summary and discussion of the research
This quantitative research examines the factors that have an impact on entrepreneurship. In order to do so, it first presents the review of literature directly linked to the examined concepts, but also some important broader practical and theoretical issues are described to allow the reader to fully understand the researched topic. After the general review of literature, a specific literature review is presented, which provides the justification for the selection of 7 factors, which effects on TEA are analysed. However, these factors are not of the same type, both individual level factors and societal level factors are analysed in this research. This solution is not very common among other research papers in this field, but as Scott, 2002 suggested, analysis of both types of factors helps to examine the factors with an impact on TEA more precisely and with a higher reliability.
The methodological chapter presents and justifies the statistical methods used in this research. Moreover, it includes a description of the variables, which are used in this research, as they represent the examined factors.
According to the descriptive statistics in Chapter 4.1, there exist enormous differences in entrepreneurship rate across countries. For example, the value of TEA is just 2.10 % in Suriname in 2014 or 2.35 % in Italy in 2010. Contrary to that the TEA rate is 51.44 % in Vanuatu in 2012 or 41.46 % in Zambia in 2012. Similarly, the values of independent variables are extremely different across countries. The existence of a correlation between independent variables and TEA rate, which can explain these differences in TEA, is examined in Chapter 4.2.
First, the model describing the effect of only log(GDP) as control variables on TEA is introduced in Table 4. This model proves that log(GDP) is a good proxy of economic development of countries, which is used as a control variable, this variable is extremely significant in this model (even on the significance level of 0.1 %) and explains almost 30 % of the variance in TEA rate. This result is expected, as it is in line with the similar research by Amorós & Bosma, 2014, which lead to the conclusion that a rise of GDP has a negative impact on the TEA rate. Another research done by Nyström, 2008, used even the same value as this research – log(GDP) – and the results of the tests also confirm a strong negative effect of log(GDP) on TEA.
Second, the model describing the effect of the independent and control variables on TEA is introduced in Table 5. The result of this model shows that the impact of Perceived business capabilities and log(GDP) is significant even at 0.1 %, which indicates an extremely strong significance. Moreover, Freedom from corruption is significant on 5 % significance level. The other variables are not significant in the model. The strong positive impact of Perceived business capabilities on TEA is in line with the GEM 2014 Global report, 2015. Country specific research by Hatakka, 2015 (examining the relationship between these two variables in India and Sweden) and Sarfaraz, 2017 (examining the relationship between these two variables in Iran) also confirms the findings of this research.
Regarding to the research one-tailed hypotheses proposed in Chapter 2, I have tested them with using the p-value obtained from the regression, which is divided by 2 (in line with Stata, 2017 or Cameron, 2006).
The significance of these one-tailed tests is the same as when they are carried out directly from the regression (despite that exact p-values naturally differ, the decision on their significance is the same), expect one variable, which is significant only as one-tailed. As the explanations of the relationship between all other significant variables are provided above, and they hold the same for one-sided testing, I now comment only on the effect of compound variable PC_BSEE, which consists of Perceived business capabilities moderated by presence of Primary and secondary education entrepreneurial activities. This significance of the moderating impact of the variable PC_BSEE can also be supported and explained by Raposo & Paço, 2011, who found that “education provides individuals with a sense of autonomy independence and self-confidence” or by O’Connor, 2013.
Overall, the only one-tailed hypotheses that are supported with the analysis of the dataset used in this research are: H1 (positive impact of Freedom from corruption), H5 (positive impact of Perceived business capabilities) and H6 (moderating effect of Primary and secondary education entrepreneurial activities).
