The effect of a change in the stock market on the entrepreneurial activity : a cross-country analysis

(1)

The effect of a change in the stock

market on the entrepreneurial

activity; a cross-country analysis

Author: Ruben Visser

Supervisor: Dr. J.C.M. van Ophem

A master thesis to fulfill the requirements for the degree of Master of Science Econometrics

in the

Department of Quantitative Economics Faculty of Economics and Businesses

University of Amsterdam

(2)

Declaration of Authorship

I, Ruben Visser, declare that this thesis titled, “The effect of a change in the stock market on the entrepreneurial activity; a cross-country analysis” and the work presented in it are my own. I confirm that:

• This work was done wholly or mainly while in candidature for a re-search degree at this University.

• Where any part of this thesis has previously been submitted for a de-gree or any other qualification at this University or any other institu-tion, this has been clearly stated.

• Where I have consulted the published work of others, this is always clearly attributed.

• Where I have quoted from the work of others, the source is always given. With the exception of such quotations, this thesis is entirely my own work.

• I have acknowledged all main sources of help.

• Where the thesis is based on work done by myself jointly with others, I have made clear exactly what was done by others and what I have contributed myself.

Signed: Ruben Visser

Date: August 5, 2016

“Have the courage to follow your heart and intuition. They somehow know what you truly want to become.”

(3)

UNIVERSITY OF AMSTERDAM

Abstract

Faculty of Economics and Businesses Department of Quantitative Economics

Master of Science Econometrics

The effect of a change in the stock market on the entrepreneurial activity; a cross-country analysis

by Ruben Visser

This study investigates the underlying factors’ return on entrepreneurial activity with specific focus on the relation between stock markets and en-trepreneurial activity. Using a panel data set consisting of 72 countries over the period 2006-2014 the rate of Total Entrepreneurial Activity (TEA) has been researched. By means of an exploratory factor analysis it was shown that two factors are affecting the entrepreneurial activity: innovative oppor-tunity and socioeconomic trends towards entrepreneurship. By means of a fixed effect regression it was shown that the stock market return has a one-year lagged effect on the TEA rates both direct as well as indirect via the under-lying socioeconomic factor.

(4)

Acknowledgements

During the proces of writing this thesis, I have had various difficult moments. While I enjoyed writing about entrepreneurship and using the knowledge obtained during the past years of study, working on a process like this takes time and requires months of self-motivation. I would proba-bly not have succeeded if it wasn’t for the support and feedback I got from the people around me.

Therefore I would very much like to thank my supervisor dr. Hans van Ophem for his positive support and continuous belief in a strong result. His feedback was a of major help and talking to him always gave me the courage to continue on this sometimes lonely trip of writing a thesis.

Furthermore I would like to thank all the other professors of the Univer-sity of Amsterdam who have helped me with smaller or bigger questions in order to overcome the problems I was facing.

Finally I would very much like to thank all my friends and family who continuously supported me during this proces. They made sure I kept be-lieving in the final steps towards the end and they made sure I was able to relax whenever I had the time.

Thank u all.

(5)

List of Figures

2.1 Three step model to evaluate entrepreneurship . . . 4

2.2 GEM model . . . 5

2.3 Full GEM model . . . 6

2.4 Relation between nascent entrepreneurship rate and Gross National Income per capita . . . 10

4.1 Histogram Total (Early-stage) Entrepreneurial Activity . . . 19

4.2 Development of median of TEA rates . . . 19

4.3 Histogram Stock Market Returns, % year-on-year . . . 21

4.4 Stock Market Returns in all countries in the data set over given time period, % year-on-year . . . 22

4.5 Average Gross National Income per capita (US $) . . . 23

4.6 Average GNI per capita per region (US $) . . . 23

5.1 Period fixed effects on TEA rates . . . 24

5.2 Screeplot of eigenvalues and the corresponding factors . . . 29

5.3 Proposed relation between factors of entrepreneurship of var-ious researchers . . . 32

5.4 Difference in TEA period efffects vs estimated differences in TEA rates based on model . . . 38

(8)

List of Tables

3.1 Variables in baseline model . . . 13

3.2 Four-step approach to analyze mediation . . . 17

4.1 TEA rates per income group . . . 20

4.2 TEA rates per region . . . 20

4.3 Average stock market return per income group . . . 22

4.4 Average Stock Market Return per region . . . 22

5.1 Pooled regressions, s.e. in brackets . . . 25

5.2 Hausman test for random effects . . . 26

5.3 Fixed effects regressions, s.e. in brackets . . . 28

5.4 Exploratory Factor Analysis, no rotation . . . 30

5.5 EFA with Kaiser-Meyer-Olkin measure of adequacy and no rotation. Factor loadings < 0.35 are left out to improve clarity. 31 5.6 Fixed effects regression second model . . . 32

5.7 Four-step approach to analyze mediation . . . 33

5.8 Dimutrescu-Hurlin Test for Granger-causality in panel data set 35 5.9 Exploratory Factor Analysis with SM Ri,t−1, no rotation . . . 36

5.10 EFA with Kaiser-Meyer-Olkin measure of adequacy and no rotation with inclusion of variable SM Ri,t−1. Factor loadings < 0.35are left out to improve clarity. . . 37

5.11 Unstandardized results of coefficient . . . 37

5.12 Estimated lagged effect of SMR on TEA . . . 38

A.1 Countries participating in the dataset . . . 45 *

(9)

1. Introduction

Entrepreneurship has always played an important role in society. It gener-ates wealth, jobs, opportunities and is a driver behind innovation. Politi-cians try to stimulate entrepreneurship and are therefore interested in the driving forces behind entrepreneurship.

Various definitions of entrepreneurship can be found. The earliest def-inition, dating from the eighteenth century, used to describe entrepreneur-ship as: ’the process of bearing the risk of buying at certain prices and sell-ing at uncertain prices’ (Cantillon, 1755). Later, Schumpeter (1945) defined entrepreneurs as: "individuals who exploit market opportunity through technical and/or organizational innovation".

Although much could be said about the various definitions of entrepre-neurship, this paper will focus on the analysis of the Total (Early-Stage) Entrepreneurial Activity (TEA). This is an indicator defined by the Global Entrepreneurship Monitor (GEM). The GEM is the world foremost study of entrepreneurship and is carried out by a consortium of various experts and universities around the world to obtain globally comparable statistics about entrepreneurship since 1999.

Due to the recent events in 2008, in which stock markets around the world collapsed, much attention is paid to the spillover effects of this global financial crisis on the financial system. However, it is still unclear what all the effects of a change on the stock market can or will have on the en-trepreneurial activity in a country.

In this research, the effect of a change in the stock market on the en-trepreneurial activity will be investigated by means of a cross-country anal-ysis. By using data from the Global Entrepreneurship Monitor, the World Bank, the Global Competitiveness Report and data about the stock market returns, a model will be developed in order to explain the differences in the Total (Early-stage) Entrepreneurial Activity (TEA) per country. By consid-ering the spillover effects of a change in the stock market it is the goal of this research to determine what a change in a very corporate environment (i.e. the companies who make up a stock index and therefore greatly impact the overall stock market returns), can mean for entrepreneurs. Although one would naturally state that in ’good economic times’ there is more trust among investors and hence better opportunities to obtain the necessary fi-nancing to start a business, one could also argue that in times of corporate downfall the way is paved for people with entrepreneurial ambitions to grab their chances.

It is expected that the overall economic situation affects the total en-trepreneurial activity and that the stock markets in turn affects this eco-nomic situation. Although much attention could be devoted to the reversed effect of the entrepreneurial activity on the economic situation this will not

(10)

be considered throughout this research. The goal of this research is to ex-plain the differences in entrepreneurial activities among countries rather than looking at the effect of these entrepreneurial activities. Throughout this research, the stock market return will be considered as an exogenous variable.

In order to investigate the influence of a change in the stock market on the entrepreneurial activity it is necessary to be able to model the en-trepreneurial activity accurately. At first the previous literature will pro-vide an extensive analysis about macroeconomic factors influencing the en-trepreneurial activity.

The second chapter starts off with an introduction of the measure of entrepreneurial activity that is being considered throughout this research. Furthermore, the model will be described which is being used by the Global Entrepreneurship Monitor nowadays to evaluate the determinants of en-trepreneurship. The next paragraphs of chapter two will provide an exten-sive overview of the currently available knowledge about the determinants of entrepreneurship, the spillover effect of changes in the stock market and the other underlying factors found by previous researchers.

Chapter three outlines the structure of the analysis of this research. By using a five step evaluation model this approach takes great care of the difficulties of possible unobserved factors affecting entrepreneurial activity. Chapter four describes the data set that is being used throughout this re-search and gives some introductory statistics about the measures and vari-ables which are used. This is done to provide an overview of the differences of the entrepreneurial activities across various regions and across various income groups. Furthermore the stock market return will be described for all the various countries in the data set.

Chapter five describes the outcomes of the analysis of this research. When relevant, the procedure and the models are being explained. This chapter will also contain the implications of the results of the various mod-els and works towards the final goal of analyzing the effect of a change in the stock markets on the entrepreneurial activity.

In chapter six, the conclusions of this research are all put together. This chapter describes the overall proces and highlights the most relevant results in order to answer the main research question.

Chapter seven describes future possibilities of further extending this re-search and can be used to obtain inspiration on how to use econometrics to evaluate (successful) entrepreneurship.

At last, the appendix contains a list of all the countries participating in the data set, the amount of available observations per country, their en-trepreneurial activity rates and their position on the Global Innovation In-dex 2015.

(11)

2. Background

2.1 Entrepreneurial Activity

The term entrepreneur (derived from the french verb entreprendre, meaning: ’to undertake’) has been used throughout the decades by many researchers in various ways. In their quest to find the meaning of entrepreneurship Hébert and Link (1989) found that Richard Cantillon (c. 1680–1734) was one of the first use the term entrepreneur. After Cantillon there have been many researchers trying to give a definitive description of entrepreneurship such as Schumpeter (1945), Schultz (1980) and Kirzner (1997).

And although much can be said about the proper definition of entrepre-neurship, during this research one definition will be followed. In 2006, The Organisation for Economic Co-operation and Development (OECD) decided to start with a program called Entrepreneurship Indicators Program (EIP). The EIP was launched in order to obtain globally comparable statis-tics about entrepreneurship and therefore a unique definition of ’entrepreneur’, ’entrepreneurial activity’ and ’entrepreneurship’ was needed. Commissioned by the OECD Ahmad and Seymour (2008) suggested the following defini-tions based on their research:

• Entrepreneurs are those persons (business owners) who seek to gen-erate value, through the creation or expansion of economic activity, by identifying and exploiting new products, processes or markets. • Entrepreneurial activity is the enterprising human action in pursuit

of the generation of value, through the creation or expansion of eco-nomic activity, by identifying and exploiting new products, processes or markets.

• Entrepreneurship is the phenomenon associated with entrepreneurial activity.

These definitions will be used throughout this research. However, in order to model the entrepreneurial activity accurately a fourth measurement, de-scribed by Wong, Ho, and Autio (2005) and used extensively by the Global Entrepreneurship Monitor (GEM) will be used:

• Total (Early-Stage) Entrepreneurial Activity (TEA). The TEA rate mea-sures the proportion of working-age adults in the population who are either involved in the process of starting-up a business or are active as owner-managers of enterprises less than 42 months old (i.e. maxi-mum 3,5 years old).

(12)

This variable is, according to the researchers from the GEM, the most accu-rate representation of the amount of active start-ups in a country. In con-trast, one could have used the nascent entrepreneurship rate or the new busi-ness ownership rate. However, data about busibusi-ness-registrations are most frequently provided by the local governmental agencies and there are ma-jor differences between countries about the restrictions on the declaration of the existence of a new business. In order to do a proper cross-country analysis, it is inevitable do use globally comparable statistics. Therefore the variable of interest will be the TEA rate.

2.2 Determinants of Entrepreneurial Activity

In order to have a better understanding of the factors affecting the en-trepreneurial activity Ahmad and Hoffmann (2008) carried out a research, again commissioned by the OECD, to come up with A Framework for Ad-dressing and Measuring Entrepreneurship. This research aimed at providing a better understanding of both the factors affecting entrepreneurship, as well as the impact of entrepreneurship.

To quantify their results Ahmad and Hoffmann (2008) used a three step model to evaluate entrepreneurship:

FIGURE2.1: Three step model to evaluate entrepreneurship

Determinants Entrepreneurial_performance Impact

In order to create a model to explain entrepreneurial activity appropriately, it is necessary to have a clear image of the determinants of entrepreneur-ship. By following the literature review done by Audretsch and Erdem (2005), the relevant factors can be divided into four categories: Individual-level (cognitive) factors, Firm-Individual-level (including university start-ups) factors, Regional (cluster)-level factors and the Impact of policy.

On the individual level Verheul et al. (2002) found that the supply and demand of entrepreneurship strongly affect the amount of entrepreneurial activity. On the supply side, they found that the demographic composition of the population and the amount of resources available to people who as-pire to become an entrepreneur, have strong effects. On the demand side, they find that the technological developments and the governmental poli-cies to support innovative initiatives effects the amount of entrepreneurs. Furthermore, Thornton and Flynn (2003) found that the available network and the geographic situation affects the amount of entrepreneurs. They found that a lack of a network can be a major obstacle when starting a busi-ness which they classify as regional level factors.

On the firm level factors, many researchers have found a significant re-lation between the amount of collaboration between industries and univer-sities, and the amount of start-ups affiliated with these universities. Lock-ett, Wright, and Franklin (2003) showed that universities who are ranked higher, have more successful spin-off start-ups from the scientific knowl-edge and a more intensified collaboration with companies.

(13)

Within the literature various conclusions can be found about the spe-cific influences of the various factors on the various levels, for example by Audretsch (2002) and Schramm (2006). However, as stated by Ahmad and Hoffmann (2008): "The differences between these various studies are often largely semantic; most agree for example that entrepreneurs and entrepreneurship are cre-ated by a combination of three factors: opportunities, skilled people and resources". The Global Entrepreneurship Monitor (GEM) has devoted much of its time on researching the various factors affecting entrepreneurship and the GEM came up with a general framework to evaluate the determinants of en-trepreneurship. This is called the GEM - model. Reynolds et al. (2005) exten-sively describe the process of creating this framework. From their research, Reynolds et al. (2005) came up with six overall determinants affecting en-trepreneurship: Regulatory framework, R&D and technology, entrepreneurial capabilities, culture, access to finance and market conditions.

Based on these six determinants the entrepreneurial performance can be measured. In order to do so, Ahmad and Hoffmann (2008) decided to use three indicators: Firm based, Employment based and Wealth. Each of these indicators is divided into various components which can be quantified and measured, and therefore can be compared cross-country.

The literature suggest that entrepreneurship impacts three things. As Wennekers and Thurik (1999) laid out, entrepreneurship is of great influ-ence on economic growth and the reduction of poverty and unemployment. Malchow-Møller, Schjerning, and Sørensen (2011) additionally found that entrepreneurship has a great influence on job creation. Therefore Ahmad and Hoffmann (2008) use these three pillars to measure the impact of en-trepreneurship.

The total model therefore becomes:

FIGURE2.2: GEM model

Although many models can be used to evaluate entrepreneurship, through-out this research, this model will provide the starting point for modeling entrepreneurial activity. This research will focus solely on the determinants of entrepreneurship. The main question is whether or not a change on the stock market affects the total entrepreneurial activity in a country and therefore no attention will be paid to the subsequent entrepreneurial per-formance or the impact of this entrepreneurial activity. The hypothesized relation between the stock market return and the total entrepreneurial ac-tivity will be described in the next paragraphs.

In order to do an empirical analysis on the six determinants of entrepre-neurship, there is a need for more specific measurable indicators of which the data is available and which can lead to pragmatic political advises.

(14)

Therefore Ahmad and Hoffmann (2008) in collaboration with the Global Entrepreneurship Monitor (GEM) came up with various measurable indi-cators.

2.3 Entrepreneurial Determinants

Over the years, various institutions have used different metrics to evalu-ate entrepreneurship. By combining the knowledge and data of the World Economic Forum (WEF), The OECD, GEM, Eurostat and Eurobarometer, Hoffmann, Larsen, and Oxholm (2006) from the International Consortium for Entrepreneurship (ICE) were able to do a quality assessment of the var-ious available indicators.

Without going in too deep into all the suggested indicators one of the most important aspects with regard to this research are the indicators which are used to quantify the level of access to finance. They pointed out that access to finance could be measured by regarding five matters: Access to debt financing, business angels, venture capital, access to other types of equity and stock markets. One of the main problems in evaluating the access to finance is the lack of available data about the possibilities to extract external financing via business angels and or venture capital.

It is not only access to finance that is difficult to measure. All other five determinants are difficult to quantify with globally comparable statistics as well. However, Ahmad and Hoffmann (2008) made a great effort to come up with underlying indicators to measure the determinants.

In Figure 2.3 the full GEM model, including all the underlying indica-tors, is presented which has been used by the OECD and various other in-stitutions to evaluate entrepreneurship and advice governments about their policy.

(15)

When modeling entrepreneurial activity it is essential to make sure that all determinants of entrepreneurship are taken into account. Nonetheless it should not be forgotten that both time effects and unobserved country spe-cific effects should also be taken into account. Although The GEM provides globally comparable statistics it does not mean that all cross-country differ-ences can be expected to be captured by modeling according to these de-terminants because there is no country specific data for every determinant on itself. These determinants have to be estimated based on available data. Furthermore, although stock markets are mentioned to be one the main in-dicators of access to finance, till now there is no conclusive evidence that stock markets affect entrepreneurial activity. Chapter three will provide a description of which determinants and corresponding indicators will be used throughout this research.

2.4 Stock Markets

As early as 1602, the Dutch East India Company (VOC) started with the exchange of stocks, according to Petram (2014). This was one of the first steps towards a tremendous change in the globalization of the world econ-omy. Nowadays virtually every developed country has their own stock market and stocks worth trillions of dollars are traded daily on the for-eign exchange market. One of the main instruments to investigate stock markets is the so-called stock market index. This index is a mathematical construct which generally computes a weighted average of the prizes of se-lected stocks and bonds. Such a stock market index cannot be traded, how-ever, very often stock-trackers are constructed which can be traded and of which the value is strictly connected to the value of the stock market index. To be able to compare stock market statistics globally it is therefore con-venient to use stock trackers. When the value of a stock tracker goes up, this means that the weighted average of the value of the selected stocks is up; hence stocks in a specific country are doing better. In the Global Compet-itiveness Report these indices are used to calculate the annual stock market return per country. This variable will be the measure of the performance of the stock markets. Strictly speaking the effect of a change in the stock market can be different among various sectors and therefore could affect entrepreneurial activity in different ways for different sectors. However, during this research the yearly aggregated data for both the stock market returns as well as the yearly aggregated data for entrepreneurial activity will be used in order to estimate the aggregated total effect of stock market returns on the total entrepreneurial activity.

By studying the effects of stock markets on the economy as a whole, Demirgüç-Kunt and Levine (1996) have made major contributions to the academic knowledge. In their research four main findings are discussed:

First of all, they constructed objective criteria to compare stock markets cross countries. They defined the criteria of performance based on liquidity, concentration, volatility, institutional development and international inte-gration. By using these criteria they were able to compare the stock markets of various countries not only on their returns, but also on their impact on the rest of the economy as a whole.

(16)

Secondly, they find that countries with better developed stock markets, tend to have better developed banks and other financial intermediaries. Therefore an evolved stock market results in a general increase in financial development.

Thirdly, they empirically showed that stock market development can be used to make predictions about long-run economic growth. This is highly relevant in this research since Carree and Thurik (2003) showed that en-trepreneurial activity is also strongly connected to economic growth in a country.

Lastly, the research shows that stock market development affects the financing choices of corporates and leads to an increase in the supply of bank loans. This indicates that stock market development actually affects the access to finance in the broadest sense.

In a earlier paper by King and Levine (1993), they investigated the re-lationship between finance, entrepreneurship and growth. They develop an endogenous growth model to link these three factors based on theories developed by Schumpeter (1913). An important aspect of their analysis is that they consider financial institutions to affect productivity enhancing activities, i.e. entrepreneurship, in two ways: they evaluate prospective en-trepreneurs and they provide funding to the most promising ones. King and Levine (1993) state the following: "Better financial services expand the scope and improve the efficiency of innovative activity; they thereby accelerate eco-nomic growth." Which supports the view that the existence and performance of a stock market affects the economic development of a country.

2.5 Stock Markets, Entrepreneurial Activity and the

difficulties analyzing their relation

Various researchers have shown that stock markets and the amount of in-vestments made by companies are positively correlated, for example Baker, Stein, and Wurgler (2002). Although not specifically pointed out, this is a clear indication that the amount of financing granted by companies to star-tups (or to stimulate entrepreneurial ideas) is positively correlated to the developments on the stock market. Furthermore, as Schröder et al. (2013) pointed out by using a panel-data analysis, highly evolved financial mar-kets have a positive effect on the amount of active venture capitalists. The fact that they use a panel data set, in contract to many other researchers who focus solely on one country, means that they account for the various cross-country differences, as well as the differences over time, hence making the results more applicable on a global scale.

Access to finance is considered to be one of the most important fac-tors affecting entrepreneurial activity (Klapper, Laeven, and Rajan, 2004). It is difficult to find one suitable definition for ’access to finance’ but anal-yses show that the availability of both private- and bank-financed credit has a substantial effect on the amount of new business entries per country (Klapper, Laeven, and Rajan, 2004). Since new business entries per year is strongly correlated with the variable Total Entrepreneurial Activity (TEA) (Wennekers et al., 2005), it is expected that the availability of credit has a

(17)

strong effect on the TEA as well. Since it is expected that the stock mar-kets affect the availability of credit, it is expected that the stock marmar-kets indirectly affect the TEA rates.

Not only the financial development in a country affects the entrepreneurs, also the financial conditions affect the amount of new business entries. Klap-per and Love (2011) showed, by using a cross-country panel data analysis that the financial crisis of 2008 had a significant negative impact on the amount of registrations of new firms. In their analysis they used data of the various local registers of companies over the years 2004-2009. They ad-ditionally conclude that countries with higher levels of GDP experienced a more severe decline in firm-registrations during the financial crisis than less developed countries. This states that there is a qualitative link between entrepreneurial activity and financial conditions, but it is interesting to see if this effect can be quantified based on specific stock performance. There is no universally accepted definition of a financial crisis, but a so-called stock-market collapse, as we experienced in 2008, can lead to a a situation in which the value of assets and financial institutions decreases rapidly. This in turn is a situation which is generally described as a financial crisis.

2.6 Global Entrepreneurship Monitor

The Global Entrepreneurship Monitor (GEM) is a collaboration of various universities around the world. Started in 2009, the Babson College (USA) and the London Business School (UK) decided to join forces to create a bet-ter understanding of factors affecting entrepreneurship. In doing so they were forced to come up with globally comparable statistics which can be equally measured throughout the world. 17 years later, the GEM is consid-ered to be the richest source of data and information about entrepreneur-ship. With over 17 years of data, more than 200.000 interviews per year and over 500 field experts spread around the world, it is fair to say that the GEM has extensive knowledge about the subject.

When using the data provided by GEM, it makes sense to also work with their main indicator (TEA). The data of the GEM is gathered by means of panel questions to random individuals and extensive questions to en-trepreneurs. This is important to realize for this research, since it affects the way the results should be interpreted. During the yearly research per-formed by the GEM, two main approaches are being considered.

Firstly, the GEM conducts a National Expert Survey (NES). The result consists of a data set in which various entrepreneurial indicators are being graded on a scale from 1 to 5.This data is gathered by means of interviews with industry experts per country.

Secondly, the GEM conducts a Adult Population Survey (APS). In this survey general statistics about entrepreneurship are created. These statis-tics are all expressed in percentage of the total adult population. This data is gathered by interviewing a random selection of inhabitants varying from 1.000-20.000 per country.

(18)

2.7 Macroeconomic approach

One of the major disadvantages of the use of the data of the GEM, is the fact that all variables in the National Expert Survey are measured on a grading scale from 1-5 as graded by the individual who is being interviewed. Al-though this provides good insights in the perception of how political poli-cies affect entrepreneurship, it doesn’t necessarily explain the cross-country differences in the levels of TEA.

In order to explain these differences, Wennekers et al. (2005) focused on macroeconomic factors affecting the rate of nascent entrepreneurship. They used the framework proposed by Wennekers, Uhlaner, and Thurik (2002) and Verheul et al. (2002) in which they control the rate of entrepreneurship for economic, technological, demographic, cultural and institutional vari-ables. They found a significant U-shaped relation between economic devel-opment (measured by national income per capita) and the rate of nascent entrepreneurship. These results imply that for countries with very low lev-els of economic development, entrepreneurship rates are high. When eco-nomic development increases, the amount of entrepreneurs declines, up until the point where the countries are well-developed. Then a significant increase in entrepreneurship can be seen again. This is intuitively plausible, since in less-developed entrepreneurship can be a way to harvest a living, whilst in very well-developed countries entrepreneurship can be a way to benefit from new innovative solutions.

FIGURE 2.4: Relation between nascent entrepreneurship rate and Gross National Income per capita

When explaining differences in TEA rates, it therefore makes sense to account for the various levels of economic development with a quadratic specification.

Since differences in entrepreneurial rates are affected by many variables, the question of which variables to use to capture the ’proper’ framework for analyzing entrepreneurship, is difficult to answer. Valliere (2010) made a great effort in order to analyze the framework conditions sketches by the

(19)

GEM. By applying structural modeling and Principal Components Analy-sis it was derived that the current framework used by the GEM is an in-sufficient representation of the various factors affecting entrepreneurship. Valliere (2010) analyzed this by using a list of operational indicators which could explain the underlying determinants as proposed by the GEM. In or-der to do so, data was used from the Global Competitiveness Report (GCR) constructed by the World Economic Forum.

In this database, annual information can be found about multiple topics affecting entrepreneurship. This data is partly gathered by interviewing business executives from multiple sectors and high-ranking government officials. Another large part of this data set consist of data from the World Development Indicator Program, which is a program initiated by the World Bank.

With the Principal Component Analysis, Valliere (2010) proposed a new structure for analyzing entrepreneurial activity. He suggest to use four components: commercial munificence, technology openness, regulatory openness and technology influx. These components can be estimated by using var-ious variables from the GCR. In this analysis the dependent variable is Opportunity-based Entrepreneurial Activity (OEA).

These results indicate that a combination of these approaches could lead to a useful model that explains the cross-country differences in TEA and with which the effect of a change in the stock index on the TEA can be investigated.

(20)

3. Model and method

3.1 Overview

Based on the underlying theory it appears that there is no definitive way of explaining the cross-country differences in TEA rates. Due to a lack of un-derstanding, lack of comparable datasets and unmeasurable differences be-tween countries, one could use multiple approaches to explain these differ-ences. In this research, a combination will be made between the approaches of Wennekers et al. (2005) and Valliere (2010).

First of all, a baseline model will be developed based on the determi-nants of entrepreneurship as found in the GEM-model and as discussed by Valliere (2010). This model will contain all the variables which are expected to have a significant impact on the TEA rates.

Next, a factor analysis will be done. It is expected that the variables included in the baseline model can be divided into clusters which explain the common variance in TEA rates. These underlying factors make up the determinants that will be used in our next model.

Third, based on the factors found in the second stage, latent variables will be constructed in order to be able investigate the individual effects of the factors on the TEA rates. This enables us to decrease the amount of mul-ticollinearity in the model and reduces the dimensions of the explanatory variable matrix. Since it is expected that these underlying factors are not observable, an exploratory factor analysis (EFA) is used and not a principal components analysis (PCA). PCA assumes that the amount of variables can be reduced by extracting principal components from the observed variables which explain most of the variance, EFA assumes most of the covariance could be captured by various unobserved underlying factors. These factors are then extracted an new variables can be composed based on these factors. With these new variables a second model will be proposed to explain the variance in TEA rates. In this second model the factors, which are extracted in the factor analysis, will be used to estimate the effect of the underlying variables.

Next, the effect of a change in the stock index on the TEA rates will be considered. It is known from Klapper and Love (2011) that the financial crisis had a significant impact on the amount of new firm registration and this provides the basis for the main hypothesis. However, it is expected that a change in the stock market affects one or more of the previously extracted factors and therefore the indirect effect via the underlying factors should be taken into account.

In order to determine whether or not a change in the stock markets effect the entrepreneurial activity directly, the panel data Granger-causality test of Dumitrescu and Hurlin (2012) will be used. This test should provide

(21)

insights into the predictive power of the variable stock market return with respect to the TEA rates, corrected for the various country specific effects.

Together this should provide a thorough analysis of the cross-country differences in entrepreneurial activity. By using direct and an indirect es-timation procedure the effect of a change in the stock index on the TEA rates will be considered in a way that the theory suggests and is intuitively plausible.

3.2 Baseline model

Within the framework of the GEM, six determinants are considered to influ-ence entrepreneurship: regulatory framework, R&D and technology, entrepreneurial capabilities, culture, access to finance and market condition. However, Valliere (2010) showed that these clusters are not sufficiently diversified to explain differences in entrepreneurial activity due to mutual correlation. He there-fore proposed to use these four factors: commercial munificence, technology openness, regulatory openness and technology influx.

In this research, the focus will be on macro-economic factors affecting entrepreneurship. This enables us to use more independently described variables by the World Bank and this should increase the explanatory value of the model. The variables used in the baseline model are all chosen to cap-ture the maximal amount of indicators of entrepreneurship, given the avail-ability of the data. Due to the availavail-ability of the data is it expected that the extracted factors, which should describe the underlying determinants, will be slightly different from the GEM model and from the research by Valliere (2010). In this case it is expected that five underlying determinants affect en-trepreneurship: economic situation, technology openness, regulatory framework, demographics and culture. This would mean that more distinct factors could be extracted than by Valliere (2010) but less than in the GEM-model. This hypothesis is a combination of the various results in the literature and will be tested in the factor analysis. The variables used are described in Tabel 3.1.

TABLE3.1: Variables in baseline model

Item Description Source Hypothesized determinant

TEA Total (Early-stage) Entrepreneurial Activity GEM Dependent variable GDP Growth Annual % growth rate of GDP World Bank Economic Situation Employment Rate % labor force employed World Bank Economic Situation Log(GNI) Gross National Income per capita in $, Atlas method World Bank Economic Situation Stock market return Annual average stock market return GFDD Economic Situation Internet % of people with access to internet World Bank Technology Openness Uni-Ind University-Industry Collaboration. Scaled 0-100 GCR Technology Openness High-tech export High-tech % of manufactured exports World Bank Technology Openness Firm-tech absorb Firm technology absorption. Scaled 0-100 GCR Technology Openness Innovation Capacity for innovation. Scaled 0-100 GCR Technology Openness Cost of startup Cost of start-up procedures as % of GNI World Bank Regulatory Framework Days to startup Days required to start-up a business World Bank Regulatory Framework Gov-regulations Burden of governmental regulation. Scaled 0-100 GCR Regulatory Framework Total tax rate Total taxes in 2nd_{year as % of commercial profit} _GCR _{Regulatory Framework}

Population growth Annual growth rate of population World Bank Demographics International Tourism Tourism expenditures % of total export World Bank Demographics Friend Entrepreneur % of people that personally know an entrepreneur GEM Culture High status entrepreneur % of people agree entrepreneur gives high-status GEM Culture (Ex-)Communist Dummy variable for (ex)-communist countries Wikipedia Dummy Crisis1 Dummy variable equals 1 for years ’08, ’09, ’10 - Dummy Crisis2 Variable equals 1 for ’07 & ’11, 2 for ’08 & ’10, 3 for ’09 - Dummy Trend Linear trend to capture long-term changes - Linear variable

(22)

These variables are chosen based on reliability, coverage and availabil-ity. In an effort to capture as many entrepreneurial indicators whilst rely-ing on independent macro-economic data, these 16 variables were chosen. We use log(GNI) rather than GNI because it is expected that the effect of a percentual change in GNI on the TEA rate is a more appropriate mea-sure than an absolute change. For technology openness the variables were relatively easy to be found and in order to use globally comparable statis-tics about the regulatory framework we have chosen to include the cost of start-up and the days required to start-up a business. Both variables should capture many of the underlying indicators of the regulatory frame-work. For the demographic factors we included the population growth, the amount of international tourism (which hypothetically both contribute to more business opportunities for entrepreneurs) and for the cultural factor we included whether someone knows an entrepreneur and the status that entrepreneurs are generally considered to have. All variables are gathered in an unbalanced panel data set covering 90 countries over the years 2006-2014 (n=472). A more thorough description of the data set will be given in the next chapter.

Following the approach of Valliere (2010) the baseline model will pro-vide an upper bound on explanatory power of the chosen variables. The individual effect of the variables on the TEA rates will be estimated by means of a pooled Ordinary Least Squares model, a pooled Generalized Least Squares model and a Fixed-Effects model.

3.3 Factor analysis

The next step in analyzing the different variables is to investigate whether or not these variables can be subdivided into groups. It is expected that various variables contribute to the covariance of the TEA rates in the same way, and therefore could be grouped together. These groups are called Hy-pothesized determinants of entrepreneurship.

This step is being taken since the theory suggests that there are latent effects present; i.e. non-observable factors that affect the TEA rates. From both the GEM model and the analysis performed by Valliere (2010) we know that these underlying factors should be present. Since these factors and corresponding effects can not be measured directly, they have to be extracted and then estimated.

If it is expected that the variables are highly correlated, one can choose to apply variable reduction techniques to reduce the possible multicollinearity. The idea is quite simple: if a group of the observed variables account for the same fraction of covariance with the dependent variable, these observed variables can be grouped and reduced to one weighted component. This component should be uncorrelated with the other components of the model in order to increase the explanatory value of factor analysis.

In the area of variable reduction techniques there are many different possibilities. Osborne and Costello (2009) compared the various techniques and provide us with the best practices. In their research it is explained that next to Principal Component Analysis (PCA), most statistical packages have six more options to choose from when deciding which factor extrac-tion method is being used, namely: unweighted least squares, generalized

(23)

least squares, maximum likelihood, principal axis factoring, alpha factor-ing and image factorfactor-ing. Due to the severe lack of scientific proof of when to use which model, many researchers stick to the default option of apply-ing Principal Components Analysis.

However, in this case, we expect there to be various underlying unob-served factors. Therefore we have to apply an Exploratory Factor Analysis. To do so the principal axis factoring method is chosen because this method makes the least assumptions about the general structure of the underlying correlation matrix of the different variables. Based on the outcomes, it will be determined whether or not the factorization will be rotated afterwards. Since by default the factorization will be done in such a way that it max-imizes the explained covariance in the factors. Rotating is therefore only used in case of difficulties with the interpretation of the results.

3.4 Factors

After validating the underlying structure of the various observed variables, the basics of the underlying model to explain the variance in TEA rates will be known. It is expected that the factors will be made up of a linear combination of the several underlying variables. These factors should all be theoretically justified and their individual influence should not overlap with the other factors. So the correlation between the various factors should be considerably small. The whole idea of applying this variable reduction is to reduce the correlation between the variables in the model, hence this should be checked an corrected for if present.

These factors should help us to obtain more knowledge of the underly-ing framework of determinants of entrepreneurship. Furthermore it should enable us to investigate changes in underlying factors on the TEA rates, in-stead of the effect of individual variables. After the extraction method, we can define new variables which are based on the underlying factors. When these newly defined variables are used in the model, a new interpretation should be given to the effect of this variable. The factor loadings determine the indirect effect of the underlying variables on this factor.

3.5 Second model

When the underlying clusters are analyzed, the next model can be con-structed. This model will be used to capture the variance in the TEA rates. The model that is being used will be based on the results of the previous analyses, however the basic structure will be like this:

yit= αi+ δτt+ βXit+ γDit+ it (3.1)

• yit:is TEA.

• τt:is the linear trend variable.

• X_it :is the matrix consisting of the various clusters/determinants of entrepreneurship.

(24)

• it:is the error term.

The underlying goal of evaluating this model is to determine the effect of a change in the economic situation, which is hypothesized to be one of the determinants, on the dependent variable T EA. The focus of this re-search is to gain insights in the spillover effects of a change in the economic situation on the entrepreneurial activity. Although there is not one defini-tive variable to capture the economic situation the factor analyses should pro-vide a solid basis to be able to use a variable which represents the economic situation.

3.6 Mediation

In the end, this research focuses on the effect of a change in the stock market on the entrepreneurial activity. Since the theory does not give strong sup-port for a direct effect of a change in the stock index on the entrepreneurial activity, it will also be investigated if there is an indirect effect of the stock markets on the TEA rates.

In order to test whether or not the direct effect should be taken into account, the test for Granger-non causality will be used (Dumitrescu and Hurlin, 2012). This test is developed to test for Granger-non causality (Granger, 1969) in a panel-data set. This test should provide insights into the difficult question if the return of the stock markets directly effects the TEA rates. If it does, the stock market returns will be included in the factor analysis. If it doesn’t, the indirect effect will also be considered.

This indirect effect is assumed to follow from the relation of the stock markets to the economic situation in a country. Many researchers have de-voted their efforts to explain the spillover effects of a change in the markets or the effects of stock markets on economic growth (Demirgüç-Kunt and Levine, 1996), however in these papers it is never mentioned to investigate the effect of the stock markets on the economic situation, when the latter is measured by using a factor analysis and analyzed as a latent variable. Therefore, analyzing this relationship should provide more insights, due to the much broader coverage of the newly defined latent variable.

When estimating an indirect effect, one should consider the mediation effect. The mediator variables are extensively described by Baron and Kenny (1986). The general idea of mediation is as follows: Mediation is a esized causal relation between between variable X, M and Y . It hypoth-esizes that X affects M and M affects Y , in such a way that indirectly X affects Y . m is here defined as the mediator. Graphically it looks like this:

X −→ Ma −→ Yb (3.2)

However, if we look at:

X −→ Yc (3.3)

It does not necessarily have to be that a ∗ b = c.

Baron and Kenny (1986) proposed a four-step approach to analyze this relation. This approach is shown in Table 3.2.

(25)

TABLE3.2: Four-step approach to analyze mediation

Step Description Regression Tested effect

1 Regress X on Y to test path c Y = a + b1X +e X c

−→ Y 2 Regress X on M to test path a M = a + b1X +e X

a

−→ M 3 Regress M on Y to test path b Y = a + b1M +e M

b

− → Y 4 Multiple regression of X and M on Y Y = a + b1X + b2M +e X, M

c0

−→

b Y

This approach is most widely used by researchers but one should be aware of the assumptions underlying the use of this approach. For example, it is assumed that the relations in step 1 to 3 are all significant. These steps are undertaken in order to be able to investigate how the first-order rela-tions between the variables are defined. If these direct relarela-tions are not all significant, it can be doubted that there is a mediation effect.

When analyzing this mediation effect, two different sources of stock-market data will be considered. First of all, the stock stock-market return from this year, and second, the stock market return of the year before will be considered. It is expected that if there is any effect of the stock market return on the economic situation, it could very well be a lagged effect. Hence, inclusion of the variable SM Rt−1should provide insights in the relation.

Altogether this analysis should provide a framework to explain the vari-ation in entrepreneurial activities among different countries. Furthermore it should provide insights in the spillover effects of changes in the stock markets on the total entrepreneurial activity per country.

(26)

4. Data

4.1 Overview

In order to properly investigate the factors affecting entrepreneurial activ-ity, it is necessary to have a good overview of the differences between coun-tries. Since the GEM provides one globally comparable statistic (TEA), this can be easily interpreted. Furthermore, as mentioned before, Klapper and Love (2011) showed that the financial crisis had a significant impact on the total amount of new business registrations, so it is insightful to have a look at these statistics as well.

Because Wennekers et al. (2005) found a U-shaped relation between the rate of nascent entrepreneurship and the gross national income per coun-try, insights will be provided about the variation of the TEA-rates when compared between income levels and geographic location.

In order to be able to add multiple explanatory variables to the model, various sources of data had to be used. Since data provided by these sources is only sporadically complete, an overview will be given of which countries and years are used in the model. To make sure that the full model could be estimated, various data-points had to be removed from the data.

One of the variables in the model is based on the Technology Openness. In order to have an idea of the technological developments per country, an overview of the Global Innovation Index (GII) will be given. Since this is an index, it can not be used to estimate the specific factors affecting en-trepreneurial activity, but it will give a good indication of the sources of innovative knowledge in the world.

When an overview of the entrepreneurial differences between countries is established, an overview of the behavior of the various stock markets will be given. Descriptive statistics of the stock markets per country, per region, per income-group and of the world can provide insights in the way the stock markets behaves on various levels of analysis.

Altogether these statistics should provide a general overview of the dif-ferences between entrepreneurial activity per country and the various vari-ables used in the models to analyze these differences.

4.2 TEA rates

First of all, we would like to have an idea of the spread of the TEA rates among the various countries. In Figure 4.1 the histogram can be seen of the various levels of entrepreneurial activity in the complete data-set. Here it can be seen that the median of entrepreneurial activity is around 9.1% if all countries are considered in the total observed period. Furthermore it can

(27)

be seen that there is a significant part of the data-set with TEA rates higher than 20% with some outliers around 40%.

FIGURE4.1: Histogram Total (Early-stage) Entrepreneurial Activity

To have an indication of the general development of the entrepreneurial activity over the years in between 2006-2014, Figure 4.2 shows a graph of the median of the TEA-rates over this period.

FIGURE4.2: Development of median of TEA rates

It can be seen that there is a strong increase in entrepreneurial activity over the measured period with a sharp decline in 2010. This shows that the TEA rates dropped two years after the beginning of the financial crisis in 2008. This suggests that the results of Klapper and Love (2011), who pointed out that there has been a significant negative impact of the amount of new busi-ness registrations during the crisis, are in a way connected to the drop in TEA rates. However, the TEA rates started to drop only in 2010, which is at least one year after the new business registrations started to drop. This can be explained by the fact that new business registrations are measured annually and the TEA rates describe a 3,5 year period.

(28)

If the various income regions are considered, it can be seen that there are great differences between these regions.

TABLE4.1: TEA rates per income group

INCOME Mean Std. Dev. Obs.

High income: OECD 7,34 3,64 206

High income: nonOECD 10,73 5,04 77

Upper middle income 14,23 7,50 142

Lower middle income 20,13 10,68 41

Low income 29,36 8,16 6

All 11,36 7,58 472

Looking at Table 4.1, there appears to exist a linear negative relationship be-tween the level of income and the amount of entrepreneurial activity. This would suggest a different relationship with respect to Gross National In-come (GNI) per capita than previously found by Wennekers et al. (2005). These differences could arise from the difference in measurement (TEA vs. nascent entrepreneurs), from the difference in observed time-period or be-cause the effects of income on entrepreneurial activity actually changed over time.

For a full overview of which countries are all included, the amount of observations per country and their ranking in the Global Innovation Index 2015 (GII), see Appendix A.1.

In Table 4.2 the TEA rates per region can be seen. Here it is observed that in Europe and Central Asia the rates of entrepreneurial activity are among the lowest in the world. In contrast, in Sub-Saharan Africa almost 1/4 of the working people is engaged in entrepreneurial activities.

TABLE4.2: TEA rates per region

REGION Mean Std. Dev. Obs.

East Asia & Pacific 11,89 6,73 49 Europe & Central Asia 6,94 2,51 228 Latin America & Caribbean 17,42 6,65 113 Middle East & North Africa 9,62 3,90 33

North America 10,83 2,34 12

South Asia 9,90 1,86 9

Sub-Saharan Africa 24,67 12,54 28

All 11,36 7,58 472

In order to obtain a data set in which all the variables were comparable per observation, some observations had to be removed over the years 2006, 2007, 2009 & 2011. However, due to the small scale of these countries (Tai-wan, Angola, Dem. Rep. Korea, Kosovo, Tonga & Vanatua), this should not have strong effects on the results.

(29)

4.3 Stock markets

Data on the stock market returns are gathered from the Global Financial De-velopment Database (GFDD) from the World Bank. This data set consists of numerous indicators for financial development for various countries over the past decades. In this data set, the financial situation per country is mea-sured in terms of access, depth, efficiency and various other metrics. One of the variables in this data set is stock market return, % year-on-year. This vari-able consists of the average stock market index per country and is therefore an excellent metric for the performance of the stock markets.

One of the disadvantages of this data set is the fact that it lacks quite a few observations. For the analysis in this research the total amount of observations for which the stock performance can be analyzed in the period 2006-2014 is 411 out of the original data set of 472.

Another interesting variable in this data set is the measure: Gross port-folio equity assets to GDP (%). This variable is an indication of the extent to which the stock markets relate to the overall economy per country. As ex-pected, for highly develop countries, this value is much higher than for low developed countries (approximately 20% vs. 0.04%). This variable however is left out of the analysis due to a great lack of observations.

In order to have a proper overview over the variation of the stock mar-kets over the observed period, Figure 4.3 presents a histogram of the stock markets returns per country.

FIGURE 4.3: Histogram Stock Market Returns, % year-on-year

From the figure it can be seen that on average, there is a positive annual stock market return of approximately 7% but there is a large standard devi-ation. We see many outliers, both negative and positive with some going as far as -85% or +160%. This huge spread of variance is a potential difficulty when analyzing the effect of stock markets since it is very likely that the shocks on the stock markets are much more severe than the (un)observed effect on the total economy.

If we look at the development of the stock markets over the given time period, great similarities can be seen among the various countries.

Figure 4.4 shows the stock market returns for all countries over the ob-served period. It can be seen that the stock markets strongly depend on each other (which is expected) and that the financial crisis of 2008 resulted in a worldwide collapse of the stock markets.

(30)

FIGURE 4.4: Stock Market Returns in all countries in the data set over given time period, % year-on-year

If we look at the variation of the stock returns per income group we see a remarkable division between the groups.

TABLE4.3: Average stock market return per income group

INCOME Mean Std. Dev. Obs.

High income: OECD 4,2647 20,9444 206 High income: nonOECD 9,2808 30,7560 56 Upper middle income 8,3505 30,6938 124 Lower middle income 18,1587 21,1419 25

All 7,0260 25,8387 411

Over the given period, the stock market of the income group Lower middle income greatly outperforms the other income groups. The income group Low income is dropped from this analysis due to the lack of existence and/or data of stock markets in those countries.

TABLE4.4: Average Stock Market Return per region

REGION Mean Std. Dev. Obs.

East Asia & Pacific 11,121 26,620 49 Europe & Central Asia 3,773 25,344 227 Latin America & Caribbean 12,277 28,581 75 Middle East & North Africa 0,446 23,516 22

North America 8,089 13,893 12

South Asia 23,566 18,170 8

Sub-Saharan Africa 15,006 22,321 18

All 7,026 25,839 411

When the various regions are considered (Table 4.4), there are many differ-ences to be seen as well. Clearly, the stock markets in South Asia perform best while the markets of Europe, Central Asia, Middle East and North Africa stay far behind. It is remarkable to see that the stock markets in

(31)

North America perform on average more than twice as good as the stock markets in Europe and Central Asia over the given time period, given the fact that these markets are expected to be highly dependent of each other.

4.4 Income groups

Since both the TEA rates and the return of the stock markets seem to have strong dissimilarities between various income groups, the differences in in-come may very well have a great impact on this research.

FIGURE4.5: Average Gross National Income per capita (US $)

From Figure 4.5 and 4.6 it can be seen that there exist great differences in terms of gross national income among the various income groups and re-gions.

FIGURE4.6: Average GNI per capita per region (US $)

Since the differences in average GNI per capita are very large it is decided to include a logarithmic transformation of the variable GNI in the model.

Altogether, these statistics illustrate the differences among countries in the data set with respect to their entrepreneurial activity, their stock market development and the variation of income levels.

(32)

5. Analysis and results

5.1 Baseline model

To start the analysis of the factors affecting entrepreneurial activity a base-line model is estimated. In the previous chapter it was shown that a base-linear trend can be seen in the TEA rates; hence we would like to correct for the systematic change over time. First the simplest model possible is being con-sidered:

T EAit = αi+ γt+ uit (5.1)

In this model there are individual dummies and time dummies for every observation, and we can easily observe the trend over time (see figure 5.1).

FIGURE5.1: Period fixed effects on TEA rates

Again it can be seen that there is a linear upward trend in TEA rates but a sharp drop around 2009 and 2010, which happens to be the two years after the start of the financial crisis. Because we would like to use as little dummies as possible, we substitute a fixed trend (τt) in the model which equals 1 for t = 2006, 2 for t = 2007, ... , 9 for t = 2014. We do not add a dummy variable for the crisis yet, because we would like to know if our included variables have explanatory value for this phenomena as well. The linear trend variable is included to correct for a long term unobserved trend in TEA rates.

Next, we consider the full model as described in chapter 3.2 with pooled effects.

T EAit= α + δτt+ βXit+ uit (5.2) Table 5.1 (p. 26) shows the results of various pooled regressions.

The first regression is a standard panel data OLS. The other two regressions are generalized least squares, where the weighting matrix is based on the

(33)

TABLE5.1: Pooled regressions, s.e. in brackets

Pooled OLS Pooled GLS PGLS G2S

Constant -22,8 -9,216 13,215*** (21,516) (13,311) (3,066) Trend 0,611*** 0,474*** 0,486*** (0,113) (0,058) (0,051) GDP growth 0,228*** 0,11** 0,101*** (0,084) (0,044) (0,025) Employment 0,241*** 0,262*** 0,245*** (0,049) (0,033) (0,022) Log(GNI) 11,946 5,055 -6,702*** (9,864) (5,267) (0,41) Log(GNI)2 -2,024 -1,404** -(1,238) (0,604)

-Stock market return -0,004 -0,004

-(0,01) (0,006) -Internet 0,038 0,055*** 0,057*** (0,027) (0,006) (0,008) Uni-Ind 0,079* 0,044** 0,035** (0,044) (0,018) (0,016) High-tech export -0,038 0,032* 0,038*** (0,024) (0,016) (0,012) Firm-tech absorb -0,103* -0,03 -(0,052) (0,026) -Innovation -0,164*** -0,166*** -0,162*** (0,036) (0,019) (0,017) Cost of startup 0,002 -0,031 -(0,032) (0,024) -Days to startup 0,046*** 0,044*** 0,045*** (0,012) (0,005) (0,006) Gov-regulations -0,035 -0,053*** -0,037** (0,034) (0,017) (0,016)

Total tax rate -0,006 0,018 0,02*

(0,016) (0,013) (0,011) Population growth -0,208 -0,084 -(0,205) (0,126) -International tourism -0,037 -0,006 -(0,029) (0,013) -Friend Entrepreneur 0,161*** 0,096*** 0,092*** (0,025) (0,015) (0,012)

High status entrepreneur 0,003 0,001

-(0,025) (0,016) -(Ex-)Communist -5,666*** -4,252*** -3,545*** (0,817) (0,524) (0,257) Adjusted R2 0,56 0,73 0,72 *: p < 0.10. **: p < 0.05. ***: p < 0.01

(34)

cross-section weights. In the last regression a general-to-specific procedure is applied until only significant variables remained in the model.

From these regressions various interesting results appear. First of all, in contrast to the research done by Wennekers et al. (2005), no significant U-shaped relation can be found between entrepreneurial activity and the gross national income. There appears to be a linear negative relation be-tween level of income and TEA rates. This could be explained by two rea-sons:

First of all, in this research a logarithmic transformation of the variable GNI is used. This changes the way the relation between the two variables is analyzed an therefore changes the interpretation.

Furthermore there is the fact that the TEA rates are a slightly different metric then the nascent rate of entrepreneurship (which Wennekers et al. (2005) used). In the TEA rates there is by definition a correction for the survival rate of start-ups. Even with these differences we could still have found proof for the U-shaped relation, but without the use of fixed effects this cannot be proven.

Next, the variables population growth and international tourism, which were both hypothesized to affect the TEA rates due to their demographic influence, do not appear to have a significant effect.

But most importantly, in neither of the models, the stock market return appears to have a direct significant effect on the TEA rates. However, this does not yet mean that the stock markets cannot have an indirect on the TEA rates via the hypothesized latent variable economic situation.

When a dummy variable for the year 2010 is added to the model, or one of the dummy variables crisis1 or crisis2 (as defined in Chapter 3.2), they have a highly significant negative result. This indicates that the large drop in TEA rates after the financial crisis of 2008 is not yet explained by the other variables included in the model. The regressions in Table 5.1 were performed in order to obtain an upper bound of explanatory value of the variables and therefore the results of the regressions with the dummy vari-able crisis are not shown.

However, due to the many different factors affecting entrepreneurial ac-tivity, there can be the problem of unobserved country-specific factors af-fecting the TEA rates, which are now not taken into account. In theory these effects could be fixed or random. However, with the Hausman test (Haus-man, 1978), the null hypothesis that a random effect model is appropriate is clearly rejected (Table 5.2). Hence, a fixed effect model is presented in Table 5.3.

TABLE5.2: Hausman test for random effects

Test Summary Chi-Sq. Statistic Chi-Sq. d.f. Prob.

Cross-section random 49,13404 19 0,0002

By using cross-country fixed effects, we automatically correct the model for unobserved individual characteristics per country that were not yet cap-tured by the variables in the model. In doing so, we overcome the problem of unobserved heterogeneity in the error term which could be correlated with the explanatory variables.

(35)

In Table 5.3 (p. 27) the results of the fixed effect regressions are shown. We immediately notice that the adjusted R2 of the various regressions is much higher than before (0.85 vs 0.73). Furthermore, it is interesting to see that the variable Log(GN I) and Log(GN I)2_{are both highly significant} now that we corrected for unobserved heterogeneity. This means that the same U-shaped relation is found as Wennekers et al. (2005) did before, even though the variable is logarithmic transformed. This means that a small increase in GNI per capita results in less entrepreneurial activity but a very high increase results in more entrepreneurial activity.

When applying the general-to-specific procedure to the regressions, we can easily see that many of the variables do not have a significant effect on the TEA rates. This is remarkable since all of the variables are included be-cause the underlying theory provided us with support for inclusion. The fact that these variables are not significant is most likely due to multicollinear-ity. Many of the variables are hypothesized to belong to the same under-lying factor and the individual effect cannot be estimated properly because of the correlation between the variables.

(36)

TABLE5.3: Fixed effects regressions, s.e. in brackets

Full model Full model G2S G2S

no crisis with crisis no crisis with crisis

Constant 89,47** 85,572** 101,579*** 93,232*** (34,967) (36,628) (28,098) (25,515) Trend 0,351*** 0,260*** 0,248*** 0,124 (0,055) (0,062) (0,066) (0,079) Crisis - -0,449*** - -0,543*** - (0,148) - (0,143) GDP growth 0,046 -0,003 - -(0,056) (0,059) - -Employment 0,040 0,040 - -(0,034) (0,036) - -Log(GNI) -50,649*** -50,628*** -55,689*** -54,017*** (13,535) (14,041) (12,527) (10,667) Log(GNI)2 _5,959*** _6,143*** _7,043*** _6,897*** (1,545) (1,615) (1,411) (1,234)

Stock market return -0,011 -0,013 - -0,011*

(0,009) (0,008) - (0,006) Internet 0,041*** 0,037*** 0,054*** 0,048*** (0,009) (0,009) (0,010) (0,011) Uni-Ind 0,145** 0,161*** - 0,151** (0,058) (0,059) - (0,067) High-tech export 0,057** 0,042 - -(0,028) (0,036) - -Firm-tech absorb -0,027 0,000 - -(0,059) (0,06) - -Innovation 0,088* 0,051 0,091* -(0,046) (0,038) (0,050) -Cost of startup 0,024 0,012 - -(0,070) (0,071) - -Days to startup 0,041** 0,048** - 0,049** (0,017) (0,019) - (0,021) Gov-regulations 0,031 0,042 - -(0,042) (0,044) -

-Total tax rate 0,101*** 0,101*** 0,096*** 0,101**

(0,038) (0,038) (0,036) (0,041) Population growth 0,395* 0,330 - -(0,225) (0,225) - -International tourism -0,188 -0,157 - -(0,148) (0,160) - -Friend Entrepreneur 0,081*** 0,098*** 0,071*** 0,087*** (0,020) (0,022) (0,027) (0,023)

High status entrepreneur -0,037 -0,041 -

-(0,026) (0,026) -

-Adjusted R2 _0,85 _0,85 _0,84 _0,85

*: p < 0.10, **: p < 0.05, ***: p < 0.01 .

The effect of a change in the stock market on the entrepreneurial activity : a cross-country analysis