The impact of the enlargement of the European Union in 2004 and 2007 on self-employment in the United Kingdom

(1)

The impact of the enlargement of the European Union

in 2004 and 2007 on self-employment in the United Kingdom

Thesis Master of Science Economics – International Economics and Globalization

University of Amsterdam Faculty of Economics and Business

Aglaia Clarissa Naumann

Ruyschstraat 8H, 1091CB Amsterdam

Tel.: +31 617711166

aglaianaumann@googlemail.com

Studentnumber: 10604537

27.08.2014, Supervisors: Maja

Micevska Scharf and Dirk Veestraeten

(2)

2

1. Introduction

As described by the Treaty of the Functioning of the European Union, free movement of workers is a fundamental principle of the European Union. Free movement includes, amongst other things, the right to search for a job in another EU country, work there without the necessity of a work permit and reside in that country for that purpose. Migration flows across the European Union are mostly due to the search for employment. Since the enlargement of the European Union in 2004 and 2007 and due to transitional restrictions, the European Union has experienced a shift in migration flows. Traditionally Germany has been the preferred immigration country, but since the German government chose to keep its labour markets restricted for immigrants, migration is increasingly directed towards other countries in the EU. Especially the United Kingdom and Ireland experienced increased inflows from migrants from the new member states.

Although it becomes easier for EU citizens to migrate and search for work, actual labour mobility remains rather low. According to the Eurobarometer from 2010, only about two percent of EU citizens live in another EU member state and only ten percent have ever lived and worked abroad at some point in their lives. Despite this low numbers of immigrants, it seems that there are still many false pieces of information and reservations found in the public discussion; especially about and towards immigrants from the EU-8, Romania and Bulgaria. In spite of the anticipated, and often not realised, negative impact on the labour market in the destination country, immigration inflow implies also many positive externalities. Immigrants bring new skills and human capital into the country; they close gaps between labour demand and supply and often have a rejuvenating effect on the host country’s society. Also, immigrants in almost all OECD countries are more entrepreneurial active than their native counterparts. As the OECD International Migration Outlook 2011 reports, on average 12.6% of migrants of working age in OECD countries were self-employed in non-agricultural activities, compared with 12.0% among natives in 2007/2008. In most countries, the difference between migrant entrepreneurship and native entrepreneurship is larger. Countries where migrants are especially entrepreneurial more active than natives are among others Denmark (10.0% vs. 7.0%), France (10.6% vs. 8.0%) and Belgium (14.7% vs. 12.0%) and most extreme cases in this context being Poland (29.4% vs. 11.2%), Czech Republic (20.3% vs. 15.1%) and the Slovak Republic (23.6% vs. 13.0%). Countries where natives are traditionally more entrepreneurial active are for example Germany, Austria, Ireland and South European countries, like Greece, Italy and Spain. The Netherlands is the only country among the OECD states, where the share of native entrepreneurs and migrant entrepreneurs was equal in 2007/2008 (10.7%), after experiencing a rather large increase in migrant entrepreneurship after 1998.

Entrepreneurship is perceived as beneficial for an economy and a society for many reasons. It is crucial for economic growth and innovation through creation of employment, increased competition and the introduction of new products and production methods. There are two main factors that influence entrepreneurship in the destination country in connection with the EU enlargement: migration and trade. Other channels of influence include a diversification of skills and knowledge and increased innovation.

Migrants and migrant entrepreneurs contribute to entrepreneurship and the economy of the destination country as a whole by bringing in knowledge and skills from different cultures, which has positive effects on international trade and innovation. Because global trade still involves significant

(5)

5 costs, immigrants can decrease those costs since they “have a good knowledge about the business culture, politics, religion and language of their former home countries” (see Hatzigeorgiou, 2010, p. 273). Also, they maintain a personal network which they can use to foster trade relations. This is particularly important when institutions are weak and informal channels for information and trade are central. In some cases, like in the United States, immigrants play a particularly important role for innovation. As discussed by Hunt (2010), skilled immigrants in the United States do better than natives in terms of wages, patenting and publishing. This is mainly due to the fact, that many immigrants first entered the United States on student visas or temporary work permits, which encourages permanent immigration later on. These are main examples of the manifold effects of immigrants and migrant entrepreneurs on the economy in the destination country. The mentioned positive effects are not limited to high-skilled and high-value activities, but are observed in one way or another in all entrepreneurial activities.

Because of the positive effects of entrepreneurship on the economy as a whole and increasing migration flows across the European Union followed by the EU enlargements, it is interesting to look further into recent trends of entrepreneurship in the core member states of the EU. The term “entrepreneur” and “migrant entrepreneur” or “ethnic entrepreneur” is not a clear term and a meaningful empirical analysis depends on a comprehensible definition. In my analysis I will adopt a widely used definition from the European Labour Force Survey: Entrepreneurs are persons who work in their own businesses, professional practices or farms in order to make profits or are in the process of setting up a business. For empirical work, the number of self-employed is representative for all entrepreneurs, since this is the most suitable approach, also with respect to data availability. Migrant entrepreneurs are then those business owners who were born in a foreign country. In my analysis tough, I will not differentiate between migrant and native entrepreneurs, but instead look at the overall level of self-employed in the UK.

Among others, the British government chose to drop all formal labour market restrictions, such as the need of work permits for immigrants from the EU-10 countries. Transitional restrictions for migrants from the EU-2 were applied until the end of 2013 and free access to the labour market was granted on January 1, 2014. Many studies have been conducted to examine migrant entrepreneurship and macro-economic effects in the host and the source country. The focus of my work will be the influence of the opening of the EU on the share of entrepreneurs in total employment in the United Kingdom. My hypothesis is that the enlargement of the EU has a positive effect on self-employment in the UK. This is mainly due to enhanced migration of mostly working-age people and the difference in entrepreneurial behaviour. In connection with an inflow of new skills and knowledge, the overall economic environment and trade conditions improve, which also has a positive impact on entrepreneurship. In my research I will therefore explore the general impact of the opening of the EU in 2004 and 2007 on entrepreneurship in the United Kingdom.

The paper is structured as follows: first, I will give an overview of existing literature about immigration and migrant entrepreneurship. Then I will move on to my empirical work by describing first the theory of influences on entrepreneurship and the model selection process, followed by the description of the used data. The fourth part will include results of the empirical analysis and finally the paper will be summed up by a conclusion.

(6)

6

2. Immigration and Entrepreneurship in the United Kingdom

a. Migration across Europe and to the United Kingdom

To give an impression of the research about migration in Europe, I have chosen two studies that provide an overview of the current discussion on this topic.

The special Eurobarometer of 2010, is a comprehensive study that aims to assess labour market mobility and motivation and disincentives for individuals to move across borders. It also covers the opinion of individuals towards cross-border movement, their experience with working, studying or living abroad and also their plans to move abroad in the future. Furthermore, it asks the participants about their knowledge of the different ways to find work in the destination country. The main results of this survey are that most Europeans have a positive attitude towards free movement in Europe and about a third believes that chances of employment are better abroad than in their home country, although there are large differences in this perception on country level. Especially citizens of the EU-10+2 countries believe their chances of employment abroad are higher. Citizens of EU-10+2 are in general more likely to consider moving out of economic reasons than EU-15 citizens, where lifestyle and cultural factors dominate in the decision-making process.

Another study that complements these results is undertaken by Holland et al (2011). Here, migration between 2004 and 2009 was examined to quantify the share of population flows that occurred due to the EU enlargement. This study also attempts to assess the macro-economic impact of migration in the host countries. Key findings of the paper are, that since the 2004 enlargement, about 1.8% of the EU-8 population has moved to the EU-15 countries, while since 2007 4.1% of the EU-2 population moved to the EU-15. Since 2004 Germany has not been the preferred destination country, rather have been Ireland, the United Kingdom and Luxembourg for migrants from the EU-8 and Spain and Italy for migrants from EU-2, due to the mentioned transitional restrictions. This and many other studies show that mostly young, educated people search employment in the EU-15 countries, which has a rejuvenating effect on the host country’s society and fills gaps between labour supply and demand. It is also mentioned here, that migrants are generally more entrepreneurial active than natives.

b. Entrepreneurship in the United Kingdom

i.

Facts about Entrepreneurship and Employment

One study that gives a broad picture about the phenomenon of native and migrant entrepreneurship is the OECD International Migration Outlook of 2011. It covers entrepreneurship in different dimension, such as age, gender, education, business sectors and region of origin.

The share of self-employed in total employment is higher for foreign-born than for native-born in almost all OECD-countries. Exceptions of this observation are nine out of the twenty-three OECD countries, including Germany, France, Greece and Italy. In Poland, the Slovak Republic and the Czech Republic migrants are far more entrepreneurially active compared with natives. For the rest of the countries, the numbers differ only slightly. Figure 2.1 shows the share of self-employed in total non-agricultural employment for all OECD countries.

(7)

7

Fig. 2.1: Source: EU Labour Force Survey, OECD International Migration Outlook 2011

In the following, I will only focus on the results for the United Kingdom. The share of self-employed in total non-agricultural employment is higher for foreign-born than for native-born throughout the years from 1998 to 2008. Table 2.1 shows the evolution of the share of self-employment by place of birth. It seems that the shares of native- and foreign-born self-employed converge slightly over time, where the share of foreign-born entrepreneurs decreases and the share of native self-employed increases. After 2000, the share of foreign-born self-employed remained rather stable around 14.2%, whereas the share of native-born entrepreneurs increases steadily since 1998. This seems like a rather small increase and insignificant, but compared with the OECD average and the development in other European countries, the situation in the United Kingdom seems unique. In most countries, the share of migrant and native self-employed follow the same trend. Instead of convergence, we can observe a rather parallel development.

Native-born Foreign-born

1998-2000 10.8 15.5

2001-2003 11.0 14.2

2004-2006 11.6 14.1

2007-2008 12.1 14.2

Table 2.1: Evolution of share of self-employment by place of birth; Source: Mestres (2010)

Entrepreneurs naturally also contribute to employment creation. A bigger share of migrants employs not only themselves but also others; 73.3% of migrant entrepreneurs employ only themselves compared to 77.8% of natives. Furthermore, each self-employed immigrant creates on average between 1.5 to 2.6 jobs, compared with 1.3 to 2.1 additional jobs created by each native entrepreneur. Still, migrant entrepreneurs are less successful and the flows in and out of

self-0 5 10 15 20 25 30 35 Au stra lia Au stria Be lgi u m Cze ch Rep u b lic De n m ar k Fran ce G erm an y G re ece H u n gary Irel an d Is re al It aly Lu xe m b o u rg N eth erlan d s N o rw ay Po lan d Portu gal Slov ak Rep u b lic Sp ain Sw ed en Sw itz erlan d U n ited Ki n gd o m U n ited Sta tes OE CD av era ge

Share of self-employed in total non-agricultural

employment in 2007-2008

Foreign-born Native-born

(8)

8 employment are bigger than for natives. Overall, migrant entrepreneurs have similar underlying age distribution as natives, although they are slightly younger. As the OECD International Migration Outlook and other studies emphasize, immigrants work in many different sectors, not only in ethnic businesses that they are usually associated with. Although a high share of migrants own traditional ethnic businesses, the majority of them are engaged outside this sector. An increasing number of immigrants are starting a business in professional, scientific and technical activities and other high-value activities.

Although the flow of immigrants from the new EU member states to the United Kingdom has risen since the enlargement, only a small percentage of these new immigrants are self-employed, as reported by Drinkwater (2010). While about 32% of Polish immigrants were self-employed in the period before the enlargement, only 4% choose entrepreneurship after 2003. This is mostly due to the fact, that before the opening of the labour market, it was easier for entrepreneurs to enter the United Kingdom than it was for immigrants searching paid work.

ii.

Determinants of Entrepreneurship for Natives and Migrants

Two theoretical approaches can be taken to assess the different determinants of entrepreneurship: examination of the individual decision-making process and the description of macro-economic circumstances.

Individual determinants

OECD (2011), as well as Mestres (2010), provide an empirical analysis on the factors behind a migrant’s individual entrepreneurial decision. A logit model is created to determine the probability of natives and immigrants of being self-employed. Socio-demographic factors, such as age, gender and education are all significant determinants. Younger individuals are generally less likely to start a business, as are women. The variable ‘foreign-born’, which includes EU citizens, as well as immigrants from non-EU countries, is significant and positive. This result complements the before mentioned developments. The years spent in the host country play a significant role in the decision making and the probability of being an entrepreneur increases with years of residence in the host country. In general, immigrants choose self-employment as a way to avoid discrimination in paid employment, mostly related to cultural differences and language, or to move out from a low-wage job. Therefore, they find self-employment as their only option given the lack of alternatives and recognize a positive wage premium connected with entrepreneurship. Another important factor for the individual decision of being an entrepreneur is the access to financial support. Immigrants usually experience restricted availability of financial support like credit from banks and rely more heavily on informal networks and family. This and complex administrative requirements decrease the probability of becoming an entrepreneur and are especially discriminating for immigrants.

Macro-economic determinants

Macro-economic factors involved in creating entrepreneurship serve, to explain the overall level of entrepreneurship. Grilo and Thurik (2004) describe the demand and supply side of entrepreneurship. In other works this is referred to as push- and pull-factors.

The demand side is considered the product market perspective. It represents the opportunities for entrepreneurship and involves costumers and firms. Diversity in demand for products increases opportunities for new businesses. Therefore, the demand side is determined by economic and

(9)

9 technological development, globalization and the industrial structure of the economy. A high level of technological development encourages entrepreneurship, as does globalization. These factors decrease marginal costs and offer opportunities for exploiting scale. Globalization also increases diversity in demand for goods. The influence of economic growth is ambiguous, since it raises wages and therefore the opportunity costs of entrepreneurship, but it also increases demand on the consumer side. The industrial structure can be characterized by the number of services firms, which is usually large in Western countries. Those businesses require low initial capital and are usually small, therefore minimizing entry barriers. Clusters of production and outsourcing of services also foster entrepreneurship.

The supply side of entrepreneurship, also considered the labour market perspective, is affected by the demographic characteristics of population, the resources and skills of individuals and the general attitude towards entrepreneurship. The individual decision-making process mentioned before is intertwined with this side. Important demographic features are the size and composition of population, which include immigration, population growth, density and rate of urbanization. Other factors determining the supply side are age structure and the share of women in the population. The wage level is included on the supply side since it can be interpreted as the opportunity costs faced by a potential entrepreneur. Population growth has a positive influence on entrepreneurship, since it is accompanied by decreasing wages and an increase in demand. The effect in population density and urbanization is ambiguous. Infrastructure in urban areas is usually better, but it also raises opportunities to take advantage of economies of scale, which has a negative effect on the creation of new small businesses. As mentioned before, younger people have a lower probability to start a business, but on the macro-level, an ageing society has a negative effect on entrepreneurship. If more women are participating in the labour force, the number of female entrepreneurs also rises. A high wage level has also an ambiguous effect, as does the unemployment rate.

In general, individuals weigh up the expected return from being an entrepreneur against the rewards of taking up paid work. Everybody has a unique risk-profile, which determines their decision. Migrants are considered to have a lower risk-aversion, since they have already made the decision to move abroad before, therefore they have a higher propensity to become an entrepreneur.

My research will contribute to the existing literature by examining the effects of a larger EU, an open European labour market and increased migration on the level of entrepreneurs in the UK. Most studies focus solely on the phenomenon of migrant entrepreneurs. I wanted to show that the opening of the EU is beneficial for the general level of self-employment in the UK.

(10)

10

3. Model and Data

a. Model

i.

Self-employment and its influence factors

In order to evaluate the effect of the EU enlargement on self-employment in the United Kingdom, I created a small model for a regression analysis. Since the number of observations in my dataset is limited, it is crucial to reduce the number of variables to a minimum and at the same time, still include all relevant factors. This is especially important with regards to multicollinearity. My approach was therefore to extract the main influence factors from the literature and theory on entrepreneurship and find explaining variables for each of them. I could extract four fields of influence combined from individual decision-making and the demand and supply side of entrepreneurship: population characteristics, technological development, globalization and industrial structure. Each field can be defined by various variables. To capture the effect of the EU enlargement I created two dummy variables, one for the enlargement in 2004 and another one for the enlargement in 2007. The dependent variable is the share of self-employed in total non-agricultural employment. This refers to the population of age sixteen and older. Instead of taking the employment of age 16-64, I chose the mentioned broader group, since I believe this will yield a more complete picture of self-employment. Many self-employed might continue with their business beyond the age of 64 and employed people might also continue working as self-employed after retirement.

The most important population characteristics are the median age, the degree of urbanization and migration. According to Mestres (2010) most entrepreneurs in all OECD countries are between the age of 35 and 44. This holds for natives, as well as for immigrants, although migrant entrepreneurs “are on average slightly younger than their native counterparts”1. Young people who just step into employment are less likely to choose self-employment. In general, self-employed are older than wage employees. According to Grilo and Thurik (2004) a high degree of urbanization has an ambiguous effect on entrepreneurship. It provides a dense network of knowledge and a good infrastructure, which is beneficial “for business start-up and development”2. But big urban areas can also give opportunity to exploit economies of scale, which is negative for small entrepreneurs. The effect of migration cannot be ignored in my analysis. As Mestres (2010) shows, immigrants are often times entrepreneurially more active than natives in OECD countries, which is also true for the UK. An increased inflow of immigrants, as it was the case for the UK since the opening of the EU in 2004, brings new skills and knowledge into a country and possibly lowers the income level. I will explain the expected effect of increased immigration below in connection with the opening of the EU. For my analysis, I included the lagged value of migration inflows instead of the migration inflow of the current year. Since it always takes some time to settle in a new country and pick up work, I believe this is a more accurate measure of the effect of immigration. Also, possible multicollinearity problems are eliminated if I include only one measure of immigration.

Technological development can be captured by the expenditures on research and development of an economy. A highly developed economy is supposed to be beneficial for possible entrepreneurs, since certain procedures, like information transfers, become easier and less cost intensive and new innovative services and products can be offered and the demand for entrepreneurs also increases,

1

See Mestres J. (2010) in OECD; Open for Business: Migrant Entrepreneurship in OECD countries, p. 29 2_{See Grilo I. and Thurik R. (2004), p. 7}

(11)

11 i.e. internet start-ups (Grilo and Thurik, 2004). Of course, the importance of the internet is a rather recent development and might not seem significant for my timeline. Nevertheless, for completeness I included the expenditures on research and development (R&D) as a share of GDP in my analysis. Globalization is a complex phenomenon and can be explained through multiple variables. According to the OECD Handbook on Economic Globalisation Indicators (2005), the degree of globalization can be captured through foreign direct investment flows (FDI), economic activity of multinational enterprises, international dissemination of technology and the globalisation of trade. Due to data availability, I chose the FDI balance to capture the effect of globalization and left out the trade balance to avoid possible multicollinearity. The more open a country is to the rest of the world, the more chances unfold for entrepreneurs, for example through a larger trade area and increased investment opportunities. Therefore, a high degree of globalization should show a positive effect on the share of self-employed in total employment.

The industrial structure of an economy is also determined by its technological development. A highly developed country will have a bigger service sector and is less dependent on manufacturing and agricultural industries than developing countries. Starting a service business is easier since the entry barriers are lower, for example a lower initial capital is needed, facilities like big productions sites are not necessary and licences are also cheaper (Grilo and Thurik, 2004). Therefore a big service sector, which in my analysis is represented by the share of service jobs, should have a positive effect on self-employment.

For completeness, I decided to include GDP growth in my analysis, although Grilo and Thurik (2004) describe an ambiguous effect of economic development on self-employment. A positive economic development can have a positive effect on self-employment, since possible entrepreneurs are more confident and expect their initial investment to be paid back. Also, a higher disposable income results in an increased demand for products and services, which gives incentives for the creation of new industries. Another effect of economic development might be the increased inflow of entrepreneurial active migrants. As the Eurobarometer (2010) describes, migrants from the new EU member states are more willing to leave their home country for economic reasons than citizens of the EU-15 states. If they expect the probability to be employed in the destination country or to have a successful business to be high, they are more willing to move there and be entrepreneurial active. On the other hand, rising real wages and a better social security system give disincentives to be self-employed, since the opportunity costs of self-employment also raise. The effect of economic development therefore needs to be examined with respect to the available data on expectations and behaviour of individuals.

The enlargement of the EU may have various influences on the level of self-employment in the UK. A greater EU can have positive effects on trade and employment. The inflow of workers increases the labour supply and has a downward pressure on wages, especially in the low-skilled sector. This has two effects on the decision of becoming an entrepreneur. First, it decreases opportunity costs for workers and gives an incentive to become entrepreneurial active (Grilo and Thurik, 2004). Second, potential employees can be hired at lower wages. On the other hand, lower wages mean also lower disposable income and could decrease demand for new products. This effect might be offset by the increased number of customers throughout Europe and hence a larger market. Through the opening of the EU, a larger trade area is created and more possibilities for self-employed evolve. The larger area is also economically connected, which makes for example investments easier. Still, most self-employed own a rather small business instead of big enterprises that engage in international trade. As Mestres (2010) shows, most self-employed, natives and immigrants alike are active in the

(12)

12 wholesale and construction sector. And in the UK, most entrepreneurs only employ themselves, where about 73% of foreign-born self-employed and 78% of native-born entrepreneurs work alone3. The effect of the larger economic area, which is partly captured by the globalization variable, will be one interesting result of my analysis.

Since I have only a limited timeline at hand, I needed to leave out some explanatory variables. These variables are, among others, the labour participation rate of women, the income level and unemployment. According to Grilo and Thurik (2004), all these variables have an ambiguous effect. The effects of these variables need to be examined in a different context and larger sample.

To summarize, the model will then look as follows: the share of self-employed of total employment (16 and older) is supposed to be determined by the median age, the share of urban population, the share of service jobs of total workforce jobs, the expenditures on R&D as a share of GDP, the FDI balance in British pound, the lagged value of migration inflows, GDP growth and the two EU enlargement dummies.

ii. Model selection

As aforementioned, the problem of multicollinearity was also an issue in my analysis. Multicollinearity in an ordinary least squares (OLS) regression leads to incorrect coefficient and standard error estimates. A valid hypothesis testing is not possible. The overall prediction power of the model is not affected.

In order to eliminate the multicollinearity problem, I wanted to choose variables that are highly correlated with the dependent variable, but show a low correlation with the other regressors. My model specification took place in two stages: first I looked at the correlation between the variables to pick out important variables, and second, I performed regression and examined the variance inflation factors (VIF) to check for multicollinearity. Since I want to examine the effect of the EU enlargements, I will keep those two variables in my model. It seems that especially the median age and the degree of urbanization are highly correlated with almost all other variables. This is not surprising, since those two variables are population characteristics that result from other economic and social factors and are by nature rather endogenous. Hence, I decided to leave out these variables from the model. I expect the share of service jobs, which shows the size of the service sector, to be very important for the explanation of self-employment and it also shows the highest correlation with the dependent variable. Therefore, I decided to keep the share of service jobs of total workforce jobs in my model. After I have already left out the median age and the share of urban population, migration inflow and share of service jobs, show a very high correlation. The variable for migration inflow is also highly correlated with the opening of the EU in 2004, which is of course not surprising. Since migration inflow shows a high correlation with two variables that I certainly want to include in my model, I decided to drop the migration inflows as well. This leaves me with a smaller set of explanatory variables: the share of service jobs in total workforce jobs, R&D expenditures, the FDI balance and the two EU dummies. The next step is then, to check the VIF for severe multicollinearity. A VIF value above 10 shows a severe multicollinearity problem. The VIF output for the regression with the remaining variables shows a multicollinearity issue with regards to the share of service jobs and a VIF close to 10 for R&D expenditures. In combination with the correlations and my decision to keep the

3

See Table 1.3 in Mestres (2010) in OECD; Open for Business: Migrant Entrepreneurship in OECD countries, p. 38

(13)

13 share of service jobs, I decided to drop expenditures on R&D from my model as well. This leaves me with the following base model, where multicollinearity is eliminated:

After I have decided on a base model without multicollinearity, I have to make sure that the other assumptions for the OLS regression hold. This can be problematic in a small sample like mine. The assumptions of the OLS are: a correct specification, according to economic theory and including all relevant variables, linear relationship between dependent and independent variable, exogeneity and no reverse causaloty, homoscedasticity and normality of error terms. I will check for each assumption separately.

I chose all independent variables according to theory known to me and described the expected effect of each explanatory variable above. Still, it is interesting to check formally for correct specifications in Stata. This can be done through the link test and the test for omitted variables. If a model is correctly specified, it should not be possible to find any other explanatory variables that are significant, except by chance. This forms the basic idea of the link test which checks for the so-called link error that occurs if the dependent variable needs a transformation or a “link” function to properly connect to the independent variables4. In the link test, _hat = is calculated, as well as _hatsq . The model is then refit with these new variables. A correct specification of the base model depends on the significance of _hatsq. If this variable is significant, there are still significant regressors not included in the model; _hat should always be significant, since this captures the model with the parameter estimates. In my case, according to the link test, _hatsq is significant, which means that the model is not complete. To double check, I performed the test for omitted variables, which is a regression specification error test and is similar to the link test. It also creates new variables and checks if any of these are significant. For my model, the null-hypothesis that there are no omitted variables cannot be rejected. This leaves me with an ambiguous result for the model specification issue. The trade-off that I was facing in my work was choosing between including all explanatory variables and reducing multicollinearity. I chose to focus mainly on the multicollinearity problem because of the small sample problem and in order to get meaningful standard errors. Therefore, the model that I am working with cannot be complete and the regression will not yield a full explanation of self-employment. I merely aimed at giving a first impression about the effect of the EU enlargements on self-employment.

In order to check linearity of the model, scatter plots can be used to detect possible problems. In a regression model with multiple predictors, it is useful to plot the residuals against each of the regressors. There should be no clear pattern visual. When plotting the residuals against the share of service jobs and the GDP growth, no clear non-linearity can be observed, although there is a hint for non-linearity for the share of service jobs. The plot for residuals against the FDI balance shows a more severe problem with linearity. In this case a quadratic form, instead of a linear relationship, seems more appropriate. To fix this problem, I used a square root transformation for the variable describing the FDI balance, which leaves me with one missing value. In 1991 the FDI balance was negative and since the square root is not defined for negative values, the value for this year is

(14)

14 missing after the transformation. The rest of the data can easily be transformed. Also, after this transformation, the estimated coefficients cannot be interpreted as straight forward as before. Since I am not primarily interested in the exact effect of a unit change in FDI on the level of self-employed, but rather in an overall well-fitting model, this transformation seems appropriate to me. The transformation seems to fix this problem, as the relationship is now fairly close to linearity. I did not perform a transformation for the share of service jobs, since the deviation is not as large and also not as clear as was the case for the FDI balance. One result of this decision could be that the effect of the share of service jobs is biased, that is the coefficient is under- or overestimated. As I have mentioned, the exact value of the coefficients is not my main concern and hence I decided to keep the share of service firms in the original form.

Next, I discuss possible endogeneity or reverse causality and homoscedasticity and normality of error terms, which forms requirements for the independent and identical distribution of observations. The problem of reverse causality occurs, if the dependent variable influences one or more independent variables. If this is at hand, the error terms and the explanatory variables are correlated with each other and the estimation is biased. A reverse causality might occur between the share of employed and the share of service jobs in total workforce jobs, since a rising number of self-employed might also cause the service sector to grow. After estimating the above described model with the transformation, I checked for correlation between the residuals and the explanatory variables. It seems that there is no correlation between the residuals and the share of service jobs, the square root of the FDI balance, the GDP growth and the EU enlargement in 2004 and 2007. It can be concluded that there is no endogeneity problem in my model.

Homoscedasticity can easily be tested when looking at the scatter plot of the residuals against the fitted values. In this graph, no pattern should be observable, which means that the variance of the error terms does not change according to a specific trend. Again, to double check I also conducted a non-graphical test for homoscedasticity: the White’s test and the Breusch-Pagan test. Both test the null-hypothesis that the variance of the residuals is homogeneous and no heteroscedasticity exists. It is always useful to include both graphical and non-graphical tests, since the statistical tests are rather sensitive to model assumptions such as normality. All three tests and plots show no signs for heteroscedasticity, since the plot shows no clear trend or pattern and for both tests the null-hypothesis cannot be rejected.

Normality of error terms can also be tested through various graphical and non-graphical checks. A graphical analysis is to compare the estimated distribution of the residuals with the normal distribution or compare the quantiles of the residuals against the quantiles of the normal distribution. There exist various statistical tests in Stata that can be conducted when testing for normality: the Shapiro-Wilk test and the Shapiro-Francia test, the Kolmogorov-Smirnov test and the skewness and kurtosis test for normality. All of them test the null-hypothesis, that the distribution of a selected variable, in this case the residuals, is normally distributed. In Stata, the Kolmogorov-Smirnov test can formally be used for testing normality, but is described as not a particularly powerful test for this purpose5. The graphical tests show a slight deviation from the normal distribution, especially in the middle range. If these deviations are significant needs to be checked with the statistical tests. According to every single statistical test, the null-hypothesis of normality cannot be rejected.

(15)

15 After checking for all assumptions for the OLS and fixing possible problems, the model that I will conduct my regression with, will be:

b. Data

i.

Overview

The data for my analysis were all taken from the database of the British Office for National Statistics (ONS) is generally freely available. Each variable can easily be selected from the data selector online. The share of self-employed in total employment is generated from data from the British Labour Force Survey (LFS). Various distinctions can be made with this measure. I chose the broadest available measure of self-employment, which includes self-employed that work full- and part-time, male and female. Since I focus on the general level of self-employment and not specifically on a certain group of entrepreneurs, I believe that this is the most appropriate measure. Data for the number of self-employed is available from 1984 to 2013, which leaves me with thirty observations. There are no missing values in this series and each annual value is calculated as a four quarter average and is seasonally adjusted. As a reference to calculate the share of self-employed in total employment I used the total number of people in employment. This is also part of the LFS and annual values are calculated in the same way as the number of self-employed. The share of service jobs in total workforce jobs is calculated from the number of service jobs and the total number of workforce jobs, which is included in different labour market statistics. In the available time line are no missing values. Data on foreign direct investment is recorded by the ONS in their balance of payment statistics. As a summarizing variable of both inflows and outflows, I chose the total balance of FDI in British pound and there is no missing value in this time line. Data on GDP growth is recorded in the quarterly national accounts and available via the website of the ONS. The EU opening in 2004 and 2007 is depicted by dummy variables.

ii.

Details on data sources

Since 1984, the British Labour Force Survey is conducted every year and covers all fields of employment. It consists of quarterly surveys and a boost survey in the spring quarter, which complements the dataset. Every quarter around 41,000 respondents take part in the survey. The target population is based on the resident population in the UK and includes all people living in private households and also National Health Service accommodations, and young people living away from home in a student hall or similar housing. Appropriate weights for each respondent make it possible to derive accurate estimates for the whole population in the UK. For the LFS a rotational sampling design is applied. This means that once a household is selected for the survey, it stays in the sample for five consecutive quarters. This way the precision of estimates of changes over time is improved and longitudinal data sets can be produced to capture different kinds of changes, i.e. if individuals become unemployed or change their profession.

Self-employment and total employment is covered in the basic economic activity indicators section and is part of the core questionnaire. Economic activity is defined according to the International Labour Organisation (ILO) guidelines, which includes all individuals who “furnish the supply of labour

(16)

16 for the production of goods and services”6. Therefore, employed and unemployed are likewise included. Professional partnerships, i.e. lawyers, doctors etc., are also considered as self-employed. The exact number of self-employed can be biased due to various reasons and can be over- or underestimated. For example if somebody who is actually a business owner, but is not considering himself as such, the number will be lower. So called “bogus employment” or “dependent self-employment” will cause overestimation. This is a case, where somebody claims to be self-employed, although the nature of her occupation is this of an employee, i.e. when she is only works for one employer or is not free to choose different contractors independently. This is especially popular in the construction sector, where due to cost pressure, many construction firms employ a lower number of direct employees (Harvey and Behling, 2009).

The number of service jobs and the total number of workforce jobs are part of a different labour market statistic, which yields a number of outputs by industry, region, and gender and full/part time status. The number of jobs is measured quarterly and is derived from different sources. Total workforce jobs is the sum of employee jobs, measured by employer surveys7_{, conducted both in the}

private and public sector, self-employment jobs, taken from the LFS, and government-supported trainees and Her Majesty’s Forces, which are both derived from administrative records. The total number of service jobs includes all jobs in the sectors G to T8, which covers all fields of services. It is important to notice, that this is a measure for jobs, not individuals or employment. Since a person can have more than one job or one job is shared by several people, the workforce job statistics differ crucially from the employment statistics in the LFS. Employment is “measured by the LFS as the number of people who worked at least one hour during the survey reference week”9. The workforce jobs statistics is a more accurate measure for jobs by industry and the development of an industry, rather than a measure of employment development. Since I am interested in the industrial structure of the British economy, I consider the share of service jobs of total workforce jobs to be a suitable variable for my analysis.

Foreign direct investment (FDI) is collected in a survey, which complements the Balance of Payment records. This survey collects information about direct investments in the UK by foreign enterprises (inward FDI) and direct investments abroad by UK businesses (outward FDI). In both cases the investment must be at least 10% of the ordinary shares or voting power. Investments of less than those 10% “are classed as portfolio investments [and] are collected in the National Accounts Division within the ONS”10. The survey consists of two separate sample frames to make a distinction between receivers and providers of FDI. About 4,400 enterprise groups take part in the survey for outwards FDI and 16,900 enterprise groups for inward FDI. Respondents of the outward population are wholly owned UK businesses with ownership of subsidiaries or branches in foreign countries. The same applies vice versa for respondents of the inward population. Since 1997 various changes were undertaken in the survey design and definition. The most important one, which will affect my time line as well, concerns the threshold of the size of investment. Before 1997, an investment of equivalent to at least 20% of shareholding was considered a FDI. After 1997 until now, this was lowered to 10%. This means, that the effect of globalization, which I want to capture with this

6_{See ILO LABORSTA Internet, http://laborsta.ilo.org/applv8/data/c1e.html , (19.07.2014)} 7

Private sector jobs are predominantly measured by Short-Term Employment Surveys (STES); Public sector jobs are measured by the Quarterly Public Sector Employment Survey (QPSES)

8_{For descriptions of industrial sectors see Appendix b.2} 9

See Office For National Statistics (2013), Quality and Methodology Information: Workforce Jobs, p.9 10

See Office For National Statistics (2013), Quality and Methodology Information: Foreign Direct Investment, p.2

(17)

17 variable, could be overestimated after 1997, since simply more investments were regarded as FDI. On the other hand, one could argue, that rapid globalization is anyway a development which especially gained importance in the mid-1990’s and after.

Data about GDP growth is collected in the quarterly national accounts. There are different measures of GDP and also different calculation methods with regard to inflation. Since I have a timeline starting in 1984, inflation is a matter that cannot be ignored. Therefore, I chose the estimate of GDP not at current prices, but calculated as chained volume measures. With this method instead of updating the base year every five years, it is done every year and gives a full real term time series. Major structural changes are more easily spotted and also more accurately depicted, which is interesting given the fact that industry and product mixes change more rapidly now than in the past. I believe that this is a suitable measure of GDP for my purposes and gives a more complete picture of the economic development.

The last two variables in my model, which capture the opening of the EU are self-generated dummy variables. In 2004, ten countries joined the EU on May 1st and in 2007 Romania and Bulgaria joined on January 1st. This means, that during the first four months of 2004, migrants were not able to enter the UK labour market freely and also other legal restriction were still in place. This could bias the result and make the influence on self-employment and other indirect developments look weaker than it actually is. On the other hand, the full influence of an enlargement unfolds and is wholly visible only after some months or even years, which make the first four missing months seem irrelevant.

iii.

Descriptive Statistics

To obtain an overview about self-employment and the explaining variables before doing regressions, I looked at descriptive statistics of each variable, which is presented in Table 3.1:

Variable Mean Min Max Std. Dev

Self –employment of total employment, in %

12.82 11.1 14.21 .8477368

Share of service jobs of total workforce jobs, in %

76.59 68.14 83.36 4.75597

FDI balance, in 1000 £ 147731.3 -550 408362 129807.4

GDP growth, annual % 2.60 -5.2 5.6 2.1607844

(18)

18 Figure 3.1.: Development of self-employment over time; Source: ONS

In the observed time frame, the share of self-employed in the UK is on average 12.8% of total employment. As a comparison, the average self-employment rate in the EU between 1991 and 2012 is about 18%. Between 1984 and 2012 the self-employment rate is about 11.3% in Germany and in Italy about 27.8%11. The UK is therefore rather on the lower end of the spectrum in comparison with other European countries. In 2012 the self-employment rate hit its peak with 14.21% and the low point was recorded in 1984 with 11.1%. The development of the share of self-employed in total employment over time can be seen in Figure 3.1. Overall, the general trend of the development of the share of self-employed is slightly positive, except for a negative development between 1995 and 2000. This decline in the self-employment rate is a reverse development after the steady increase in the 1980s until the early 1990s. Although there is only little research about the causes for the drop in self-employment, three main influences or changes can be assumed to have caused this development12. First, the factors that positively influenced the growth of self-employment in the 1980s had a weakened impact in the 1990s and the potential for further growth of the self-employment rate was limited due to structural limitations, i.e. in the construction sector or outsourcing of services. Also the enforcement of the income tax collection was altered in 1996 onwards, which reduced incentives to become self-employed. With tighter capital markets in the 1990s, it was also more difficult to get access to credit for start-up businesses and a slowed down housing boom reduced net personal wealth. Also, the UK experienced a policy change, where rather than stimulating start-ups, supporting existing small businesses was emphasized. Second, the recession in the early 1990s lead to significant job loss, especially in the service sector. This might also have implications for the development in the second half of the 1990s. Third, people who started a business in the 1980s included also people who traditionally had a lower survival rate in self-employment due to age, gender and employment background. These “new self-employed” were different than the traditional entrepreneurs and were more likely to come from unemployment and entered highly competitive service-sector activities. Hence, it can be assumed that these entrepreneurs formed an outflow of self-employment in the 1990s. The largest drop in two consecutive years was recorded between 1998 and 1999, where the share of self-employed of total

11

See World Bank Data, http://data.worldbank.org/indicator/SL.EMP.SELF.ZS?page=6 12_{See Arum and Mueller (2004)}

0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

Development of self-employment of total

employment 1984-2013

(19)

19 employment fell from 13.01% to 12.49%. After 2000, the self-employment rate recovered and apart from some small fluctuations rose steadily to the peak in 2012. The service sector is steadily growing, with a constant increase in the share of service jobs of total workforce jobs. The highest value of 83.46% is therefore recorded in 2013. GDP growth shows the most variation between consecutive years, and no clear trend. The FDI balance also underlies a general positive trend, with the lowest value in 1991, which is also the only time where this balance was negative. The ONS calculates the FDI balance as the difference between FDI outflows and inflows. A negative balance means therefore that FDI inflows were larger than FDI outflows. In all other years, outflows were larger than inflows of FDI.

After deciding on an appropriate model and looking at the trend of self-employment and other variables, the regression can be conducted.

4. Results

a. Regression Results

I conducted my empirical analysis with the base model described before in Stata. Due to the short time line of the dependent variable, my model is not complete and does not capture all aspects of self-employment and its influence factors. Nevertheless, it gives a first impression of the influence of the opening of the EU and provides a basis for further research. I used a simple OLS regression which yields the following results:

Source SS Df MS

Model 16.8344565 5 3.3668913

Residual 3.99455147 23 .173676151

Total 20.829008 28 .743893141

Self-employed Coefficient Standard Error t-value P> [95% Confidence Interval]

Share of service jobs .3364147 .0442571 7.60 0.000 .2448618 .427967

(Sqrt) FDI balance -.0064588 .0008591 -7.52 0.000 -.008236 -.0046815

GDP growth .0155044 .0500406 0.31 0.759 -.0880124 .1190212

EU opening 2004 -.7553351 .3236675 -2.33 0.029 -1.424892 -.0857779

EU opening 2007 .7262554 .3200812 2.27 0.033 .0641171 1.388394

constant -10.74787 3.140644 -3.42 0.002 -17.24478 -4.250948

Table 4.1: Regression output full model

According to this output, the share of service jobs, the square root of FDI balance and the constant are significant on a 5% significance level, whereas GDP growth is insignificant. According to this analysis, both EU enlargements are also significant for the explanation of self-employment. To reveal a possible misestimation, I conducted the regression separately for the EU dummies, with the rest of the model remaining the same. This way, I can check if the variables behave in the same way in different situations. Number of observations 29 F (5, 23) 19.39 Probability > F .0000 R-squared 0.8082 Adjusted R-squared 0.7665 Root MSE .41674

(20)

20 First, I conducted the regression without the dummy for the EU opening in 2007. In this separate estimation with the dummy for the EU opening in 2004 remaining, the share of service jobs, the FDI balance and the constant behave in the same way and are all significant. Now, the EU dummy for 2004 is insignificant on a 5%-significance level. The other separate regression shows similar results. The behaviour of the other variables remains constant, whereas the EU dummy for 2007 is now insignificant. This leads me to the conclusion that the dummies for the EU enlargements are in some way related to each other and in the single regression the effect of the other dummy is associated with the remaining dummy. Both variables need to be included in the model and in the regression, in order to obtain good estimates and separate the effects. The relationship between the enlargements of the EU is not very surprising and can also be observed when looking at the correlation matrix. Since each enlargement of the EU is a lengthy political process, that allows individuals and countries to prepare to expected changes, it is natural that the opening of the EU in 2007 was also associated before. These associations and expectations can be captured by the dummy for the enlargement in 2004. Also, adaptions that might have taken longer, like the aforementioned delay in migration, can be included in the dummy for the later enlargement. Together included in a regression, the majority of the effects associated with the EU enlargement and their effect on self-employment in the UK should be captured. Therefore, I will explain my results on the basis of the regression of the full model.

b. Interpretation and Explanation

According to my results, the share of service firms has a positive and significant effect on the share of self-employed in the UK. This was also expected according to economic theory due to the aforementioned lower entry barriers. Therefore, the bigger the service sector in the UK, the higher the share of self-employed will be. The square root of the FDI balance is also significant, and has a very small negative influence on the share of self-employed. As I have mentioned before, the coefficient cannot be interpreted as straight forward due to the undertaken transformation. Nevertheless, it is significant for the overall model and therefore necessary for a most complete picture as possible. The FDI balance combines inflows and outflows of foreign direct investment and will be negative if the inflow of FDI is larger than the outflow and positive if the opposite is true. Here, this measure is solely included to capture the influence of globalization, not specifically the influence of FDI in- or outflows. What can be concluded from this analysis is that globalization indeed has an effect on self-employment in the UK. The exact magnitude and direction of the effect of FDI needs to be examined in a different model. In this context it will also be interesting which kind of investments have a positive or negative influence on self-employment. It is possible that outsourcing of service jobs might have negative effect on domestic self-employment, since the share of service jobs determines the self-employment rate. Then again, most self-employed are engaged in the wholesale/ retail sale and construction sector, which are both branches that are not primarily affected by outsourcing. According to my results, the economic development has no significant effect on self-employment in the UK. This hints the ambiguous causality between these economic indicators, which I have already discussed before.

According to the results of my regression, the opening of the EU in 2004 is significant and has a negative effect, whereas the EU enlargement in 2007 has a positive effect on the share of self-employed in the UK. This is an interesting result and can have various implications and reasons.

(21)

21 The effect of intra-European migration is possibly captured by those two dummies. Hence, it is important to review migration patterns connected to the enlargement and discuss possible effects of this on employment and self-employment. There is a broad spectrum of literature examining the phenomenon and the effects of migrant entrepreneurship. As I have discussed in the literature review, in most OECD countries, migrants are more entrepreneurial active than their native counterparts. This also applies for the UK, although as one can see in Table 2.1 the share of migrant self-employed remained rather stable after 2001, whereas the share of native self-employed increased constantly. Still, the share of self-employed among migrants is always higher than the native self-employment rate, also when looking at most recent developments (Jones et al., 2014). Also, not every group of migrants is the same. Some groups of immigrants choose more often self-employment than other immigrants. In the UK, the most entrepreneurial active ethnic groups are Bangladeshis, Chinese and Pakistanis (Drinkwater, 2010). Whites, which also include EU citizens and British, are moderately entrepreneurial active. As the Eurobarometer (2010) on geographical and labour market mobility reports, for citizens of the new member states that joined the EU in 2004, Germany and the UK are the most preferred destination countries. Since there were no transitional restrictions applied, the UK experienced the largest inflow of migrants from EU-8 countries, particularly from Poland, after 2004. This results from a shift away from traditional destination countries, namely Germany, towards more easily accessible countries like the UK (Holland et al., 2011). Changes in employment patterns of new migrants could also be observed. According to Drinkwater (2010) the self-employment rate among migrants from the EU-8 countries was higher before the opening of the EU than afterwards. This was due to easier immigration policies for entrepreneurs to the UK. More new migrants arriving after 2003 chose paid labour instead of self-employment and a majority is successful in finding work (Jones et al., 2014). Nevertheless, especially Polish immigrants are entrepreneurial active, also after the EU enlargement, with 9% of Polish migrants that arrived between 2004 and 2013 being employed. As a comparison, the self-employment rate for Polish migrants arriving between 1994 and 2003 was 25%13. These changes could bear one explanation for the negative impact of the EU opening in 2004. Since immigration was very selective in the pre-enlargement era, a strong positive effect should be observed when looking at the effect during this time, whereas the effect of immigration due to the enlargement in 2004 is compared to that negative. Migration from Romania and Bulgaria underlies different patterns. Since transitional restrictions were still in place for citizens migrating from the EU-2 states until the end of 2013, the migration distribution is rather selective. For Bulgarians Spain, Germany and Greece are the most preferred destination countries, whereas most Romanians choose Italy and Spain, followed by Germany and the UK as a destination. The majority of Bulgarian migrants are medium-skilled and Romanians are low- and medium-skilled. Germany on the other hand attracts highly-skilled migrants from Romania. Still, an increase in the number of migrants from the EU-2 to the UK could be observed after the accession in 2007. EU-2 citizens experienced similar restrictions as did citizens from the EU-8 states. Before the complete opening of the labour market on January 1st 2014, individuals had to apply for a work permit and it was easier for self-employed and highly-skilled to settle down in the UK (Holland et al., 2011). Therefore, a positive selection of entrepreneurially active migrants should be observed when looking at migration patterns from the EU-2 to the UK, which could partly explain the positive effect of the opening of the EU in 2007 in my analysis.

Differences in migration behaviour are one explanation for the different effects of the EU enlargements. Other explanations could be the enhanced trade and investment, a bigger market for

(22)

22 domestic entrepreneurs and a diversification of skills and knowledge, which is beneficial for the overall economic development and also affects self-employment.

When comparing bilateral trade patterns between the UK and EU-10 and EU-2 countries, differences can be observed. Since 2006, exports to Bulgaria experienced the highest average annual increase compared with exports to other countries of the EU14. Exports to Romania reached their peak in 2012 with £940 million and an average annual growth rate of 5.5%. For Bulgaria, exports grew on average 7.1% annually and reached their peak also in 2013 with £389 million. The same picture evolves for imports. Imports from Romania grew between 2006 and 2013 on average 7.0% annually with a peak in 2013 of £1,441 million and imports from Bulgaria grew with 8.0% to a peak in 2013 of £367 million. Among the EU-10 countries, Poland is the most important trade partner for the UK. Exports to Poland experienced an average annual growth rate of 3.4% and imports grew on average 11.7%. Only the export volume to Slovakia shows another high increase comparable to that of Bulgaria and Romania, with average annual growth of 6.8% and average annual growth of imports of 13.6%. Other EU-10 countries show even a decrease in trade with the UK. The sharpest decline can be observed in export volume from Latvia (-5.6%) and Cyprus (-9.7%). Import flows from the EU-10 countries are similarly diverse. Imports from Lithuania grew on average annually by 15.7% and from Slovakia by 13.6%. Imports from Estonia and Cyprus decreased substantially, by -11.3% and

-21.6% respectively. Overall, the enlargement in 2007 was beneficial for trade to and from the UK. The group of EU-10 countries is more diverse and each country underlies individual developments. Especially trade with countries that were intensely affected by the recent financial crisis, like Cyprus, experienced a sharp decline in trade volume with the UK. The observed positive development of trade with Romania and Bulgaria and the ambiguous effect of the EU enlargement in 2004 on trade relations is another explanation for the difference in effects on self-employment. Unfortunately, data on investment flows to and from the EU-2 countries and also partly to and from EU-10 countries is incomplete and does not allow drawing conclusions.

Other possible impacts of the enlargements, like an inflow of new skills and knowledge, cannot be observed directly. This is also closely connected to migration. As mentioned before, since citizens from the EU-2 states were affected by transitional restrictions, this group of migrants is also in average slightly higher educated than those coming from the EU-10 countries. This also contributes to the positive effect of the enlargement in 2007, compared to the opening in 2004.

One motivation of enabling citizens of the EU to move freely and pick up work in every member country of the EU, was also to close gaps between labour demand and supply and ensure a better distribution of workers. This is a development that I cannot measure in this context and might also require different data sources and a different analysis. Migrants from the EU-10 as well as from the EU-2 are very successful in finding work in the destination country and their employment rates are higher than the employment rate of the population in their home country (Rolfe et al., 2013). Also the employment rate of mobile workers of the two country group is comparable to the employment rate in the EU-15 states. This hints to a good match between labour demand and supply, which would result in a positive effect of the EU enlargements on self-employment in the UK.

The overall economic situation in Europe, which was highly affected by the recent financial crisis, also needs to be borne in mind. As mentioned before, among the EU-10 are also countries that were hit hard by the crisis, like Cyprus, although small. Also, the UK itself was affected, which also resulted in negative GDP growth of 5.2% in the year 2009.

The impact of the enlargement of the European Union in 2004 and 2007 on self-employment in the United Kingdom