Labour Emigration Brain Drain and The Role of Labour Market Policies: An Empirical Study of the Greek Case

(1)

Photo: Young People in Greece (Tserepi, 2016) Leiden University

Faculty of Governance and Global Affairs

MSc in Public Administration Economics & Governance MASTER THESIS

Labour Emigration Brain Drain and The

Role of Labour Market Policies

An Empirical Study of the Greek Case

By Marinela Tserepi

(2)

2

Acknowledgements

Writing this thesis has been a great learning experience for me, both at an academic and personal level. It is especially thanks to my supervisor that I have reached this final point of my master’s studies, as she has been very supportive. I want, therefore, to first express my gratitude to Dr. Kim Fairley for her valuable guidance during these past months.

At the same time, I feel extremely thankful to my parents who have always believed in me and have been by my side, no matter the challenges across the road and the distance. And, I want to thank my dear sister Irini and brother in law, who have encouraged me and offered meaningful advice.

Finally, I want to thank my friends with whom I have shared this chapter of my life and who have repeatedly shown how much they care, in beautiful creative ways.

(3)

3

1

Introduction

Educating a young person is quite an expensive activity both for the taxpayers as for the students and their families (Cerdeira, Cabrito, Machado-Taylor, & Patrocínio, 2016, p. 790). It requires specialized material resources and highly qualified human resources while the foregone earnings and the costs of living during education will take long until they turn into financial gains (Cerdeira, Cabrito, Machado-Taylor, & Patrocínio, 2016, p. 790). The emigration of the highly skilled is therefore a loss for the country of origin as the amount of resources the country has spent on their education is unlikely to bear any fruits if the emigrants do not return to their country (Cerdeira, Cabrito, Machado-Taylor, & Patrocínio, 2016, p. 778). The phenomenon that “marks the international transfer of resources in the form of human capital and mainly applies to the immigration of highly skilled people from less-developed to developed countries” has been termed as ‘brain drain’ (qtd. in Theodoropoulos et. al 2015).

Greece has been on the spotlight as a case of brain drain in the recent years as its weak financial situation that led the country to almost default on its sovereign debt, also led thousands of people to seek better opportunities abroad. The annual emigration flows have risen from 40,000 persons prior to 2010 to above 100,000, and in contrast to the previous waves most of the late emigrants are young, and highly educated (OECD, 2018). This leads to a vicious cycle where the highly skilled individuals prefer to leave the country for better opportunities because of the crisis, while their emigration could slow the progress of the country’s situation (Labrianidis & Sykas, 2017, pp. 126-127). And, the problem gets intensified if we consider the important role that the highly skilled can play for the country’s growth and development (Barro and Sala-Martin 1995; Driouchi, Azelmad, and Anders 2006; Jones 2002). The basic argument is that countries that utilise smartly their human capital are able to take advantage of the benefits of technological advances and raise their productivity because of the high level of knowledge that the highly educated possess and can apply (Labrianidis & Vogiatzis, 2013, p. 529).

The high numbers of the Greek highly skilled emigrants can be partially explained considering the country’s high unemployment rates this last decade. Youth unemployment (aged 15-24 years) as a percentage of the youth labour force reached in 2016 the level of 47.4% while the

(5)

5

highest unemployment rate among the tertiary educated has also been observed in Greece in 2017 (OECD, 2018). The persisting high unemployment rates of the young and the highly educated people has been showcasing an oversupply of human capital on the one hand, and the inability of the country’s labour market to make use of their skills on the other. The highly educated people more than the lower educated ones are, then, willing to migrate to escape unemployment because they are better equipped in terms of knowledge and skills to handle the uncertainty encountered in a foreign country (Bauer & Zimmermann, 1999, p. 15). For the policymakers in Greece to plan forward on addressing youth emigration and brain drain, this study aims to answer primarily the following question: “To what extent is labour emigration

influenced by the level of education attainment?”. Other important determinants, especially

concerning previous occupational experiences and unemployment characteristics, are also studied.

Moreover, given the role that the necessity and willingness to start working can play in the migratory behaviour, it is meaningful to explore how policy measures designed to mitigate unemployment spells could influence this intention. The young people are more susceptible to unemployment and they have a greater difficulty in job finding because of their inexperience and lack of connections in the local labour markets (Arntz, 2005). The labour market policies, which are meant to help individuals during unemployment, could assist the young people in maintaining a decent life in their country while searching for employment and support them in the initial stage of their career and the integration in the local labour markets. In the case that the unemployed are willing to migrate to find a job, to what extent do labour market policies influence this intention, given the relief they can provide to unemployment? This is a secondary question that this study tries to answer with the aim of offering an insight into the confrontation of unemployment induced emigration. Simultaneously, the purpose of this thesis offers an addition to the brain drain literature, where the role of the unemployment characteristics in emigration have been only limitedly investigated.

Aiming at contributing to the knowledge and understanding of the young Greeks emigration intention, this research study follows the format of a quantitative empirical analysis approach. We begin with a presentation of the case, highlighting the existing gap in the literature and how the study will contribute in the scientific knowledge (Chapter 2). In Chapter 3 we describe the labour market policies that are present in Greece in the period of interest (2016) that are meant for the unemployed individuals in the country. In the next chapter (Chapter 4), the theoretical

(6)

6

background and the causal mechanisms that inform and guide our study are analysed and from that analysis we derive the key hypotheses that we will test in the next step of our work. Research Design (Chapter 5) then follows, where we describe the method of data collection and the dataset we are using, the population of interest and sample characteristics, and operationalize the variables whose relationships we try to quantify. In Chapter 6 of the Empirical Methodology, we explain what statistical methods we are using in order to study those relationships and what equations are used for this purpose. In Chapter 7 we present the results of our research methods and in Chapter 8 we look back at our hypothesis to test to what extent they hold true. At the same time, we analyse the findings through the lenses of the theory presented in Chapter 3 and we try to explain the aspect of the reality of the Greek brain drain which we have decided to zoom in. Finally, we proceed with the Conclusion (Chapter 9) where we summarize the research’s main findings, talk about the limitations of our work, and provide an analysis of the policy implications. The chapter closes with recommendations for future research that could further contribute in this area.

(7)

7

2

The Case of the Greek Brain Drain and the

Research Gap

The severity of the Greek emigration during the crisis and particularly its brain drain can be spelled out by the striking numbers. It is difficult to precisely present the volume of the emigration flow as the Greek statistical Authority stopped collecting migration data in 1977, let alone looking at the demographic characteristics of these populations (Cavounidis, 2015, p. 5). However, a lot of estimations have been made trying to capture a picture of reality. In 2010/11 Greek emigrants residing in OECD countries amounted to 655,000 of which 143,000 were highly educated, as the OECD (2013) has estimated. In 2012 nearly 33,000 Greeks arrived in Germany and other popular destinations have been the United Kingdom and the Netherlands with 6,000 and 3,000 respectively, while 80% move within the EU (Cavounidis, 2015, p. 6; OECD, 2018). Worth highlighting is the outflow of the medical doctors, with Germany being the main destination country as the numbers of Greek trained doctors working there increased from 1,700 in 2008 to 2,600 in 2012 (Cavounidis, 2015, p. 8).

Figure 1. Emigration from Greece 2008-2017. Information Source: Eurostat Database ‘Emigration by sex, age

(8)

8

The high numbers of the Greek highly skilled emigrants can be justified considering the country’s high unemployment rates this last decade. The unemployment rate of the 25-64 year-olds having received tertiary education reached 15.7% in 2017, the highest among all OECD countries where the average is 4.1% (OECD). From the figures 1 and 2 we can see an association between the trendlines of the unemployment rates and the emigration flows which has been widely confirmed (Ifanti, Argyriou, Kalofonou, & Kalofonos, 2014; Labrianidis & Sykas, 2017; Triandafyllidou & Gropas, 2014; Lafleur & Stanek, 2017). Unemployment is one of the most common cited reasons but not the only one that accounts for high Greek emigration flows. Other causes that have been documented with the use of surveys include low wages, the willingness to improve one’s educational/professional training, poor overall quality of life, and seeing no future for one’s country (Triandafyllidou & Gropas, 2014, p. 1623).

Figure 2 Unemployment of the Tertiary educated Greeks as a % Rate of the Active Population by Year (age 15-74).

Information Source: Eurostat Database ‘Unemployment by sex, age and educational attainment - annual averages’. http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=une_educ_a&lang=en

The Greek case of brain drain has been especially on the spotlight lately due to the recent economic crisis, but the phenomenon is rather old. As early as 1972, Kourvetaris had distinguished Greece as an important case of brain drain. To give an example of the magnitude of the situation at the time, between 1957-1961, over one fifth of Greek engineers had migrated to the United States (Kourvetaris, 1972, p. 6). Today, it is broadly stated that while the quantification of the outflow is challenging on its own, it is even more difficult to analyse the

(9)

9

human capital aspect of the outflow, or the specific characteristics of the individuals deciding to migrate (Ilić & Milosavljević, 2017). The final report ‘Outward migration from Greece during the crisis’ written by Labrianidis and Pratsinakis (2015) gives some of the most extensive information on the post emigrants’ profile. They realized a nationwide representative telephone survey that draws data by asking households about family members who live abroad or those who have returned in the country after migration. The questions allow for drawing a picture of the socioeconomic background of those who had left Greece including among others an inquiry in the reasons for emigration, the existence of connections with people in the destination countries, and the accessibility to the destination country’s labour market (Labrianidis & Pratsinakis, 2015, p. 5). The study finds that 75% of the emigrants in crisis period hold university degrees while 25% of the post-crisis emigrants hold post graduate or doctor degrees (Labrianidis & Pratsinakis, 2015, pp. 12-13). It is also found that in the post-crisis period ‘low to very low’ income households are more prone to emigrate and the mean age of the emigrant in the post-2010 period is 30,5 years old (Labrianidis & Pratsinakis, 2015, p. 14).

While quantifying the realized migration flows is important, equally of interest is the estimation of potential flows and the individuals’ willingness to leave the country so that the policymakers can get a picture of the coming future. In the “Why High School Students Aspire to Emigrate: Evidence from Greece” Labrianidis and Sykas (2017) study the reasons behind the Greek high school students’ aspirations to emigrate for university studies, especially focusing on their education and socio-economic background. They find that the students who express a higher intention to emigrate are ambitious and have high educational and professional expectations. They also have high-mean school grades and come mostly from middle and upper middle social class with highly educated parents. The authors highlight what these findings matter a lot for the case of Greece which has been facing a severe recession. According to their analysis, the emigration of this dynamic part of the Greek youth constitutes a significant loss of developmental human resources for the country and can aggravate the effects of the long recession. This leads to a vicious cycle where the highly skilled (in this case potential) individuals prefer to leave the country for better opportunities because of the crisis, while their emigration could slow the progress of the country’s situation (Labrianidis & Sykas, 2017, pp. 126-127).

(10)

10

We conduct this empirical study with a similar scope, that of offering information about emigration determinants and the likelihood of the persistence of the Greek brain drain. What differs in our work is the target population which includes only the unemployed young people in Greece. As far as we are concerned the brain drain of the highly skilled unemployed individuals is a new addition to the Greek brain drain literature. The previous empirical Large-N quantitative studies have approached brain drain broadly but we argue that by focusing on this specific group we can also examine how unemployment as an experience of its own nature drives emigration depending on its various characteristics. The survey from which we draw our data focuses on the position of the young people in the labour market and this permits an examination of all those characteristics of unemployment, which are explained in the next chapter, that can influence the intention to migrate abroad. We do not study in general if the highly skilled are more mobile than the lower skilled as studied before. Our question is focused on the individual’s willingness to move abroad to start working. This is a very specific reason for migration and differs from other migration experiences such as studying abroad. The country loses human capital due to its inability to provide employment opportunities. Therefore, it is also vital to understand how the government’s response to unemployment, through the labour market policies, influences if at all the intention to seek employment abroad. Shedding light on these specific aspects of brain drain could be beneficial for incentivising the young people, including the highly skilled, to stay in the country and try to position themselves in a place within the local labour market at difficult times of unemployment.

(11)

11

3

The Labour Market Policies

For the alleviation of unemployment effects, governments can take different measures and apply diverse policies. Following the classification of the European Commission (2006), labour market policies consist of Passive Labour Market Policies (PLMP) which refer basically to financial support and benefits, and Active Labour Market Policies (ALMP) which include a) services such as advice on job searching, and b) measures such as training and work placements (Bozani & Drydakis, 2015, p. 131). The common active labour market policies that have been provided in Greece during the period that we study include both PLMP and ALMP. These are administered by O.A.E.D or the Hellenic Manpower Employment Organization, the main public organization in Greece for managing labour market policies.

In Greece, PLMP refer to the income support for the unemployed, which is provided to counteract the negative influences of unemployment on the individual’s living standards (Bakirtzi, 2016). This income assistance is provided to the eligible unemployed individuals under certain conditions. In general terms, the first requirement is the status of unemployment of an individual who has legally exited employment after termination or expiration, is looking for employment, agrees to work within a employment sector suggested by O.A.E.D., or agrees to participate in vocational training or retraining (Bakirtzi, 2016, p. 121). Second, the individual should be able to work, or in other words does not have a disability that prevents them from working (Bakirtzi, 2016, p. 121). Third, they need to be available for work and open to accept any appropriate position offered (Bakirtzi, 2016, p. 121). Fourth, the unemployed must have worked for a specific number of days (generally 125 full-time working days) required according to their profession in the period of 12-14 months before dismissal (O.A.E.D., 2016; Bakirtzi, 2016, p. 121). The unemployed person needs also to be registered in O.A.E.D and after dismissal apply for the benefits within 60 days (Bakirtzi, 2016, p. 121).

The benefits take the form of regular unemployment benefits as basic insurance, and unemployment assistance benefits are special unemployment benefits for various instances such as for cases of employer bankruptcy (Bakirtzi, 2016, p. 121). Also, limited unemployment benefits are provided for the case of youth unemployment for individuals between 20-29 years

(12)

12

of age who are new labour market entrants and are degree/diploma holders (O.A.E.D., 2016). Finally, for the long term unemployed (more than 12 months) benefits are provided on the condition that their family income is within a certain limit (Bakirtzi, 2016, p. 121).

As regards the ALMP, these can be applied as complementary to PLMP or they can be provided to the ones that have exhausted their unemployment benefits (Bakirtzi, 2016, p. 126). Amongst the eligible, the most targeted groups include the young individuals, the long term unemployed, and the older unemployed (Bakirtzi, 2016, p. 126). A big part of these policies are the “Employment Programmes for Εnterprises – Employers (Eligible Claimants) and Unemployed People / Workers (Beneficiaries)” whereby the employers receive subsidies for hiring the unemployed (O.A.E.D., 2016). Also, popular in this category of policies are the training programs, the vocational training programs, and training vouchers (O.A.E.D., 2016). And, finally, O.A.E.D. offers counselling services in the form of career choice advice, job searching advice and support, entrepreneurship advice on starting a business and support on entering self-employment (O.A.E.D., 2016).

(13)

13

4

Theoretical Framework

The description in the previous chapters of the Greek case of unemployment and the brain drain phenomenon, has offered an understanding about the environment facing the young unemployed individuals. In our effort to bring a deeper understanding on the driving forces of the youth emigration we offer in this section a theoretical explanation of the possible factors and the underlying mechanisms. The neoclassical migration framework offers such explanatory tools as it is concerned with both the macro-level and the individual micro-level characteristics that explain the migratory behaviour. On the macro-level, the focus is mostly on the geographical differences in supply and demand for labour and on the micro-level individuals are viewed as rational actors who decide to move on the basis of a cost-benefit analysis (Haas, 2011, p. 9). For the policymakers who need to respond effectively to the effects of migration and mitigate some of its potential negative effects, understanding the macro-level factors is crucial but in the short term it could prove more valuable if they could influence the decision making cost-benefit processes of individuals by using economic instruments and applying the principles of behavioural economics.

i. The Human Capital Theory of Migration

Among the individual characteristics that have received considerable attention is education attainment. The question of how education influences a person’s decision to emigrate has been the subject of a considerable number of scientific studies (Long 1973, Lambrianidis & Sykas 2015, Ozden & Schiff 2006, Grapsa 2018). This is a topic of the neo-classical micro-model of migration and more specifically the branch of the theory that primarily deals with this question is the so-called human capital theory of migration. The theory suggests that the socio-demographic characteristics of the individual are important determinants of migration and at the core of the theory lies the assumption that the individual is a benefit maximiser (Kurekova, 2011, p. 6). According to the human capital theory, individuals that choose to migrate tend to be relatively more skilled than the ones who do not. The causal explanation is that the more skilled individuals are better able to collect and process information, given their higher

(14)

14

education levels, and can therefore reduce the risks associated with the migration process (Bauer & Zimmermann, 1999, p. 15). It is thus assumed that individuals with higher levels of education attainment are more likely to emigrate than the ones of lower education levels (Mora & Taylor, 2006, p. 21). As Ahn, De la Rica, and Ugidos (1999) estimate, university graduates are about six to nine times more willing to migrate than the ones with a lower level of education. This leads us to the main hypothesis:

H1: Ceteris paribus, highly educated individuals are more likely to emigrate than the lower educated. As the level of education is expected to influence the willingness to migrate, so is the field of studies. Assirelli, Barone, and Recchi (2019) point out that degrees differ in their marketability, explaining that some fields produce knowledge and skills that are more nation specific such as law and the ones that generate national civil servants such as social workers. And, they identify other degrees as involving highly internationally transferable skills such as scientific fields. There are also those fields with a cosmopolitan orientation including foreign languages and international relations that are both more likely to attract students who are internationally oriented and at the same time boost the students’ cosmopolitan view (Assirelli, Barone, & Recchi, 2019, p. 9). It is reasonable then to claim that these graduates will be more open to emigration. Some majors and degrees are more in labour market demand than others which means that those graduates that are less needed in the labour market experience higher unemployment risks and difficulties in job finding. As the local economy is not able to offer adequate occupational opportunities, those individuals might be led to consider more than others the decision to move elsewhere for employment.

The human capital theory of migration does not only focus on how skill influences migration but also on various other individual characteristics. The theory’s central assumption is that apart from skill, the individual’s capital endowments such age, marital status, gender, occupation, labour market status, and expectations, strongly influence the decision to migrate (Kurekova, 2011, p. 6). All these socio-demographic characteristics have received extensive consideration in studies of migration determinants and rightly so as they can play an important role in the migration decision as has been empirically proven. Each one of them is said to influence this decision towards opposing directions, and this requires a theoretical explanation. Firstly, as studies have shown, age matters. It has been empirically shown that it is more possible to migrate at younger ages (Bauer & Zimmermann, 1999). Only the younger ages of 16-19 have been found to be less likely to migrate and this can be due to their financial

(15)

15

dependency on their families (Ahn, De la Rica, & Ugidos, 1999, p. 340). Regarding gender, also differences have been found between the likelihood of migration among males and females. In general terms, the findings show that women are less likely to migrate than men which could be attributed to the fact that women are more often restrained by family relationships (Assirelli, Barone, & Recchi, 2019, p. 14). However, when gender is interacted with education level, it is found that women have a higher likelihood of migration, possibly because highly educated women are more ambitious than men (Ahn, De la Rica, & Ugidos, 1999). An opposite result is found by Arntz (2015), according to whose estimates higher education attainment leads to an increase in migration probability which is greater by 8.9% for males compared to that of women. Like gender, marital status also can be influential for migration, in a positive direction in case of unmarried individuals and in a negative otherwise (Ahn, De la Rica, & Ugidos, 1999, p. 346). This can be intuitively explained considering that married people have more responsibilities than the unmarried. Other researchers have distinguished between the migration behaviour of nationals and non-nationals. In theory, it is assumed that non-nationals are more likely to emigrate than nationals, both because they already have a migration experience and because they have the option of returning to their origin country in case the circumstances get unfavourable in the residing country (Assirelli, Barone, & Recchi, 2019, p. 8). The theory has been empirically tested and as suggested, it has been found that the effects on migration of the unemployed are larger for the families with migration history than those without (Assirelli, Barone, & Recchi, 2019; DaVanzo, 1978). As these socio-demographic characteristics can be determinants of migration in different degrees and directions, as it is apparent in previous research studies, we have an important reason to include them in our analysis and estimate their impact. This is not just meaningful for its own respect but also for clearing out the noise from the main relationship under study that is the migration differentials by education attainment.

ii. Unemployment Features Determinants

Unemployment is a socio-economic condition that has been recognized in extensive literature as a significant factor in leading people to migration (Arntz, 2005; DaVanzo, 1978). According to the International Organization for Migration “the movement of persons from one state to another, or within their own country of residence, for the purpose of employment” is defined as labour migration, which differs from other types of migration such as forced migration, when

(16)

16

a person’s life is threatened in their own country due to natural or man-made causes (International Organization for Migration, 2011). Additionally, when migration refers to the act of exiting one country in order to settle in another, the relevant term is ‘emigration’ to distinguish it from immigration whereby non-nationals move into a country with the purpose of settling in (International Organization for Migration, 2011). These concepts are important for the purpose of this study as we mainly focus on the movement of individuals outside of Greece for the purpose of finding a job, thus appropriate is to refer to them as ‘labour emigrants’. Here, however, we do not refer to realized but to potential labour emigration. It is very important to understand clearly this difference as we aim at examining the highly educated willingness or intention to migrate for work abroad rather than actual migration behaviour. Among the external determinants that can affect the willingness to migrate are local economic variables such as local unemployment rates. In theory, it is expected that regions with higher unemployment rates, lower vacancy rates, lower real wages, and higher housing prices will showcase higher outmigration rates (Ahn, De la Rica, & Ugidos, 1999). However, empirical evidence contradicts such theory especially with regards to the local unemployment rate effects. Numerous studies find no significant difference in the willingness to go between those living in regions with high unemployment rates and those living in regions with lower rates (Lowry & Land, 1968; Gallaway, Lansing, & Mueller, 1968; Ahn, De la Rica, & Ugidos, 1999). Testing a similar hypothesis, DaVanzo (1978) finds that origin unemployment rates affect outmigration, but this effect is visible only for the unemployed individuals.

Some of the previous research studies that have examined the determinants of migration among the unemployed, have put emphasis on how the factor of unemployment duration affects the migration decision. The evidence presented in the literature is, however, contested. On the one hand, there is the assumption that the longer an individual is unemployed the more likely he is to migrate, given that he/she has made complete use of unemployment insurance and other benefits and needs to find other sources of income to depend on, likely abroad (Ahn, De la Rica, & Ugidos, 1999, p. 341). This assumption has been verified in the study of Hughes and McCormick (1985) who find a positive relationship between unemployment duration and UK households’ intention to migrate. On the other hand, the longer an individual is unemployed the lower his/her budget is, thus rendering the choice of migration and its related costs unbearable (Ahn, De la Rica, & Ugidos, 1999, p. 341). Goss and Schoening (1984) find evidence in favour of this assumption as in their results it appears that in the U.S. regions with

(17)

17

long term unemployment, resides long-term unemployed population that is less willing to realise a geographic move. Here, the length of unemployment is operationalized as the time spent looking for a job and more specifically measures the number of weeks an unemployed worker has been looking for a job at the time of the interview (Goss & Schoening, 1984, p. 574). In the research study “Willingness to Move for Work and Unemployment Duration in Spain”, Ahn, De la Rica, and Ugidos (1999) find that generally unemployment duration does not seem to determine the migration willingness, just that women with longer unemployment spells are somewhat less likely to migrate. In this case unemployment duration is operationalized as the number of months elapsed before an individual finds a job.

Given the focus in the literature on the factor of duration of unemployment in influencing migration, we include this variable in our study by looking at the length of unemployment. As we cannot derive from the literature a dominant hypothesis and taking into consideration the characteristics of our sample of individuals, our reasoning leads us to adopt the assumption that longer unemployment negatively impacts migration willingness. This statement lies on the fact that the individuals in our sample are young and therefore less likely to be financially independent to undertake the costs of migration when they have been for a long time unemployed.

iii. Previous Occupation Determinants

There is also evidence suggesting that one’s previous occupational status can impact their labour emigration willingness. When one has been previously employed it is assumed that they have higher chances of finding re-employment because of their experience and their connections with businesses and local job providers (Arntz, 2005). This experience can make someone more confident about looking for a job locally and persisting in this endeavour. On the other hand, when someone has just graduated or has not worked before then they might get more easily discouraged due to the job finding difficulties. And, previous profession, for those that had been employed appears to matter in the labour emigration intention. Arntz (2005), for example, finds that white collar workers are a lot more mobile than skilled blue collar workers.

(18)

18

iv. The Role of Labour Market Policies

What the existing brain drain studies have been trying to do is to establish a relationship between skill level and emigration, verifying or refuting the brain drain hypothesis. What we identify as missing from the brain drain literature is the study of other variables that can influence this relationship as the main issue facing policymakers is how to handle brain drain and influence individual decision making. Given that unemployment is considered as the main driving force of emigration from Greece, we would like to examine whether the government active labour market policies can influence a highly skilled unemployed individual’s decision to move abroad to start working. The assumption that labour market policies can influence emigration is not an original one and has been researched in previous studies which are relatively limited compared to those focusing on other migration determinants.

Westerlund (1998) has stated that labour market programs could be substituting migration for the unemployed. This may happen because the provision of the labour market policy measures pushes the reservation wage upwards and this in turn lowers the out-migration rate (Westerlund, 1998, p. 370). But different labour market policies have different goals and effects and Westerlund draws the attention on some other types of active labour market policies, those with a scope of stimulating migration. The example in the study comes from the “Swedish model” of active labour market policies, especially the starting assistance grant which was offered exclusively to the unemployed with the aim of increasing labour mobility and helping the recipients meet travelling costs to job interviews or get access to private housing. From a different perspective, the study titled “The Geographical Mobility of Unemployed Workers: Evidence from West Germany” by Melanie Arntz (2005) sheds light particularly on the question whether the accommodation of active labour market policies reduces the interregional mobility of the unemployed. This is described by the author as a locking-in effect whereby the unemployed persons may delay or cancel their movement if they are beneficiaries of active labour market programs (Arntz, 2005). However, the results of the study indicate that this is not the case and the locking-in effect is not verified.

As has been hypothesized in the literature, labour market policies can influence the migration decision by way of providing to the individual a substitute for the opportunities sought in another region (Westerlund, 1998). Following a similar reasoning, we can assume that labour market policies can influence the migration decision as the unemployed individual may be less

(19)

19

willing to migrate if he/she receives government support and can meet his/her basic needs. This leads us to the following hypothesis:

H2: Ceteris paribus, unemployed individuals who are beneficiaries of passive/active labour market policies are less likely to emigrate than non-beneficiaries.

As in our study we focus on the highly skilled unemployed individuals’ willingness to migrate, it is important to study how the labour market policies influence this group’s attitude towards migration. We have made clear in the previous section that the highly skilled have extended training and have higher reservation wages than the lower skilled. This consideration could imply that the labour market policies, both active and passive, have a negative effect, but overall lower effect on this group’s willingness to migrate (as their expectations are higher) than those of the low skilled individuals. Therefore, we hypothesize that:

H3: Ceteris paribus, highly educated unemployed individuals who are beneficiaries of passive/active labour market policies are less likely to emigrate than highly educated unemployed individuals who are non-beneficiaries.

H4: Ceteris paribus, highly educated unemployed individuals who are beneficiaries of passive/active labour market policies are more likely to emigrate than lower educated unemployed individuals who are beneficiaries of passive/active labour market policies.

The theoretical analysis that has been presented in this chapter can be summarized in the following visual illustration of causal relationships; the study’s theoretical model guiding the empirical investigation:

Level of Education Labour Emigration

Labour Market Policies

• Other Socio-demographic determinants • _{Unemployment Characteristics Determinants} • Previous Occupation Determinants

(20)

20

5

Research Design

v. Population of Interest and Unit of Analysis

The population of interest in this study comprises of the young and unemployed individuals residing in Greece. Therefore, our study applies a cross-sectional individual level analysis approach that is time specific and aims at covering all regions in the country.

An unemployed individual, according to Eurostat, and in line with the guidelines of the International Labour Organization, is considered someone aged between 15 (in some countries 16) and 74, without work during the reference week, available to enter employment within the next two weeks (or has already found a job and is going to start in the next three months), actively having sought employment at some time during the last four weeks (Eurostat, 2010). As young we define the individuals between 18-34 years of age. Starting with the age of 18 is a choice made considering that it is improbable that individuals below that age could have completed tertiary education level studies.

vi. Method of Data Collection and Sample Characteristics

For the purpose of this research, appropriate and useful has proven the 2016 EU Labour Force Survey (EU-LFS) ad hoc module which studies the topic of young people on the labour market. The EU-LFS is an inquiry that is realised through a series of personal interviews directed to households and the objective is to collect information on the labour market (Eurostat, n.d.). It was established by the Council Regulation (EC) No 577/98 of March 1998 and is conducted in the European countries by the National Statistical Institutions (Eurostat, n.d.). The ad hoc modules of the EU-LFS are separate surveys that add each year an extra set of variables to the core surveys to provide information on different topics (Hungnes Lien & Antuofermo, 2018, p. 5). The topic of the young people on the labour market has been included three times in 2000, 2009, and 2016 (Hungnes Lien & Antuofermo, 2018, p. 5). Although the EU-LFS anonymized micro-datasets are publicly available on the Eurostat website, the latest for which

(21)

21

access is possible are from 2013. In order to obtain the 2016 most recent micro-data on the ad hoc module on the young people on the labour market for the case of Greece, we directly communicated with the Hellenic Statistical Authority of the country, which provided us with the survey micro-dataset in SPSS format.

The target population of the ad-hoc survey was individuals of 15-34 years of age and the response rate of the survey reached 97.2%, which equals 11,363 persons. Although the survey was addressed to all individuals in this group including those in employment, the ones not in employment, and the unemployed, we have filtered the cases of the unemployed as the question of labour emigration is exclusively addressed to them, and they constitute the suitable sample for inferring conclusions about our population of interest. After leaving out the irrelevant cases we are left with a sample of 2,064 individual cases.

vii. Operationalization of Variables

Dependent Variable: Labour Emigration

The dependent variable in our study is ‘labour emigration’. For the purpose of measuring labour emigration, we make use of two questions in the survey questionnaire that are only addressed to the unemployed individuals. The first question asks about the individual’s willingness to move in order to start working and reads as follows: “Would you be willing to change place of residence in order to start working?” with the possible answers being ‘Yes’, ‘No’, ‘No answer’. And, the other question inquires on the moving distance that the respondent is willing to undertake for this purpose, asking: “How far would you be willing to move in order to start working?”. The possible answers include ‘Only inside Greece’, ‘Inside Greece or in another EU country’, ‘Anywhere, even outside EU’. To combine the responses of both questions, we create a dummy variable where 0 indicates a ‘No’ answer to the willingness to move abroad to start working and the ‘Only inside Greece’ response. And, if the individual answers positively to the first question and in the second question chooses anything except for the ‘Only inside Greece’ answer, we consider that he/she is willing to migrate abroad to start working. The combination between these answers of the two separate questions is what can be expected to indicate the overall willingness of a person to move abroad to start working and so we have given it the value 1 in the dataset.

(22)

22

Explanatory Variables:

a) Socio-Demographic Variables

i. Level of Education: The main independent variable we want to look at in order to add to the information of the Greek brain drain in the recent years, is the education level of the individuals which is also the most common way of operationalizing skill level in the literature, especially from the human capital theory perspective. Three measurement methods of education level are used in the survey following the International Standard Classification of Education – ISCED 2011. We use the measurement that codes the responses ‘Did not go to school’, ‘Did not complete primary education level’, ‘Primary education’, ‘Lower secondary education’, ‘Higher secondary education’, ‘Post-secondary vocational education training’, ‘Bachelor’s education’, ‘Master’s education’, ‘Doctorate studies’ from 0 to 8 respectively. We then create a dummy variable for the highly educated which we define as the individuals with tertiary education attainment. This refers to education level attained after high school. Therefore, cases from 5 to 8 are valued as 1 and all the rest take value 0.

ii. Field of Education: We categorize field of education by using the ISCED 2011 Classification of field of education: General Programs, Education, Humanities and Arts, Social Sciences, Business/ Law, Science/Computing, Engineering/Manufacturing/Construction, Agriculture, Health and Welfare, Services. We use General Programs as the base category.

iii. Region of Residence: In order to observe where the potential labour emigrants reside, we include the categorical variable of region of residence which has been measured with the usage of the NUTS classification (Nomenclature of territorial units for statistics) system which has been established by the European Union (EUR-Lex, 2018). Specifically, NUTS2 (which holds a classification population threshold between 800,000 and 3 million) is used which divides the Greek territory into 13 regions. These can be viewed on the figure below and include East Macedonia and Thrace, Central Macedonia, West Macedonia, Epirus, Thessaly, Ionian Islands, West Greece, Central Greece, Attica, Peloponnese, North Aegean, South Aegean, Crete. The region of Attica is used as the base category.

(23)

23

Figure 3. Greek Regions (European Union NUTS2 classification). Source: (Tsimbos, Kotsifakis, Verropoulou, & Kalogirou, 2011).

iv. Financial Independence: As income plays an important role in individuals’ decisions, and due to the lack of data on the respondents’ level of finances (as they are unemployed they do not respond to the questionnaire’s income questions), we take advantage of the question that asks about their main financial source. The main responses include the following sources: Work (past), Income from property, Income from other household members, Income from persons that are not household members, Benefits or allowances. Using the responses regarding the source of finance, we create a dummy variable that measures the individual’s financial independence. If the individual has been using income from his work experience, property, and benefits or allowances, then we code that entry as 1-financially independent because public financial support is also mostly contributory, while when the individual is dependent on other persons (family members or non-family members), we code that entry as 0-financially dependent.

v. Other socio-demographic characteristics: We also include other explanatory variables in our analysis in order to measure their effect on the dependent variable, controlling in this way for

(24)

24

their effect on the main relationship of interest. These include Age (18-34), Gender (1=Female, 0=Male), Nationality (1=Immigrant, 0=National), Marital Status (1=Single: Not married/ Widow/er, Separated, 0=Married/Civil Partnership).

b) Previous occupation determinants

i. Has worked in the past: This variable is a dummy and takes the value of 1 if the individual has worked before in his/her life otherwise it is valued as 0.

ii. Previous occupation: This nominal variable separates the responses according to occupational activity before unemployment. The responses we are studying here include: Working, and Was in full time education. We create a dummy variable with the value of 1 for the latter and 0 for the former.

iii. Previous Profession: If the individual was working just before unemployment, this variable nominally distinguishes amongst the previous profession: Managers, Professionals, Technicians/Associate Professionals, Clerical Support Workers, Skilled Agricultural/Forestry/Fishery Workers, Crafts and Related Trades Workers, Plant/Machinery Operators/Assemblers, and Elementary Occupations. The Elementary Occupations are set as the base.

c) Unemployment characteristics determinants

i. Regional unemployment Rates: In order to determine the effect of local unemployment rates on labour emigration we create a variable that registers the unemployment rate in each of the 13 regions (NUTS2). This registration is possible with the information about regional unemployment rates derived by the same survey and shown in the reports of the Hellenic Statistical Authority. Having translated the region of residence of the individual into the regional unemployment rate, we can then measure the impact of regional unemployment on labour emigration.

ii. Duration of unemployment: We create a duration of employment ordinal variable by coding the survey response ‘Expected to start working, unemployment duration less than 6 months’ as 0, ‘6-11 months’ as 1, and ’12 months or more’ as 2.

(25)

25

iii. Employment preference: This variable is a dummy and takes the value of 0 if the individual prefers working in a part-time position or the value of 1 if full time employment is preferred. iv. Type in employment preferred: This variable is a dummy and takes the value of 1 if the individual would prefer to work as an employee or the value of 0 of he/she would prefer to be self-employed.

v. Turned down a job offer: This variable indicates whether the individual while being unemployed has turned down a job offer (0=No, 1=Yes).

vi. Reason for job rejection: This variable is nominal and gives more information about the reason why the individual has turned down a job offer. It includes the following responses: Not convenient workplace, Not flexible working hours, There were not proper preconditions for career, The job did not correspond to the individual’s typical qualifications, Payment was not satisfactory.

d) Labour Market Policies

As we have explained in the theoretical framework, we expect that an unemployed person’s willingness to emigrate is affected by the reception of public support. The reception of public support is thus a moderator variable of the basic relationship and since its study could be informative for government policy making, we are including it in the empirical methodology. In the survey there are questions that allow the operationalization of both passive and active labour market policies.

i. Passive Labour Market Policies: The survey asks, ‘Are you registered at a public employment office?’ and the response could be ‘Yes, and receives benefit or assistance’, ‘Yes, and he/she does not receive benefit or assistance’, ‘No’. The dummy variable that we make out of these responses, for the purpose of indicating the reception of passive labour market benefits, takes a 1 value for the first response and 0 otherwise.

ii. Active Labour Market Policies: Regarding the participation in active labour market programs, the question in the survey that is relevant for its operationalization is the following: “During the last 12 months did you receive any support from OAED or another public agency (for example, university) in order to find a job?”. And, the answer is binary “Yes/No”. The following question in the questionnaire asks “What kind of support you consider more useful?”,

(26)

26

where the choice could be: ‘Advice on how to apply for a job/ Advice on how to find job vacancies/ Advice on education and training opportunities/ I got into a work program/ I got into an educational program/ Other/ I got support, but it was not useful’. The setup of this question allows us to group all of them except for the last one into an expression of the individual’s participation in an active labour market program (1=Yes/0=No).

iii. Type of Active Labour Market Policy: Active labour market policies applied include: Advice on how to apply for a job, Advice on how to find a job, Advice on education and training opportunities, Got into a training program, Got into a work program. The first three are grouped as an indication of advice reception to which we give the value of 0, whereas the last two indicate the participation in a training/work program and is given the value of 1.

iv: Usefulness of Active Labour Market Policy: The question on the survey that asks which type of program the individual has participated (the types mentioned above), offers also the choice of the following response: “I received support, but it was not useful”. This response is useful for us to create a dummy variable that measures the usefulness of the programs: the positive answer to this response is valued as 0, otherwise the response is valued as 1.

(27)

27

6

Empirical Methodology

viii. A Logistic Regression Model (MODEL A)

In order to test our hypotheses concerning the effect of individual characteristics, previous occupational status, and unemployment characteristics on labour emigration, we use a logistic regression model. This statistical method is widely used in studies with similar purposes (Assirelli, Barone, & Recchi, 2019; Ahn, De la Rica, & Ugidos, 1999; Labrianidis & Vogiatzis, 2013). But, apart from its broad application in similar studies, a logistic regression model is preferred for our analysis for various reasons. In this section we explain the basics of a logistic regression as well as the importance of odds ratios in order to clarify the method used to examine the truthfulness of our hypotheses.

Logistic regression models are used to analyse the relationship between multiple independent variables and a categorical dependent variable. The independent variables can be either continuous or categorical while the dependent variable is either dichotomous or multinomial, meaning that it consists of more than two categories. As our dependent variable ‘labour emigration’ is dichotomous we follow a binary logistic regression model or more specifically a multivariate logistic regression as we include many independent variables in the model. A logistic regression, contrary to a linear regression does not require some restricting conditions to hold for its application (Hyeoun Ae, 2013, p. 156). Specifically, logistic regressions can be applied for non-linear relationships between the dependent and the independent variables as they transform relationships in a non-linear logistic curve (Hyeoun Ae, 2013, p. 156). A logistic regression is a way of fitting a regression curve y=f(x) in the case that y is a binary variable coded 0/1. The formula of a logistic function in its simplest form looks like this:

𝑦 = 𝑒

𝛼+𝛽𝜒

1 + 𝑒𝛼+𝛽𝜒 =

1 1 + 𝑒−(𝛼+𝛽𝜒)

Where α and β determine the logistic intercept and slope (Hyeoun Ae, 2013, p. 156). The logistic regression, then, finds a fit for α and β and can be written as:

(28)

28

When α and β take the values of 0 and 1 respectively, the logistic function looks as shown in figure 4.

Figure 4 Graph of logistic curve where α=0 and β=1. (Hyeoun Ae, 2013)

What the logistic regression does is to calculate the log of odds of the dependent variable.

𝐿𝑜𝑔 ( 𝑝

1−𝑝) = 𝛽0+ 𝛽1𝐿𝑒𝑣𝑒𝑙_𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 + . . . + ε (1)

In the above equation, 𝑝

1−𝑝

is the odds which is “the ratio between probabilities: the

probability of an event favourable to an outcome and the probability of an event against the same outcome” (Sperandei, 2014, p. 14). 𝛽0 is the intercept and 𝛽1 is the coefficient of our

variable of interest, while Level_Education is the dummy explanatory variable that takes the value of 1 for the highly skilled individuals and 0 otherwise.

Although we use this equation to estimate the effect of the level of education on labour emigration, for the interpretation of the results we will receive we will use odds ratios (OR) because they are more appropriate for explanations. An OR as the term suggests is the ratio of odds, which is comparing the two odds relative to different events. As Hyeoun-Ae (2013) puts it, an OR is “a measure of association between an exposure and an outcome”. This simply means that the ratio represents the odds that an outcome (here labour emigration) will occur in the presence of a specific exposure (here high level of education), compared to the odds of that outcome occurring when the exposure is not present (low level of education). So, for two events A and B the OR is (Hyeoun Ae, 2013):

(29)

29 𝑂𝑅 (𝐴 𝑣𝑠 𝐵) = 𝑜𝑑𝑑𝑠 {𝐴} 𝑜𝑑𝑑𝑠 {𝐵} = 𝑝_𝐴 1 − 𝑃𝐴 𝑝_𝐵 1 − 𝑃𝐵

ix. A Logistic Regression with A Moderator or Interaction Effect (MODEL B)

When the effect of an independent variable on the dependent variable changes depending on the value of a third variable, it is said that an interaction effect exists (Jaccard, 2011, p. 10). This third variable is usually called a moderator. The most common way of modelling interactions in logistic regressions is by means of product terms. Therefore, the basic logistic regression in equation (1) transforms into the following equation:

𝐿𝑜𝑔 ( 𝑝

1 − 𝑝) = 𝛽0+ 𝛽1Level_Education + 𝛽2PLMP + 𝛽3ALMP +

+ 𝛽4Level_Education × PLMP + 𝛽5Level_Education × ALMP + ⋯ + ε (2)

Here PLMP stands for passive labour market policies and ALMP for active labour market policies. By introducing the interactions Level_Education × PLMP and Level_Education × ALMP, we try to see whether labour market policies have an impact on the main relationship of interest that is the effect of education level on labour emigration. As the interpretation of logistic regression coefficients is alone challenging, even more so is the interpretation of the coefficients in models with interaction terms. Again, for this purpose ORs will be used because they can offer a clearer comparison between the different events. In the Results chapter we will comment on the exact steps we followed to derive the coefficients, ORs, and tables presented. We have considered meaningful to explain the logic of the formulas of the logistic regression model. However, as we use the statistical software Stata for deriving the coefficients and ORs, we will not dive deeper in explaining the equations, but we suggest a look at the relevant bibliography for a better understanding (Kleinbaum & Klein, 2010; Sainani, 2014; Hyeoun Ae, 2013).

(30)

30

7

Empirical Findings

x. Descriptive Statistics

The average age of the respondents is 27 years with the minimum age being 18 and the maximum 34 years old while almost 53% of the participants are female. The big majority (93%) are nationals and mostly (80%) not married. Regarding education, around 46% of the respondents are tertiary education graduates, therefore highly educated. Most of the respondents were educated in general programs for which we have no specific information and the next biggest groups consist of social science educated individuals (11%) and science, and computing majors (16%). Also, the respondents are spread around the whole country but most of them reside in Eastern Macedonia (11%), Central Macedonia Region (18%), and Attica (21%). They are almost equally spread across the three levels of urbanization areas, with similar size groups (around 30%) residing in rural areas, towns and suburbs, and cities. And, significant is the fact that the young unemployed persons are observed to be greatly financially dependent on others. More specifically, only 1,4% use financial resources gained through salary, while 75% get financially supported from other household members and 14% receive financial support from persons that are not household members. Table 1 below summarizes the main socio-demographic variables descriptive statistics.

Table 1

Descriptive Statistics: Main Socio-Demographic Variables

Variable Mean Standard Deviation

Age 27.206 4.329 Gender .5261 .4994 Nationality .0688 .2532 Marital Status .7868 .4096 Level of Education .4629 .4987 Financial Independence .0146 .1202

(31)

31

As far as previous occupational status of the unemployed is concerned, more than half (58%) of the individuals have worked at least once in their life and the majority had either been in employment or students just before being unemployed. From the ones that were previously employed their profession was mostly in the field of services and sales (35%) and the second most common profession was that of clerical support workers (17%). Finally, a small minority were self-employed in their last work with the 96% having been employees.

Moreover, the unemployed individuals differ in their unemployment characteristics. The majority (65%) of the respondents have been unemployed for more than a year and have been looking for a job for that long. The vast majority of the unemployed persons (98%) would like to be full-time employed and they mostly (98%) prefer to be employees rather than self-employed. Few have rejected a job offer and the most common reason has been the non-convenience of the workplace, the not-flexible working hours, and the not-satisfactory payment. Most of the survey participants, around 75%, have contacted and registered at a Public Employment Service. Among them 7% are beneficiaries of passive labour market policies and financial assistance while 14% have participated in active labour market policies and have benefited from public labour market programs. The most common such programs have been advice on how to find a job vacancy and advice on how to apply for a job vacancy. It is noteworthy that 73% of the active labour market policies beneficiaries find them useful. Regarding the willingness to migrate abroad for work, we observe that 27,5% of the unemployed persons have responded positively. In real terms, out of the 2,055 cases in the study, 565 persons are willing to emigrate for work comparing to 1,490 others who do not appear to share this approach. This group is not heterogeneous in its composition with respect to the level of education. It appears that half of this group is highly educated, and the other half not highly educated. However, if we look within the group of the highly educated 30% of them are willing to emigrate for work abroad compared to a 25% expression of that will among the not highly educated. Then looking within the different educational fields, we observe that in the group with a background in business, and law there are mostly potential emigrants. Also, there is only a slight difference in age between the potential emigrants (26.5 years old) and the not willing to emigrate for work persons (27,5 years old). The potential labour emigrants are in their majority male (56%), single/not married (90%), and nationals (92%). The degree of urbanization of the residence region of the potential labour emigrants is slightly higher than the rest. This means that more city residents have responded positively to the willingness to the

(32)

32

‘emigrate for work’ question while the most come from the region of Central Macedonia and Attica.

Apart from the description of the individual characteristics of the persons that are willing to emigrate for work abroad, it is important to observe this group in terms of their previous occupational status and their unemployment characteristics. The group consists mainly of people who have worked at least once in their life while looking at the responses of the participants from a view of their previous profession, it is apparent that among the previous technicians and associate professionals the response towards labour emigration is mostly positive (28 out of 54), compared to the mostly negative responses within the rest of the professions. The duration of unemployment between the two groups is similar but we can see a slightly lower search time among the not willing to emigrate respondents.

Both the potential labour emigrants and the rest of the survey participants are registered in a public employment service, benefit from passive labour market policies and active labour market policies. From those receiving passive labour market policy support 19% are willing to emigrate and from the active labour market programs beneficiaries 24% are willing to emigrate. And, among those that find the active labour market programs useful 19% are willing to emigrate while among those that do not find them useful 38% are willing to emigrate.

(33)

33

xi. Logistic Regressions Results

Model A

The application of Model A of the logistic regressions, without the interaction terms, generates the results that we can see in Table 2. We run four logistic regressions using the main predictors of the four categories of explanatory variables in this study, namely: the socio-demographic/individual characteristics, the previous occupational experience of the individuals, the unemployment characteristics, and the labour market policies. We proceed with the presentation of these findings.

As it appears from Table 2, the ORs of the ‘Level of Education’ variable are very similar and significant in all four regressions. Focusing on the first basic regression, we see that the OR of this variable equals 1.486, which means that the odds of labour emigration for individuals with high level of education are around 1,5 times higher than the lower educated individuals or, in other words, highly educated individuals are 50% more likely (in terms of odds, not probability) to emigrate for work abroad than the individuals with lower levels of educational attainment.

Looking at the other socio-demographic characteristics our findings are also statistically significant in their majority. Regarding the effect of age, the OR appears to be a bit less than 1 which shows that with one year of addition in age, the odds of labour emigration are lower. However, age is a continuous variable and the presentation of the result in this case requires special attention. Here, what we observe is the average change in odds with a one-year addition in age. To get a deeper understanding, one needs to either choose a specific age and look at the probability of labour emigration at this age, or a dummy variable creation can also be informative. We ran an extra logistic regression where we used age as a dummy, where 27 years or younger takes a value of 1, otherwise a 0. What we found is that other variables’ ORs remain similar, while the group of 27 years old or younger is 20% more likely to emigrate for work than the older ones. For the gender determinant we find that the odds of women to emigrate for work are around 35% lower than the odds of men emigrating. Nationality also plays a significant role, where an immigrant is 55% more likely to emigrate again. Also, the individuals that are not married are 2.3 times more likely to emigrate for work than the married ones. The degree of urbanization appears also to matter, with city persons being more prone (26%) to emigrate than the ones living in less urbanized regions. Finally, financial

(34)

34

independence’s role appears quite high but not significant, with around 82% higher likelihood of emigration for the financially independent.

In regressions 2 and 3, the additions of variables concerning previous employment effects and unemployment features effects appear to have no statistical significance. Having worked before lowers the intention to emigrate. Also, local unemployment rate does not seem to affect labour emigration, neither does unemployment duration, as we observe. Additionally, those that prefer a part time job are more likely (OR≈ 10%) to emigrate than the ones who prefer a full time job, and those that have rejected a job offer are more (OR ≈ 22%) prone to emigration than the ones who have not rejected an offer.

The addition in regression 4 of the labour market policies variables show that labour emigration is less likely for the recipients of these policies. However, only passive labour market policies seem to have a significant effect. Specifically, the odds of labour emigration for passive labour market policies recipients are 43% less than the ones that are non-beneficiaries. For active labour market policies, the respective likelihood is around 18% and not statistically significant.

(35)

35

Table 2

MODEL A: Logistic Regressions of Labour Emigration

seEform in parentheses *** p<0.01, ** p<0.05, * p<0.1

Estimated Odds Ratio

(1) (2) (3) (4) VARIABLES OR OR OR OR Level of education 1.486*** 1.484*** 1.500*** 1.532*** (0.169) (0.168) (0.174) (0.181) Age 0.970** 0.978 0.977 0.974* (0.0129) (0.0140) (0.0149) (0.0151) Gender 0.648*** 0.644*** 0.669*** 0.669*** (0.0713) (0.0710) (0.0753) (0.0765) Nationality 1.554** 1.543** 1.526* 1.503* (0.335) (0.332) (0.333) (0.337) Marital status 2.336*** 2.307*** 2.350*** 2.268*** (0.398) (0.394) (0.409) (0.402) Degree of urbanization 1.255*** 1.259*** 1.250*** 1.251*** (0.0821) (0.0824) (0.0838) (0.0851) Financial independence 1.817 1.967* 1.896 1.623 (0.727) (0.794) (0.787) (0.696)

Has worked before 0.853 0.848 0.940

(0.0986) (0.103) (0.117)

Local unemployment rate 1.009 1.010

(0.0101) (0.0102)

Duration of unemployment 0.992 0.962

(0.0731) (0.0736)

Employment preference 0.892 0.934

(0.304) (0.322)

Has rejected a job offer 0.778 0.810

(0.203) (0.214) Contacted a public employment service 0.942 (0.118) PLMP 0.574** (0.150) ALMP 0.822 (0.132) Constant 0.332** 0.299*** 0.318** 0.369* (0.144) (0.132) (0.178) (0.209) Observations 1,986 1,986 1,897 1,850