The role of regional differences in immigration: the case of OECD countries

190 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 THE ROLE OF REGIONAL DIFFERENCES IN IMMIGRATION: THE CASE OF OECD COUNTRIES Domicián Máté, University of Debrecen, Faculty of Economics and Business, Debrecen, Hungary, E-mail: [email protected] Imran Sarihasan, University of Debrecen, Károly Ihrig Doctoral School Debrecen, Hungary, E-mail: [email protected] József Popp, University of Debrecen, Faculty of Economics and Business, Debrecen, Hungary, E-mail: [email protected] Judit Oláh, University of Debrecen, Faculty of Economics and Business, Debrecen, Hungary, E-mail: [email protected] Received: April, 2018 1st Revision: May, 2018 Accepted: August, 2018 DOI: 10.14254/2071789X.2018/11-3/12 ABSTRACT. Despite physical, cultural, economic and other obstacles, millions of people have recently emigrated from one country to another in search of a better life. Consequently, the importance of this research topic has grown over time. Analysis of this regional approach is based on the OECD censuses, and its unique Database on Immigrants (DIOC), which makes it possible to generate a wide variety of cross-tabulations on the characteristics of asylum-seekers by the country of their birth. The aim of this study is to contribute to literature by analysing the role of regional differences in migration more closely in terms of the importance of migration regulation policies in the OECD countries. According to the results based on Binary Logistic (Logit) and Linear (OLS) Regression models, the educational attainment, age, sex and the place of birth of migrants are related in different ways to their region of origin. Moreover, male migrants are more skilled than females, migrants who are more educated are older than the less skilled, and foreign-born migrants seemed to be older than native-born migrants. There is also additional evidence suggesting that different regions and countries might follow different policies and norms for admission of migrants. Therefore, forthcoming migration governance programs that aim to facilitate the labour market integration of migrants should also take into consideration their regional characteristics. JEL Classification : K37, J21, J15 Keywords: migration; regionalism; educational attainment, aging, OECD. Máté, D., Sarihasan I., Popp, J., & Oláh, J. (2018). The Role of Regional Differences in Immigration: The Case of OECD Countries. Economics and Sociology, 11(3), 190-206. doi:10.14254/2071-789X.2018/11-3/12 191 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 Introduction Regions of the world account for different proportions in the global number of migrants. The origin of international migrants has become increasingly diversified over the past two decades (United Nations, 2017). Asia and Europe hosted over 60% of all international migrants worldwide in 2017, with nearly 80 million living in Asia and 78 million in Europe. North America hosted the third largest number of international migrants (58 million), followed by Africa (25 million), Latin America and the Caribbean (10 million), and Oceania (8 million) (UN, 2017). Asia, Europe and North America accounted for over 85% of the increase in the number of international migrants between 1990 and 2017. Thus, between 2010 and 2017, Africa experienced the second fastest annual growth, with the average of 1.1 million per annum (Figure 1). Figure 1. Average annual change in the number of international migrants by major regions of origin, 1990-2017 (millions) Source: Authors’ own compilation based on United Nations (UN, 2017). The latest wave of immigration in the post-crisis period (2010-2015) is unlike previous waves in terms of skill and origin structure. Burzynski, Docquier, & Rapoport (2018) have shown that the welfare effects of immigration are very heterogeneous across skill groups, countries of destination, and immigration waves. For instance, the latest wave period of immigration is less skilled and brings fewer benefits. The welfare impact of recent migration flows have been beneficial for 69% of the non-migrant OECD residents, and for 83% of nonmigrant citizens in 22 richest OECD countries. The leading beneficiaries are mainly traditional immigration countries; their gains are extensive due to the entry of immigrants from non-OECD countries (Aubry, Burzyński, & Docquier, 2016). There are various economic, social and physical reasons why people emigrate and they can usually be classified into push (negative incentives) and pull (positive) factors (Bodvarsson & Van den Berg, 2013). The main economic reason for choosing to be a migrant is to seek an improved life by finding a job, and the driving forces of migration are higher wages, human security and work opportunities (Castles, 2013). The economic push factors tend to be the exact opposite of the pull factors that are associated with the area of destination (overpopulation, insufficient number of jobs, low wages etc.). However, other social reasons, i.e. civil and political rights, religious freedom, law and order, social mobility, personal 192 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 safety, and peace, together with other attractive environmental factors (climate, seaside location etc.) all make a destination country attractive to potential immigrants. These economic, political and social incentives exist worldwide, thus influencing international migration over recent decades (Ozgur & Deniz, 2014). The economic characteristics of migrants and their motivations for moving vary by age group and by migrant status (Philip, Macleod, & Stockdale, 2013). Several studies (McInnis 1971; Millington, 2000 etc.) demonstrate there is a positive relationship between the age of responsiveness and the economic gains from migration. Hence, younger migrants have great adaptability to new challenges when entering the labour market of a destination country. Moreover, regional differences - including wage and age disparities etc. - can affect migration as well (Hunt, 2012). Essentially, people migrate from low-income to high-income regions, which increases the mobility cost of migration. This presumption focuses almost exclusively on the size and direction of population flows across regions (Borjas, Bronars, & Trego, 1992). Thus, the relationship between migration, age, education, segregation and distance have been discussed in many economic studies (Schwartz, 1976;, Mueser, 1989; Iceland, 2017 etc.). Although migrants typically transfer from lower to higher income regions, migration flows increase with higher education and reduce with aging. Nevertheless, the economic impact of migration depends on the number and the education level of migrants and also on how well education of migrating individuals matches their future occupations (Quinn & Rubb, 2005). The skilland origin-mix of migration to the OECD has evolved over time. Meanwhile, many OECD countries have been attempting to attract qualified human resources from abroad, since knowledge-based and progressive economies need to sustain their economic growth (Dumont & Lamaitre, 2005). It is beneficial for these countries to acquire migrants who have solid education and well-developed skills in order to improve their economic performance, since it has been widely acknowledged that education and skills together are the major sources of economic prosperity (Flisi, Meroni, & Vera-Toscano, 2016). These intentions are highly dependent on the time frame involved, i.e. education has a greater influence on long‐term migration decisions (Williams et al., 2018). Moreover, the influence of migration may lead to changes in the average levels of education in destination countries and generate educational externalities and new incentives for human capital investments (Dustmann & Glitz, 2011). Furthermore, each migrant is a source of personal consumption expenditure, which is also transferred among the regions. Di Giovanni, Levchenko, & Ortega (2015) compared welfare with levels of migration from the cross-country global analysis perspective, and found that natives in those countries (Canada, Australia) that received high levels of migration are better off due to a greater product variety available in consumption and as intermediate inputs. Meanwhile, in the short run, skilled and unskilled natives tend to experience welfare changes with opposite signs. Initial studies have discussed the process of understanding the relationships between trade, migration, and regional changes (Fournier, 1989). Meanwhile, education systems and social policies are not similar across regions, while socioeconomic status and ethnocultural arrangements of immigrants diverge considerably. Such characteristics of migration may be associated with regional differences in the education outcomes. Migration also has economic, social and cultural consequences for both the origin and the destination countries. This phenomenon is signalled by the absence of social barriers, participation in mainstream institutions, intermarriage with host country natives, and degrees of self-identification by immigrants as natives (Beck, Corak, & Tienda, 2012). The aim of this study is to contribute to literature by analysing the role of regional differences of migration more closely in terms of the importance of migration regulation policies in the OECD countries. The significance of our regional perspective is to explain the complexity of migration characteristics in different societies, taking into account spatial and 193 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 geographic aspects. Considering the consequences of population distribution, age, and gender structure etc. can be compatible with the integration purposes of migration policies (Findlay & Mulder, 2015). Assuming also that common migration policies do not work properly, there is no unique solution for social and economic problems of migration in terms of inequalities. The effects of these inequalities in migration on the places of origin and destination are also influenced by a process called migration selectivity. Hence, certain individuals are more likely to migrate, based on their personal and sociodemographic characteristics. The rest of this paper is organized as follows. The next (1) section summarizes the brief conceptual framework of migration theories with a regional perspective. Then, in section 2, logistics (Logit) and linear (OLS) regression analyses are carried out with cross-country statistics in order to investigate how gender, aging, place of birth, and educational attainment differences are related to regional characteristics of migrants. Finally, based on these results, the paper ends with some policy implications and conclusions (Section 4). However, our motivation is not only to suggest a feasible point of reference for researchers, to enhance the effectiveness of policies, to reduce the negative consequences of modern slavery, but also to outline further research directions in this international perspective related to the quality of migration regulation across countries in the interests of human development over time. 1. Literature review The notion of migration is commonly understood as a movement from one location to another for the purpose of either temporary or permanent settlement. This may involve longdistance traveling, across or within the borders of a country from one region to another with a distinctly different cultural, political or social etc. environment (Biswas, McHardy, & Nolan, 2005). The development of migration occurs when an individual decides that it is preferable to move rather than to stay, and where the difficulties of moving seem to be more than offset by the expected rewards. Population movements are responses to distinct conditions and related incidents, which people encounter both in places of origin and destination. The early theories of migration were mostly focused on domestic migration, and were closely related to location models from regional economics and the economics of geography. E. G. Ravenstein first developed the ‘law of migration’, which classified migrants by the distance of relocation (Ravenstein, 1885). According to Ravenstein’s laws of migration, most migrants move only a short distance and usually to large cities that grow rapidly and tend to be populated by migrants from nearby rural areas. Thus women are more likely to migrate than men, the out-migration is inversely related to in-migration, and a major migration wave will generate a compensating counter wave (Greenwood, 1997). In this early approach, local migrants moved within the country, or only from the country of their birth to border countries (Koser & Salt, 1997). Therefore, a series of analyses attempted to explain and predict migration patterns both within and between nations. The basic and additional laws of migration subsequently emerged to serve as the starting point for all serious models of migration patterns. One of these laws relating to the gender gap asserted that short-distance migrants far outnumbered longer distance ones, and women dominated within countries. The outcome of these tendencies led to an inverse relationship between the volume of migration and the distance between the home and destination countries. The theoretical background to international migration is mainly grounded in the consumption approach favoured by urban and regional economists and which focuses primarily on domestic migration. Greenwood (1993) points out that by the early 1980s, the ‘equilibrium’ perspective on migration suggested people migrate to take advantage of regional income differences. The elementary idea behind this perspective is that people migrate to adjust their consumption to frequent changes in incomes, prices, the supply of 194 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 goods and services and their utility functions etc. (Rosen, 1974). The notion that people migrate in response to spatial differences in amenities was also proposed by Tiebout (1956), who argued that an important factor explaining why people move from one locality to another is differences in the quality of public goods, such as police and fire protection, education, hospitals, courts and other facilities. Nevertheless, the idea has not been applied to the studies of international migration in terms of its regulated nature and the relatively higher costs of international movement. Thus, the empirical results from national analyses only contributed to the verification of migration theories, and to identifying potential differences between (inter)national determinants of migration, so as to become the groundwork for regional economic policies (Jandová & Paleta, 2015). Nevertheless, according to the ‘push and pull’ general theory of migration, immigration is determined by a summarized comparison of a wide range of positive factors in destination countries and various negative ones in the original place of residence (Rahmandoust, 2011). In other words, migration is selective and tends to consider positive (pull) factors which occur in the migration destination attracting migrants to the country of destination (Kainth, 2009). Moreover, negative (push) reasons also occur, which in some circumstances drive migrants to leave their countries of origin to destination regions. Meanwhile, in the sociological literature on migration, the role of the community of family and friends at the destination is often referred to as a relationship. The banding together of previous migrants from a similar ethnic or regional background is also referred to as a migrant network, based on the linguistic similarity between specific areas (Chiswick & Miller, 2014). Evidence also supports the idea that immigrants tend to concentrate where earlier nationals have settled because the cost of adapting to a novel society is mitigated by the presence of residents familiar with both the source and destination country cultures. The level of spatial mobility is larger, ceteris paribus, when the verbal and cultural environment in the destination country is familiar (Rephann & Vencatasawmy, 2000). In our opinion, no unique immigration theory has yet been devised which can completely explain regional migration disparities. Thus, there is still a big gap between theory and empirical evidence, and much needs to be done on the theoretical side of this literature, as well. One of the greatest challenges to migration theorists is to organize all hypothetically relevant factors into one coherent framework that will specify their interaction with each other in an empirically testable form and thereby serve as a guide for future research. Consequently, more studies are needed using many econometric models, data sets, time periods, countries, and replications of existing studies before economists can have enough confidence in the results of their own statistical studies of immigration. 2. Data and methodological approach This research uses the Database on Immigrants in OECD Countries (DIOC) which provides detailed information on the census data of immigrants (OECD, 2018). This unique database makes it possible to generate an extensive variety of cross-tabulations on the characteristics of the migrant inhabitants in OECD countries by their country of birth. The latest available release includes a specific area covering age, gender, nationality, place of birth and duration of stay etc. The estimations are mainly based on binary logistic (Logit) regression models (see Equation 1, 2, 3, 4, 5 and 6) that are frequently used to apply to a binary dependent variable of different regions. The Logit regression method was first developed by Cox (1958) to estimate the probability of a binary response based on one or more predictor (or independent) variables. In order to analyse whether educational attainment, age, sex and the place of birth 195 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 of migrants relate differently to their region of birth the following models are tested in each case for the migrants observed i: 𝐷𝐴𝑓𝑟𝑖𝑐𝑎𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐴𝑔𝑒𝑖+ 𝛽3𝐷𝑆𝑒𝑥𝑖+ 𝛽4𝐷𝐹𝑏𝑜𝑟𝑛𝑖+𝑖 (1) 𝐷𝐴𝑠𝑖𝑎𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐴𝑔𝑒𝑖+ 𝛽3𝐷𝑆𝑒𝑥𝑖+ 𝛽4𝐷𝐹𝑏𝑜𝑟𝑛𝑖+𝑖 (2) 𝐷𝐸𝑢𝑟𝑜𝑝𝑒𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐴𝑔𝑒𝑖+ 𝛽3𝐷𝑆𝑒𝑥𝑖+ 𝛽4𝐷𝐹𝑏𝑜𝑟𝑛𝑖+𝑖 (3) 𝐷𝑁_𝐴𝑚𝑒𝑟𝑖𝑐𝑎𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐴𝑔𝑒𝑖+ 𝛽3𝐷𝑆𝑒𝑥𝑖+ 𝛽4𝐷𝐹𝑏𝑜𝑟𝑛𝑖+𝑖 (4) 𝐷𝑂𝑐𝑒𝑎𝑛𝑖𝑎𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐴𝑔𝑒𝑖+ 𝛽3𝐷𝑆𝑒𝑥𝑖+ 𝛽4𝐷𝐹𝑏𝑜𝑟𝑛𝑖+𝑖 (5) 𝐷𝑆&𝐶_𝐴𝑚𝑒𝑟𝑖𝑐𝑎𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐴𝑔𝑒𝑖+ 𝛽3𝐷𝑆𝑒𝑥𝑖+ 𝛽4𝐷𝐹𝑏𝑜𝑟𝑛𝑖+𝑖 (6) The dependent variables are dummies distinguished by the migrants’ different region of birth, i.e. DAfrica is = 1 if the migrant is born in Africa, 0 = otherwise. Respectively, DAsia refers to Asia, DEurope to Europe, DN_America to North America, DOceania to Oceania, and DS&C_America indicates South and Central America. The first independent variable is (Edu), where the detailed educational levels (1-6) are based on the International Standard Classification of Education (ISCED) as defined by UNESCO (2011). The next control variable is (Age), which is divided into the following age group categories: 1 = 0-14; 2 = 15-24; 3 = 25-34; 4 = 35-44; 5 = 45-54; 6 = 55-64; 7 = 65+. The (DSex) dummy notes the gender differences; namely, 1 if the migrant is female, 0 if male. The subsequent control dummy variable (DFborn) indicates whether the migrant is foreignor native-born. According to the OECD (OECD, 2017), foreign-born immigrants include all those who have ever migrated from their country of birth to their current country of residence. Thus, a native-born citizen of a country is one whose migrant parents’ place of birth is the host country. (ε) is the error term. Based on the previously reported (1-6) Equations, the current study forms the following hypotheses: H1: African migrants are less likely to be educated, older, foreign-born and female than other migrants. H2: Asian migrants are less likely to be educated, older, foreign-born and female than other migrants. H3: European migrants are less likely to be educated, older, foreign-born and female than other migrants. H4: North American migrants are less likely to be educated, older, foreign-born and female than other migrants. H5: Oceanian migrants are less likely to be educated, older, foreign-born and female than other migrants. H6: South and Central American migrants are less likely to be educated, older, foreign-born and female than other migrants. In order to exemplify the validity of our results, and verify that educational attainment, and age of migrants relate differently to their region of birth, additional methodologies are needed to clarify the regional characteristics of migration. For instance, linear OLS regressions. Consequently, a linear (OLS) regression model is used to highlight the regional differences among our evaluations. 196 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 𝐸𝑑𝑢𝑖= 𝛽𝑜+ 𝛽1𝐴𝑔𝑒𝑖+ 𝛽2𝐷𝑆𝑒𝑥𝑖+ 𝛽3𝐷𝐹𝑏𝑜𝑟𝑛𝑖+ 𝛽4𝐷𝐴𝑓𝑟𝑖𝑐𝑎𝑖+ 𝛽5𝐷𝐴𝑠𝑖𝑎𝑖+ 𝛽6𝐷𝐸𝑢𝑟𝑜𝑝𝑒𝑖+ 𝛽7𝐷𝑁_𝐴𝑛𝑒𝑟𝑖𝑐𝑎𝑖+ 𝛽8𝐷𝑂𝑐𝑒𝑎𝑛𝑖𝑎𝑖+𝑖 (7) 𝐴𝑔𝑒𝑖= 𝛽𝑜+ 𝛽1𝐸𝑑𝑢𝑖+ 𝛽2𝐷𝑆𝑒𝑥𝑖+ 𝛽3𝐷𝐹𝑏𝑜𝑟𝑛𝑖+ 𝛽4𝐷𝐴𝑓𝑟𝑖𝑐𝑎𝑖+ 𝛽5𝐷𝐴𝑠𝑖𝑎𝑖+ 𝛽6𝐷𝐸𝑢𝑟𝑜𝑝𝑒𝑖+ 𝛽7𝐷𝑁_𝐴𝑛𝑒𝑟𝑖𝑐𝑎𝑖+ 𝛽8𝐷𝑂𝑐𝑒𝑎𝑛𝑖𝑎𝑖+𝑖 (8) Note that the South and Central American region variable was omitted in order to control and avoid the ‘dummy-trap’ problem. The dummy variable trap is a phenomenon in which the independent variables are multicollinear and two or more variables are highly correlated, i.e. one variable can be predicted from the others. Moreover, based on Equations 7 and 8, the following forms of the hypotheses can be also verified: H7: Female migrants are less educated than males. H8: Foreign-born migrants are less educated than native-born migrants. H9: Migrants that are more educated are older than less educated migrants. H10: Male migrants are older than females. H11: Foreign-born migrants are younger than native-born migrants. 3. Conducting research and results Table 1, 2 and 3 represent the corresponding results of our estimations in each model. In order to take account of the effect of regional differences we focus separately on migrants by region of birth and assume that they have different educational, age, gender and origin of birth characteristics compared to the other regions. In all observed logistic (Logit) regression models (see Tables 1 and 2), the dependent variable indicates the migrants’ region of birth. At the bottom sections of these tables are the proportion of variances explained by the predictors (measured by Cox and Shell’s, and Nagelkerke’s pseudo R2). However, the R2 statistics of these models are relatively small, thanks to a large number of observations. According to the Omnibus (F-test) and HL-tests, the explained variances in a set of data are significantly greater than the unexplained variance. In the case of Africa and Europe, we found a significant negative relationship between the educational attainment and the region of birth. Although the effect of education does not seem to be large in North, South and Central America and Oceania, in these models there was a positive correlation. In other words, African, Asian and European migrants tend to be less educated than others. In the case of Africa, Asia, Oceania, and South and Central America age negatively associated with region of birth variables, which indicates that in these regions migrants are more likely to be younger than the other regions. Nevertheless, if there are two migrants and one of them comes from Europe or North America, he/she will be significantly older than the other. Moreover, gender differences can be found in only a few cases. African migrants tend to be male, with European and South and Central Americans tending to be female. Measuring the origin of birth differences, we can also claim that Asian, and South and Central American migrants are significantly more likely to be foreign-born. From the results it can be claimed that North Americans asylum seekers and migrants from Oceania tend to be native-born. Although, the strong H1, H2, H3, H4, H5 and H6 hypotheses should be rejected, based on the results it can be claimed that the educational attainment, age, sex and the place of birth of migrants relate differently to their region of birth. 197 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 Table 1. Results of the binary logistic (logit) regressions in the examined OECD countries in Africa, Asia and Europe Dependent DAfrica DAsia DEurope Independent Beta Wald EXP(B) Beta Wald EXP(B) Beta Wald EXP(B) Constant -20.98 0.001 0 -4.281 1072.6*** 0.013 1.493 1259.2*** 4.452 Edu -0.011 31.6*** 0.989 0.001 0.238 1.001 -0.008 21.2*** 0.991 Age -0.027 215.9*** 0.974 -0.012 40.8*** 0.988 0.038 539.3*** 1.039 DSex -0.048 50.4*** 0.953 -0.005 0.471 0.995 0.016 6.7*** 1.016 DFborn 20.011 0.001 0 3.138 582.6*** 23.04 -2.503 3893.2 0.081 Observations 491813 Cox and Shell R2 0.06 0.04 0.13 Nagaike R2 0.08 0.06 0.18 Omnibus-test 2766.5*** 1984.6*** 6346.1*** HL-test 103.2*** 32.87*** 212.6*** Source: Authors’ own compilation, based on (OECD, 2018) Notes: Heteroscedasticity robust Wald-statistics are in parentheses. Letters in the upper index refer to significance: ***: significance at 1 per cent, **: 5 per cent, *: 10 per cent. P-values without an index mean that the coefficient is not significant even at the 10 per cent level. HL: Hosmer and Lemeshow χ2 test. Table 2. Results of the binary logistic (logit) regressions in the examined OECD countries in North America, Oceania and South and Central America Dependent DN_America DOceania DS&C_America Independent Beta Wald EXP(B) Beta Wald EXP(B) Beta Wald EXP(B) Constant -3.417 1423.1*** 0.033 -2.755 1719.3*** 0.064 -3.075 1870.3*** 0.046 Edu 0.017 6.6** 1.017 0.029 56.4*** 1.030 0.013 33.2*** 1.013 Age 0.034 33.1*** 1.035 -0.010 8.2*** 0.990 -0.010 23.6*** 0.990 DSex -0.022 0.995 0.978 -0.001 0.003 0.999 0.046 35.6*** 1.047 DFborn -0.834 113.9*** 0.434 -0.246 16.3*** 0.782 1.390 402.1*** 4.016 Observations 491813 Cox and Shell R2 0.01 0.01 0.01 Nagaike R2 0.02 0.02 0.02 Omnibus-test 134.3*** 76.98*** 711.5*** HL-test 9.123* 97.25*** 7.854* Source: Authors’ own compilation, based on (OECD, 2018) Notes: Heteroscedasticity robust Wald-statistics are in parentheses. Letters in the upper index refer to significance: ***: significance at 1 per cent, **: 5 per cent, *: 10 per cent. P-values without an index mean that the coefficient is not significant even at the 10 per cent level. HL: Hosmer and Lemeshow χ2 test. 198 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 Measuring the effects of aging, gender, and region of birth differences on educational attainment of migrants (in Models 1 and 2), based on Equation 7, can also provide efficient empirical tools for regional migration policy reforms. Thus, using Equation 8 in our regression models (3 and 4), the influence of these examined variables on the age of immigrants is also tested in a regional context. In the case of Models 2 and 4 we only add the variables which were significant in Models 1 and 3; the others are omitted. However, the adjusted R2 values are quite low because of the large number of observations (491,813); the significant F-test statistics suggest that our linear regression specification should be preferred in all models. The multi-collinearity amongst the independent and control variables is tested by the variance inflation factor (VIF) in each case. The maximum values of VIF for each regression coefficient range from a low of 1.181 to a high of 1.971. This suggests that the VIF values are at acceptable (less than 10) levels (Hair, 2010). At the bottom section of Table 3, as a goodness of fit (GOF) test of our regression the Kolmogorov–Smirnov (K-S) normality tests of the non-standardized residuals are also reported, to check one of the assumptions of the linear regression. Table 3. Results of the OLS regressions of Equations 7 and 8 in the OECD countries examined Independent variables Model 1 Model 2 Model 3 Model 4 Constant 3.163*** 3.172*** 3.885*** 3.894*** 117.981 306.573 181.981 136.407 Edu 0.081*** 0.081*** 50.744 50.737 Age 0.064*** 0.064*** 50.744 50.857 DSex -0.034*** -0.034*** 0.007 (-7.115) (-7.148) 1.229 DFborn 0.008 -0.101*** -0.101*** 0.321 (-3.612) (-3.611) DAfrica -0.057*** -0.058*** -0.043*** -0.041*** (-7.508) (-7.891) (-4.946) (-6.002) DAsia -0.031*** -0.031*** -0.002 (-4.029) (-4.293) (-0.236) DEurope -0.049 -0.051*** 0.125*** 0.126*** (-6.831) (-7.278) 15.293 19.981 DN_America 0.014 0.148*** 0.150*** -0.71 6.861 7.13 DOceania 0.045*** 0.044*** -0.003 3.706 -3.669 (-0.252) Observations 491813 Adj. R2 0.06 0.06 0.07 0.07 F-test 344.18*** 460.77*** 407.21*** 651.22*** VIF 1.971 1.821 1.971 1.181 Normality test of the non-standardized residuals K-S test 0.123*** 0.123*** 0.083*** 0.083*** Source: own compilation based on (OECD, 2018) Notes: Heteroscedasticity robust t-statistics are in parentheses. Letters in the upper index refer to significance: ***: significance at 1 per cent, **: 5 per cent, *: 10 per cent. P-values without an index mean that the coefficient is not significant even at the 10 per cent level. From the regional perspective, the age and educational attainment related differently to the region of birth. For example, as we found previously, African, Asian and European 205 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 UN. (2017). United Nations Statistics Division - Classifications Registry, ISIC Rev. 4. UNESCO. (2011). International Standard Classification of Education (UIS/2012/INS/10/REV). UNESCO Institute of Statistics. United Nations. (2017). International Migration Report 2017. van Dalen, H. P., & Henkens, K. (2012). Explaining low international labour mobility: the role of networks, personality, and perceived labour market opportunities. Population, Space and Place, 18(1), 31-44. https://doi.org/10.1002/psp.642 Williams, A. M., Jephcote, C., Janta, H., & Li, G. (2018). The migration intentions of young adults in Europe: A comparative, multilevel analysis. Population, Space and Place, 24(1), e2123. https://doi.org/10.1002/psp.2123 Zaiceva, A. (2014). The impact of aging on the scale of migration (99). IZA Workd of Labor. https://doi.org/doi: 10.15185/izawol.99 206 Domicián Máté, Imran Sarihasan, József Popp, Judit Oláh ISSN 2071-789X RECENT ISSUES IN SOCIOLOGICAL RESEARCH Economics & Sociology, Vol. 11, No. 3, 2018 Appendix Table A.1. Descriptive statistics of the examined variables Minimum Maximum Mean Std. Deviation Skewness Kurtosis Statistic Statistic Statistic Statistic Statistic Std. Error Statistic Std. Error Age 1 7 4.13 1.863 -.013 .003 -1.116 .006 Edu 1 6 3.35 1.667 .095 .003 -1.239 .007 DSex 1 2 1.50 .500 -.002 .003 -2.000 .006 DFborn 0 1 .99 .098 -9.988 .003 97.751 .006 DAfrica 0 1 .23 .423 1.266 .003 -.398 .006 DAsia 0 1 .23 .421 1.281 .003 -.359 .006 DEurope 0 1 .30 .458 .875 .003 -1.234 .006 DN_America 0 1 .02 .127 7.635 .003 56.298 .006 DOceania 0 1 .05 .220 4.084 .003 14.683 .006 DS&C_America 0 1 .16 .370 1.822 .003 1.318 .006 Source: own compilation based on (OECD, 2018) Table A.2. Pearson correlation matrix of the examined variables Edu Age DSex DFborn DAfrican DAsian DEuropean DN_A. DOceanian Edu 1.000 Age 0.072*** 1.000 DSex -0.01** 0.001 1.000 DFborn 0.000*** -0.009*** 0.000 1.000 DAfrica -0.009*** -0.022*** -0.01*** 0.054*** 1.000 DAsia 0.000 -0.01*** 0.000 0.05*** -0.3*** 1.000 DEurope -0.004*** 0.034*** 0.004*** -0.114*** -0.362*** -0.361*** 1.000 DN_America 0.004*** 0.009*** 0.000 -0.016*** -0.072*** -0.071*** -0.086*** 1.000 DOceania 0.01*** -0.003** 0.001 -0.006*** -0.126*** -0.126*** -0.152*** -0.03*** 1.000 Source: own compilation based on (OECD, 2018) Notes: Letters in the upper index refer to significance: ***: significance at 1 per cent, **: 5 per cent, *: 10 per cent. P-values without an index mean that the coefficient is not significant even at the 10 per cent level.