1 Introduction
To answer this research question, a Finite Mixture of Matrix-Normals model has been applied to cluster the units, taking into account the longitudinal dimension along 6 years, on the 52 available countries for 7 of the 8 dimensional indicators of the MIPEX. We relied on an unsupervised parametric clustering approach to minimize the risk of arbitrariness1 in the choices made and to be able to better evaluate the results.In order to improve the comparison between the countries regarding their migrant integration policies, is it possible to identify homogeneous groups over time among them, i.e. groups of countries which behave similarly across and within time?
2 Theoretical Framework and Related works
2.1 Immigrants Integration Framework
2.2 Immigration Policies Indexes: A Literature Review
-
First released by Banting et al. (2006), the Multiculturalism Policy Index (MCP) is a scholarly research project that monitors the evolution of multiculturalism policies in 21 Western democracies. The MCP is designed to provide information about multiculturalism policies in a standardized format that aids comparative research and contributes to the understanding of State-minorities relations. The project provides an index at 3 points in time: 1980, 2000, 2010, and for 3 types of minorities: one index relating to immigrant groups; one relating to historic national minorities; one index relating to indigenous peoples.
-
The Migrant Integration Policy Index (MIPEX) (Niessen et al., 2007; Solano & Huddleston, 2020) is a complex system of 167 policy indicators across 8 domains of citizenship and integration combined into a single composite indicator, in order to evaluate the migrant integration policies of each considered country (for details, see Sect. 3).
-
Based on the selection of data for 9 countries, between 1999 and 2008, and with the aim of measuring and comparing immigration, asylum, and naturalization policies across countries, the International Migration Policy and Law Analysis (IMPALA) database collects comparable data on immigration law and policy across 6 major areas of migration legislation: economic migration, family reunification, humanitarian migration, irregular migration, student migration, and the acquisition and loss of citizenship for migrants resident (Gest et al., 2014; Beine et al., 2016).
-
Helbling et al. (2017) presented the Immigration Policies in Comparison (IMPIC) project, which proposes a data set that allows to measure immigration regulations.
-
The Canadian Index for Measuring Integration (CIMI), is an interactive tool that allows for measuring the outcomes of immigrants in Canadian regions. It is a data-driven index that examines 4 dimensions of immigrants’ integration in Canada to assess the gaps between immigrants and the Canadian-born population. The CIMI identifies factors that underline successful immigrants’ integration, assesses changes and trends over time (currently from 1991 to 2020), enables detailed examination of 4 dimensions of integration and provides rankings based on empirical evidence for Canadian geographies.
-
The Immigration Policy Lab (IPL) (Harder et al., 2018) is a survey-based measure of immigrant integration, to provide scholars with a short instrument that can be implemented across survey modes, with the aim to strike a pragmatic compromise to help generate cumulative knowledge on immigrant integration. The IPL captures 6 dimensions of integration: psychological, economical, political, social, linguistical, and navigational.
3 Data
-
Labour Market Mobility (X1)
-
Family Reunion (X2)
-
Education (X3)
-
Political Participation (X4)
-
Long-term Residence (X5)
-
Access to Nationality (X6)
-
Anti-discrimination (X7)
-
Health4
3.1 Labour Market Mobility
3.2 Family Reunion
3.3 Education
3.4 Political Participation
3.5 Long-term Residence
3.6 Access to Nationality
3.7 Anti-discrimination
4 Methodology
4.1 Mixture of Matrix-Normals
5 Analysis and Results
-
Cluster 1 Estonia and Slovenia.In Cluster 1, we observe relatively low levels of temporal correlation, and this is due to the fact that Estonia has important changes in Family indicator in 2016 and 2017 and Residence in 2017, while Slovenia has important changes in the Anti-discrimination in 2016, in Education in 2018 and Politics in 2019. Cluster 1 is characterized by lower correlations in time between the first 3 years (2014–2016) and the second ones (2017–2019). Moreover, it has negative correlation between Labour Market Mobility and the other dimensions, with the exception of Family Reunion. Countries in this cluster have the lowest score for the Access To Nationality and rank low for Political Participation as well, while ranking high for Family Reunion, Long-term Residence and Anti-discrimination legislation.
-
Correlation in time with respect to the other clusters, Cluster 1 is the one with the lowest correlations within time.
-
Means this is the cluster with the lowest mean values in the Citizenship strand. With respect to the other clusters, it has low values in the Politics indicator but high values for Family, Residence and Anti-discrimination.
-
Correlation among indicators the Labour indicator presents negative correlations with almost all the other indicators except for Family. The correlation is particularly high between the indicators Labour and Anti-discrimination.
-
-
Cluster 2 Belgium, Canada, Chile, Hungary, India, Indonesia, Israel, Japan, Mexico, New Zealand, North Macedonia, Poland, Portugal, Romania, Slovakia, Sweden, Switzerland.During the analysed period, countries belonging to this cluster did not change much their policies, and they usually rank high in all the areas. The countries of this group tend to have good policies for Residence, Family and Anti-discrimination, but rank low for Education and Politics.
-
Correlation in time Cluster 2 presents high correlation values in time.
-
Means with respect to the other clusters, the values of the means of this group are quite low in Politics and Education and high in Family, Residence and Anti-discrimination.
-
Correlation among indicators almost all the indicators of this cluster are positively correlated, with particularly high values between Education and Labour, Politics and Labour, Politics and Education, Citizenship and Education and Citizenship and Politics.
-
-
Cluster 3 Albania, Austria, China, Croatia, Cyprus, Finland, Germany, Greece, Iceland, Ireland, Italy, Korea, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Russia, Serbia, Spain, Ukraine, UK, USA.The characteristic of Cluster 3 is its high stability in time, that is the tendency to not make huge changes in the legislation, with some remarkable exceptions such as Iceland in Anti-discrimination in 2018 and Citizenship and Anti-discrimination in Luxembourg in 2017. To this cluster, belongs the countries that reformed less their immigration legislation during the study period. They tend to rank average in most of the policies areas, with the exception of Residence and Anti-discrimination laws, where they tend to rank higher. This group could be seen as the “average” cluster, grouping countries which could be located at the middle of the MIPEX overall rank. This does not mean that any country of this cluster do not present high or low values in any indicator, but that overall, among the indicators the tendency is towards the center. However, low correlation among variables signals that countries do not move homogeneously among the policies areas.
-
Correlation in time Cluster 3 presents the highest correlations in time with respect to the other clusters.
-
Means with respect to the other clusters, this group does not present low mean values for any indicator. It presents medium values in Politics, Labour, Family, Education and Citizenship indicators and quite high values in Residence and Anti-discrimination.
-
Correlation among indicators almost all the correlations values among indicators are low, with exception for Residence and Family.
-
-
Cluster 4 Bulgaria, Czech Republic, France, Turkey.Cluster 4 is mainly characterised by its relatively low values of Politics in every country, including France. Important positive improvements in Education across time for all the countries mostly explaining the time-correlation behaviour. Despite ranking generally high for Anti-discrimination policies, countries within this cluster tend to rank low for policies in Education, Citizenship and Labour, while scoring average for Residence legislation. Yet, low correlation among variables indicates that the countries do not move homogeneously among the dimensions, with the exception of policies regarding Residence and Anti-discrimination, that have high positive correlation. Countries belonging to this cluster have seen their score moderately changing in time, indicating that some changes in the legislation have happened.
-
Correlation in time it presents high values but they shades with time.
-
Means with respect to the other clusters, Cluster 4 have the lowest mean values for Politics and quite low values in Education, Citizenship and Labour. It has high mean values in Anti-discrimination.
-
Correlation among indicators it generally presents low correlations with the exception for an high positive value between Anti-discrimination and Residence.
-
-
Cluster 5 Argentina, Australia, Brazil, Denmark, Moldova.Cluster 5 collects countries with smooth evolution, in both positive and negative directions and it generally presents low values in Education (with the exception of Australia). Changes are to be noted in Residence, where all the countries (with the exception of Argentina) see their values change in time (in both directions). Countries belonging to this cluster have high correlation values in time, but they tend to decrease faster with time, meaning that some changes in the policies have been made especially in the last years. Countries of this cluster, are characterized for generally ranking low in policies related to Educational support for foreign pupils and Politics, but high in Family, Residence, Citizenship and Anti-discrimination. However, the low correlation among the dimensions, means that the countries tend not to move homogeneously among them.
-
Correlation in time it presents high values but they shade faster.
-
Means with respect to the other clusters, the values of the means of Cluster 5 are quite low in Education and Politics, medium in Labour and high for the other indicators.
-
Correlation among indicators the values of the correlations are generally low.
-