Economic and Social Reports
Retention of government-assisted refugees in designated destinations: Recent trends and the role of destination characteristics

Abstract

Refugees resettled in Canada as part of the Government-Assisted Refugees (GARs) Program are assigned to designated communities across the country. As the dispersion of refugees across Canada continues to be a key objective of the government’s resettlement strategy, it is imperative to understand and identify the factors that may facilitate refugees’ retention in their designated community. This study asks if the retention of GARs in their designated destinations has increased since the early 2000s. And if so, to what extent can the change in the retention rate be accounted for by changes in refugees’ sociodemographic characteristics and the context of the designated destination, including factors such as the existence of resettlement service agencies, economic conditions, ethnic enclaves and cluster resettlement? The study found that the overall rate of retention among GARs in designated destinations by the end of the first full year after landing has increased considerably from 71% for the 2000 to 2004 cohort to 85% for the 2010 to 2014 and 2015 to 2018 cohorts while it has declined among economic and family class immigrants. The increase in retention rates across GAR cohorts was not a result of changes in their sociodemographic characteristics or the context of their designated destination; rather, the increase in retention rates occurred despite changes in those factors that, on their own, may have contributed to a decline in retention rates. Cross-sectionally, the retention of GARs was strongly associated with the number of GARs resettled in the same community in the same year (cluster resettlement) and the presence of co-ethnic communities. Cities with Resettlement Assistance Program Service Provider Organizations (RAP SPOs) that have been in operation for more than 10 years were associated with high retention rates compared with cities that did not have RAP SPOs.

Keywords: government-assisted refugees, geographic distribution, mobility

Authors

Yasmin Gure is with the Research and Evaluation Branch at Immigration, Refugees and Citizenship Canada. Feng Hou is with the Social Analysis and Modelling Division, Analytical Studies and Modelling Branch, at Statistics Canada.

Acknowledgements

This study was conducted in collaboration with Immigration, Refugees and Citizenship Canada. The authors would like to thank Cédric de Chardon, Marc Frenette, Lisa Kaida, Rebeka Lee and Li Xue for their advice and comments on an earlier version of this paper.

Introduction

A key policy objective of Canada’s immigration program is to promote a balanced geographic distribution of immigrants and refugees across Canada (IRCC, 2020). This concept of regionalization aims to ensure that the benefits of immigration, including economic growth and ethnocultural diversity, can be shared throughout the country. While Immigration, Refugees and Citizenship Canada (IRCC) continues to invest extensively in various policy tools to help steer the initial settlement of newcomers to medium and small metropolitan areas,^Note the department has a direct role in determining the resettlement destination of refugees. In particular, refugees resettled as part of the Government-Assisted Refugees (GARs) Program are assigned to designated communities across Canada prior to their arrival. Through this program, the specific resettlement needs of GARs are balanced with IRCC’s commitment to disperse refugees to cities throughout the country (IRCC, 2011).

The existing literature on the retention and secondary migration of newcomers has focused on immigrants more broadly, with analfysis often limited to data at the provincial level. Empirical research that specifically analyzes the retention of refugees and their mobility at the city level is limited (Kaida et al., 2020). This is an important knowledge gap, as analysis on interprovincial mobility cannot capture the migration trends of refugees that may be occurring at a lower geographic level. A more nuanced understanding of secondary migration among refugees is particularly relevant from a policy perspective, as the dispersion of refugees across Canada is a key component of IRCC’s overall objective of encouraging a more balanced geographic distribution of newcomers.

This study asks if the retention of GARs in designated destinations^Note has increased since the early 2000s. And if so, to what extent can the change in the retention rate be accounted for by changes in refugees’ sociodemographic characteristics and the context of the designated destination, including factors such as the existence of resettlement service agencies, economic conditions, ethnic enclaves and cluster resettlement? Answers to these questions would provide relevant information for the understanding of the factors that may facilitate refugees’ retention in their designated community.

Background

The Resettlement Assistance Program

Refugees who are selected abroad for resettlement in Canada are admitted through one of three resettlement streams: GARs, privately sponsored refugees (PSRs) and blended visa office-referred refugees (BVORs). Under the GAR Program, IRCC provides refugees with financial support and immediate essential services^Note under the Resettlement Assistance Program (RAP). GARs receive resettlement assistance services within the first four to six weeks of arrival in Canada, with income support being provided for up to one year^Note (IRCC, 2011).

Because of the highly specific needs of GARs, RAP services differ from the federally funded settlement services offered to all newcomers. As such, RAP services are exclusively offered through Resettlement Assistance Program Service Provider Organizations (RAP SPOs), which are located in 23 designated communities across Canada, outside of Quebec^Note (IRCC, 2016a). Prior to their arrival, GARs are “matched” or assigned to resettle in one of the 23 designated communities. The decision about where a GAR may be resettled in Canada is often determined by factors such as whether they have family or friends living in the area; the presence of ethnic, cultural and religious communities; and the availability of specific settlement services. In addition to these considerations, IRCC also determines where to resettle GARs based on discussions with provinces that help inform the department’s pre-established annual plans and quotas to resettle refugees across Canada (IRCC, 2016b).

The effectiveness of the destination assignment among GARs depends to a large extent on the match between the preferences articulated by GARs prior to their arrival and where they are assigned to resettle upon landing. In a study of refugees who arrived in 2000 and 2001, Simich et al. (2002) found that over 90% of GARs who changed their destination upon landing at the airport believed that they were not accommodated according to their expressed preferences. Among GARs who moved within the first year of arrival, almost 60% reported that they had stated a preferred destination prior to arrival but were not assigned accordingly. In more recent years, IRCC’s evaluation of the GAR Program (IRCC, 2011) was more positive, stating that among GARs who participated in the survey, an estimated 75% were content with their assigned community. This likely indicates that the match between refugees’ preferences and their assigned destination may have improved since the early 2000s, and this improvement may increase the retention of refugees in their designated destination.

The effectiveness of the destination assignment of GARs may also be affected by the capacity of RAP SPOs in the designated community. To bolster the efficiency of RAP SPOs, IRCC conducts an annual capacity survey to gather information on the ability of these organizations to provide services to refugees. The results of this annual survey would help to continuously improve matching GARs’ preferences with destination communities that can accommodate them.

Factors affecting retention and secondary migration

The ability of a destination city to retain immigrants and refugees can be attributed to a number of factors. One important factor is the size of the destination city. Krahn et al. (2005) found that larger cities have a higher retention rate than small cities. IRCC’s evaluation of the GAR Program (IRCC, 2011) also came to a similar conclusion. Specifically, the evaluation found that 30% of GARs who resettled in small communities chose to leave, compared with 14% of those who resettled in large cities. More recently, Kaida et al. (2020) found that the likelihood of GARs undertaking a secondary migration increased when the size of their initial city of settlement decreased.

The general tendency for immigrants and refugees to prefer larger cities is related to real or perceived economic opportunities (Derwing & Krahn, 2008; Hou, 2007; Hyndman et al., 2006; Newbold, 1996). Sherrell et al. (2005) emphasized that while immigrants and refugees may be encouraged to settle in smaller communities, it is difficult to retain them without suitable employment opportunities. Krahn et al. (2005) observed that among those who chose to leave their initial destination, the most frequently cited reasons for leaving were employment-related. Moreover, these individuals often chose larger cities as their preferred destination to resettle.

Large cities are also where refugees may find existing ethnic communities, a factor frequently cited in the literature as a strong motivation for secondary migration among immigrants (e.g., Hyndman et al., 2006). The role of large ethnic communities can be explained through the group affinity hypothesis that suggests that new immigrants prefer to move to and remain in areas where co-ethnic communities are concentrated (Derwing & Krahn, 2008; Hou, 2007; Nogle, 1994). Simich (2003) argued that refugees have a strong need for community support in their initial period of resettlement. While government support is an integral part of the assistance refugees receive, it can rarely meet all of their support needs. Thus, refugees may be particularly drawn to cities with large co-ethnic communities where they may receive familiar social and emotional support during their resettlement.

To some extent, research suggests social connectedness may play more of a role than economic conditions in the retention of refugees. In the case of PSRs, Kaida et al. (2020) found that unemployment was not significantly associated with leaving the initial destination. PSRs are resettled in communities with their sponsor where they may benefit from increased support in their initial year of arrival, as well as have access to their sponsor’s extended social network. Thus, this finding seems to indicate that for PSRs, social supports may supersede poor economic conditions.

Specific to GARs, there are considerable drawbacks if they choose to leave their destination community. For instance, because of limited funds, GARs may be forced to use their RAP income to finance their secondary migration. This leads to subsequent challenges in being able to afford accommodations and other necessities in their new city. Additionally, if GARs were receiving RAP services in their assigned community, they are no longer eligible to receive them when they move to a new city. IRCC’s evaluation (IRCC, 2016a) of the resettlement program noted that this rule meant that many GARs who left their destination community were unable to navigate resettlement in their new city.

This study seeks to build on the existing literature by further analyzing the retention and secondary migration patterns of GARs. Specifically, this study focuses on the recent trend and the effect of contextual factors in the retention of GARs in their designated destinations. This study makes several contributions to the literature.

First, this study is among the first to analyze the retention and secondary migration of refugees using census metropolitan areas (CMAs) and census agglomerations (CAs) as the geographic unit of analysis. This study follows the methodology employed by Kaida et al. (2020), who constructed consistent CMA and CA boundaries using data from the Longitudinal Immigration Database (IMDB).

Second, through the use of the IMDB, this study is able to identify refugees at the admission category level. This study compares the retention and secondary migration trends among GARs, PSRs, and economic and family class immigrants.

Lastly, this study examines how the individual sociodemographic characteristics of refugees and the contextual factors of the designated destination, such as economic conditions, presence of ethnic communities, cluster resettlement and resettlement services, are associated with the trend in refugee retention and secondary migration. In particular, this is the first Canadian study that uses the measures of cluster resettlement and resettlement services in analyzing refugee retention.

Data, measures and methods

Data

This study used the IMDB, which combines the landing records and annual tax information of immigrants (Statistics Canada, 2020). Immigrants who have filed at least one tax return since 1982 are included in the database. The landing records contain immigrant characteristics at the time of landing, such as intended destination,^Note admission category (e.g., economic class, family class or refugee), age, education, marital status, source country and official language capabilities. Tax records provide information on annual income, current marital status and place of residence. The IMDB data in this study cover landing information up to 2020 and tax information up to 2019.

The analysis focused on 76,500 GARs who were aged 20 to 54 at the time of immigration and arrived in Canada between 2000 and 2018. The choice of the age range is partly based on the pattern of filing income tax returns. Refugees (and immigrants in general) who arrived as children or seniors tend to have a low rate of filing income tax. Since retention is measured by comparing the intended destination of an individual with the place of residence reported on their income tax return form, the retention status cannot be determined for those who did not file income tax. Furthermore, in the analysis of factors affecting retention, some of those factors do not apply to children (e.g., level of education and employment status). GARs who were aged 20 to 54 at the time of immigration accounted for about 47% of all GARs who arrived during the study period. Descriptive analysis showed that the distribution of intended destinations among GARs who were aged 20 to 54 closely resembled that of all GARs (Appendix Table 1). Since children generally live with their parents, and many senior immigrants and refugees reside with their adult children, it is reasonable to expect the retention pattern observed among GARs who were aged 20 to 54 at the time of immigration to generally apply to all GARs.

Measures

CMAs and CAs were used as the geographic units to measure refugees’ intended destinations and subsequent mobility. A CMA or CA consists of one or more adjacent municipalities that are situated around an urban core and have a strong economic connection with the urban core as measured by commuting flows. Some CMA or CA boundaries may change from one census to another, and this change may pose a challenge when using CMAs and CAs to study mobility. In the Immigrant Landing File, the intended destinations are coded following the most recent census boundaries, but the tax file uses concurrent census boundaries for place of residence at the end of a tax year. For instance, for refugees who landed in 2010, their intended destinations are coded with the 2016 Census boundaries in the 2020 version of the IMDB, but their places of residence in the tax file are coded with the 2006 Census boundaries in 2010 and with the 2011 Census boundaries in 2011 to 2016. Such inconsistencies in CMA or CA boundaries between the Immigrant Landing File and the tax file may result in the misclassification of refugees’ retention status. To overcome this problem, this study created consistent CMA and CA boundaries in the tax file using 2016 Census boundaries as the base.^Note

The main outcome variable in this study was the retention in the intended destination.^Note For GARs, intended destinations are generally the communities to which they were assigned when they first arrived in Canada. Such an assignment is often based on the match between GARs’ characteristics and available resources in a community (settlement supports, family and ethnic community) that would best facilitate GARs’ socioeconomic integration. As will be shown, retention in the intended destination among refugees and immigrants decreased the most in the first year after their arrival. After the first year, the decline in the retention rate tended to be small and gradual. Therefore, this study focused on the retention status by the end of the first full year after landing. The retention status was determined by comparing a refugee’s intended destination with the CMA or CA where the refugee resided at the end of the calendar year of their arrival (hereafter the “landing year”), if they filed an income tax return in that year. For those who did not file income tax in the year of arrival but filed in the following year, the retention status is measured based on the CMA or CA where refugees resided in the year after the landing year. In the study sample, about 97% of refugees who filed an income tax return for the year after the landing year also did so in the landing year. Therefore, for the majority of the sample, the place of residence used to compare with the intended destination was that recorded in the landing year.

In the multivariate analysis, the focal independent variable was the refugee arrival cohorts, coded in four periods: 2000 to 2004, 2005 to 2009, 2010 to 2014 and 2015 to 2018. There were three sets of explanatory variables that could possibly affect the likelihood of a refugee staying in the intended destination and the change in the retention pattern across arrival cohorts.

The first set was the size of the intended destinations, coded into six groups: the three gateway cities of Toronto, Montréal and Vancouver; medium-sized CMAs with populations over 500,000 (Ottawa, Calgary, Edmonton, Hamilton, Winnipeg and Québec); small CMAs with populations between 100,000 and 500,000 (e.g., Victoria, Saskatoon and Halifax); and small urban areas (CAs). Refugees who initially settled in rural areas were excluded because it is difficult to create consistent boundaries for scattered rural areas. As evidenced in Table 1, very few refugees initially settled in rural areas.

The second set of explanatory variables included individual-level sociodemographic characteristics: age at the time of landing, gender, marital status, number of children, source region, language at the time of landing, level of education at the time of landing, employment status after landing and school attendance after landing. The source region was coded into nine categories: South America, Southern Europe, Eastern Europe, Africa, Southern Asia, Southeast Asia, East Asia, West Asia or others. Language skills were based on the combination of mother tongue and self-reported knowledge of official languages at the time of landing: English mother tongue, French mother tongue, other mother tongue but spoke English, other mother tongue but spoke French, other mother tongue but spoke English and French, or did not speak English or French. The education variable was coded in four categories: less than high school, high school graduation, some postsecondary education, or bachelor’s degree or higher. Employment status was coded as employed (had at least $500 in employment income in the landing year or the year after) versus others. The school attendance variable was coded as 1 if an immigrant attended full-time postsecondary education in the last tax year or as 0 if otherwise.^Note

The third set of explanatory variables was used to measure the context of the intended destinations: the presence of RAP SPOs,^Note the economic conditions, ethnic enclave and cluster resettlement. The RAP SPOs variable was coded in four categories: none, 1 to 5 years in operation, 6 to 10 years in operation or over 10 years in operation. The years in operation were based on the difference between the year a RAP SPO centre was first established and the year when a refugee first arrived in Canada.^Note This variable was used to capture the effects of not just the existence of RAP SPOs but also their cumulated experience, established network and proven capacity as proxied by years in operation.^Note Local economic conditions were proxied by the percentage of adults (aged 18 to 64) with annual earnings over $500 (i.e., employment incidence in a year) among the Canadian-born population and long-term immigrants (immigrants who arrived in Canada more than 20 years earlier).^Note This variable was included to examine whether the retention rate is higher in local labour markets with more job opportunities.^Note

The ethnic enclave was measured by the percentage of immigrants from the same origin region in the CMAs or CAs where a refugee was assigned to settle.^Note This variable was used to test whether the presence of existing own-ethnic communities is associated with a higher retention of refugees in the intended destination. The cluster resettlement was measured by the number of GARs who arrived at the same intended destination in the same year. Different from ethnic enclave, which is about pre-existing own-ethnic communities, cluster resettlement was used to capture social connections originating from close ties to family and relatives, a shared refugee experience, and common resettlement challenges and opportunities among refugees who arrived in the same community in the same year. In the multivariate analysis, the logarithm transformation of this variable was strongly associated with the retention of refugees in the intended destination, while its original form was not statistically significant.

Methods

The analysis started with descriptive tables showing the distribution of intended destinations and how the distribution changed across arrival cohorts for GARs aged 20 to 54 at the time of landing and, as a comparison, for PSRs and immigrants in the economic and family classes. Descriptive statistics were further produced to show the mobility status of GARs by the end of the first full year after landing and the rate of retention in the intended destination for up to 10 years after arrival.

In multivariate analyses, linear probability models^Note were used to examine changes in retention by cohort in the intended destination and the effects of the selected explanatory variables, particularly contextual variables. Three sequential models were estimated. Model 1 included only arrival cohorts; the coefficients of the cohort variable reflect the observed change in retention in intended destinations. Model 2 added the size categories of the intended destinations and individual-level covariates. The changes in the coefficients of the cohort variable from Model 1 to Model 2 reflect the extent to which the observed change in retention can be accounted for by the differences across arrival cohorts in the distribution of initial destinations and individual-level covariates. Model 3 added community-level contextual variables. The changes in the coefficients of the cohort variable from Model 2 to Model 3 would indicate the possible effects of changes in the context of initial destinations on the observed cohort differences. Model 3 is multi-level, with individual- and community-level variables. Because dependency among observations within a community (CMA or CA) can underestimate standard errors of regression coefficients, robust variance estimation was used to mitigate possible bias (Steenbergen & Jones, 2002).

Recent changes in retention in intended destinations among government-assisted refugees

Table 1 presents the distribution of GARs, PSRs, and economic and family immigrants aged 20 to 54 at the time of immigration across the size categories of intended destinations.^Note

Unlike economic and family class immigrants, the majority of GARs were intended to settle outside the three gateway centres. For instance, in the 2000 to 2004 landing cohort, 66% of GARs were intended to settle outside the three gateway centres, compared with 57% of PSRs, 20% of economic immigrants and 30% of family immigrants. The share of GARs intended to settle outside the gateway centres remained stable among the 2005 to 2009 and the 2010 to 2014 cohorts but increased to 73% among the 2015 to 2018 arrivals. The share of economic and family class immigrants who intended to settle outside gateway centres also increased across successive cohorts but remained much lower than the corresponding share of GARs in the 2015 to 2018 cohort.

Among GARs intended to settle in non-gateway centres, the majority (85%) went to medium-sized and small CMAs. The share of those intended for small CMAs was slightly smaller than the share of those intended for medium-sized CMAs, particularly among the 2015 to 2018 arrivals. This pattern was different from those for PSRs and economic and family class immigrants, who were primarily attracted to medium-sized CMAs. For each of the four arrival cohorts, less than 0.6% of GARs were intended to settle in rural areas, compared with 2% to 4% of economic and family class immigrants.

The above patterns of distribution among prime-aged refugees and immigrants were generally the same for the population as a whole in each group (Appendix Table 1).



Table 2 presents the tax filing status and the rate of retention in the intended destination among GARs, PSRs, and economic and family class immigrants who were aged 20 to 54 at the time of landing and filed an income tax return by the end of their first full year after landing (as discussed, mobility status can be determined only among tax filers).

About 97% of prime-aged GARs filed an income tax return for the landing year or the first full year after the landing year for all the selected cohorts (Table 2). The tax-filing rates of GARs were similar to the rates of PSRs but much higher than those of economic and family class immigrants. In particular, the overall tax-filing rates among economic immigrants ranged from 83% to 90%, depending on arrival cohorts. Previous studies show that among prime-aged individuals who did not file income tax in the initial years after arrival, some remained in Canada and would file income tax in later years, but most might have left Canada. Economic immigrants tended to have a higher emigration rate than refugees (Aydemir & Robinson, 2008). Note that when the population of all ages at the time of landing was considered, tax-filing rates were much lower than those of prime-aged arrivals, and GARs had lower rates than the other three groups of refugees and immigrants (Appendix Table 2). This highlights the importance of restricting the examination of retention and mobility to prime-aged individuals.

Among prime-aged individuals who filed income tax for the landing year or the first full year after landing, there were several salient patterns in the rate of retention in intended destinations among GARs relative to PSRs and immigrants.

First, the overall retention rate among GARs increased considerably from 71% for the 2000 to 2004 cohort to 79% for the 2005 to 2009 cohort and further to about 85% for the 2010 to 2014 and 2015 to 2018 cohorts. In comparison, the overall retention rate among PSRs rose only marginally from 74% for the 2000 to 2004 cohort to 76% for the 2015 to 2018 cohort and decreased among economic and family class immigrants.

Second, the retention rate among GARs was not related to the size of the intended destination but was generally lower in smaller destinations among PSRs and economic and family class immigrants. The retention rate of GARs tended to be highest in Vancouver and medium-sized CMAs, followed by small CMAs, with Montréal having the lowest rates for most arrival cohorts. In comparison, economic and family immigrants had the lowest retention rates in small urban centres and the highest retention rates in the three gateway centres (except for Montréal, which had a lower retention rate than medium-sized CMAs for economic immigrants who arrived in the 2010s).

Third, as their retention rates increased across successive cohorts, GARs had higher retention rates than PSRs and the two immigrant groups in small CMAs and small urban areas among those who arrived in the 2010s. GARs who arrived in the 2010s also had a higher retention rate in medium-sized CMAs than PSRs and economic immigrants. High retention rates in non-gateway areas are pertinent to the geographic balance of immigrant distribution.



An important question related to the retention of refugees or immigrants in their intended destinations is, where did they move to when they left their intended destinations? Did the majority of them move to gateway centres or did the majority of them move to other areas outside gateway centres? To answer this question, Table 3 presents the cross-tabulation of intended destinations and new destinations for refugees and immigrants who left their intended destinations by the end of the first year after their landing year.
The results show that when GARs left their intended destination, the majority moved to a location outside the three gateway centres. Specifically, 93% of GARs who moved from Toronto moved to a non-gateway location. The corresponding rates for GARs who moved from Montréal and Vancouver were 84% and 77%, respectively. The rates of staying out of gateway centres were lower for GARs who moved from non-gateway locations, ranging from 65% to 72%, but these rates were much higher than the corresponding rates for economic and family immigrants. For instance, among economic immigrants who moved from medium-sized and small CMAs, 40% to 47% stayed outside gateway centres.
One possible reason why GARs who moved were more likely to stay outside gateway centres than economic and family immigrants could be that GARs might tend to move to a city with RAP SPOs, and these are mostly located in medium-sized or small CMAs (although Toronto and Vancouver have had RAP SPOs since 1999).^Note For instance, about 93% of GARs who moved from Toronto moved to a location with RAP SPOs. The corresponding rates for those who moved from Vancouver, medium-sized CMAs and small CMAs were 78%, 83% and 69%, respectively. The number of medium-sized or small CMAs with RAP SPOs in a province could also be a reason why a much higher share of GARs who moved from Toronto stayed away from gateway centres than their counterparts from Vancouver. In Ontario, there were 10 medium-sized or small CMAs with RAP SPOs, many of which have been in operation for several years. In comparison, there have been only two small CMAs with RAP SPOs in British Columbia since 2016.



The results so far show that 15% to 30% of GARs left their intended destinations by the end of the first full year after the landing year. This observation raises an important question about whether GARs would continue to move away from intended destinations at a similar pace in subsequent years. If this is the case, efforts that are aimed at retaining GARs in their intended destination would need to be sustained for several years. Conversely, if the high mobility was limited to the first year, efforts of retention could be confined to improving the initial settlement experiences.

To answer the above question, Chart 1 plots the proportion of GARs who were aged 20 to 54 at the time of admission, who arrived between 2000 and 2009, and for whom 10 years of mobility history was observed in their intended destinations by size of the intended destination and the number of years since landing. The chart shows that regardless of the size of the intended destination, the majority of out-migration occurred in the landing year (zero years since immigration) and the year after the landing year (one year since immigration). For instance, among GARs intended for medium-sized CMAs, 24% left by the end of the year after the landing year, but it took the subsequent nine years for another 11% to leave. For those who were assigned to small urban areas, where the lowest long-term retention rate was observed, 37% of GARs left by the end of the year after the landing year, and it took the subsequent nine years for another 24% to leave.^Note

These results suggest that the period from landing to the end of the year after the landing year is the most critical in terms of understanding the retention of GARs in their intended destinations. Accordingly, the following multivariate analyses focus on retention in this initial period.

Data table for Chart 1



Data table for Chart 1
Table summary
This table displays the results of Data table for Chart 1. The information is grouped by Years since immigration (appearing as row headers), Montréal, Toronto, Vancouver, Medium-sized census metropolitan areas, Small census metropolitan areas and Small urban areas, calculated using percent units of measure (appearing as column headers).

Years since immigration	Montréal	Toronto	Vancouver	Medium-sized census metropolitan areas	Small census metropolitan areas	Small urban areas
	percent
0	66.3	67.8	88.3	82.7	71.4	72.4
1	63.6	66.2	84.0	77.5	62.8	62.6
2	62.6	65.9	82.0	74.6	57.8	56.2
3	61.9	65.1	79.7	72.3	54.2	51.6
4	61.6	64.2	78.4	70.7	51.4	48.3
5	61.2	63.7	77.7	69.2	49.1	45.6
6	60.3	63.3	77.2	68.3	47.3	43.5
7	59.4	62.8	76.7	67.4	45.8	42.3
8	59.4	62.6	75.8	66.4	45.0	40.5
9	59.1	62.7	75.6	65.5	44.3	39.9
10	58.5	61.8	75.3	65.0	43.8	39.2
Source: 2020 Longitudinal Immigration Database.

Results of multivariate analyses

This section first examines whether differences in sociodemographic characteristics and the context of intended destinations can account for some of the observed increase in retention rates across arrival cohorts of GARs. This is followed by analyses of the associations between the context of intended destinations and the retention of GARs.

Table 4 presents the sociodemographic characteristics and the context of the intended destinations for four arrival cohorts of GARs. One clear change across successive cohorts was the decrease in the level of education.^Note The share of those with less than a high school education increased from 66% among the 2000 to 2004 cohort to 83% among the 2015 to 2018 cohort, and the share of those with a university degree decreased from 13% to 6%. Another clear trend was the decrease in the share of GARs who were employed (with at least $500 earnings in a tax year) in the landing year or in the year after the landing year, from 17% for the 2000 to 2004 cohort to 10% for the 2015 to 2018 cohort. Both changes are consistent with the shift in refugee selection criteria from adaptability to vulnerability (Kaida et al., 2021). The source region composition of GARs varied considerably and became more concentrated across arrival cohorts. In the 2000 to 2004 cohort, West Asia, Africa, Eastern Europe and South America were the four major regions, accounting for 93% of the cohort population. In the 2005 to 2009 cohort, Southeast Asia replaced Eastern Europe as one of the four top source regions. In the 2010 to 2014 cohort, West Asia, Africa and South Asia accounted for 91% of GARs. Among the 2015 to 2018 arrivals, 67% came from West Asia (mostly Syria) and 27% came from Africa.



The changes in contextual variables show a gradual increase in the size of ethnic enclave (percentage of own-group immigrants in the region), reduced cluster resettlement (log of number of GARs who arrived in the same year) and more established RAP SPOs. The share of own-group immigrants in the CMA or CA’s population (as a measure of ethnic enclave) increased on average from 1.2% for the 2000 to 2004 cohort to 2.4% for the 2015 to 2018 cohort, reflecting the growing size of immigrant populations from regions where refugees originated. The log of the number of GARs who arrived in the same community in the same year decreased across arrival cohorts. This decrease was a result of the decline in the number of GARs admitted annually from 10,700 in 2000 to fewer than 6,000 in 2012 and 2013 and 7,600 in 2014. The number of GARs increased considerably from 2015 to 2018, particularly in 2016 (23,600), when a large wave of Syrian refugees arrived. However, GARs in this period were much more likely to be assigned to small and medium-sized CMAs, where, each year, the average number of GARs was much smaller than the number in each gateway centre.

The majority (over 76%) of GARs who arrived in the 2010s were assigned to cities with RAP SPOs that had been in operation for over 10 years, while 62% of the 2000 to 2004 arrivals were resettled in cities with RAP SPOs that had been in operation for 1 to 5 years.^Note Across successive cohorts, the share of GARs who were destined to CMAs or CAs without RAP SPOs decreased from 28% to 14% as more cities established RAP SPOs.

To what extent were the above changes in GARs’ sociodemographic characteristics and the context of the intended destinations associated with the observed trend in the rate of retention in intended destinations? This question is addressed by multivariate models in Table 5.

Model 1 in Table 5 replicated the observed cohort differences in the retention rate as in Table 2. The observed retention rate increased by 7.9 percentage points from the 2000 to 2004 cohort to the 2005 to 2009 cohort, and increased another 6.2 percentage points for the 2010 to 2014 cohort, then changed little for the 2015 to 2018 cohort. In Model 2, after changes in the distribution of initial destinations and sociodemographic characteristics were controlled for, the trend of rising retention rates across successive cohorts became stronger. In particular, the difference in retention rates between the 2015 to 2018 cohort and the 2000 to 2004 cohort increased to 15.0 percentage points, a 1.7 percentage point increase from the observed difference in Model 1. Further decomposition analysis^Note showed that the increase in the coefficient of the 2015 to 2018 cohort was related primarily to changes in source regions of GARs. Specifically, GARs who came from Eastern Europe and South America tended to have relatively high retention rates, but very few of them were in the 2015 to 2018 cohort. The results suggest that holding the source region constant would increase the retention rate more than the observed pattern for the 2015 to 2018 cohort.

The retention rate increased another 0.6 percentage points for the 2015 to 2018 cohort in Model 3. Further decomposition analysis showed that this increase related primarily to the decrease in cluster resettlement, i.e., the average number of GARs assigned to a city in a year. Put differently, since a higher level of cluster resettlement was significantly associated with a higher retention rate, a cohort with a lower level of cluster resettlement would be expected to have a lower retention rate.

Since the cluster resettlement variable was strongly correlated with the size of the intended destinations, the control of the cluster resettlement variable also changed the coefficients of the size categories of the intended destinations from Model 2 to Model 3. With similar levels of cluster resettlement, non-gateway centres would have much higher retention rates than Toronto and Montréal.

As mentioned earlier, the cluster resettlement variable was statistically significant in its logarithm form but not in its actual level. This indicates that the association between cluster resettlement and retention was not linear and became smaller as the number of GARs assigned to a city increased. Chart 2 illustrates the estimated relationship between retention rates and the number of GARs who arrived in a city in a given year, based on the results from Model 3. The chart shows that the retention rate increased quickly before the number of GARs assigned to a city reached 500 but grew more slowly as the number of GARs reached beyond 500. For instance, the retention rate would increase 14.9 percentage points from 59.2% to 74.1% when the number of GARs in a city increased from 20 to 500 but would only increase another 4.1 percentage points when the number of GARs in a city increased from 500 to 1,000.

In addition to the cluster resettlement variable, an ethnic enclave was also significantly associated with GARs’ retention rate in a city. The estimate in Model 3 shows that a 1 percentage point increase in the share of immigrants from the same source region in the city’s population was associated with a 1.9 percentage point increase in the retention rate. The regional economic conditions variable (employment incidence among Canadian-born citizens and long-term immigrants) was not statistically significant.

The results for the fourth contextual variable regarding RAP SPOs show that the mere presence of RAP SPOs was not necessarily associated with a higher retention rate. However, cities with RAP SPOs that had been in operation for over 10 years were associated with a 4.2 percentage point higher retention rate than cities without RAP SPOs.^Note



Data table for Chart 2



Data table for Chart 1
Table summary
This table displays the results of Data table for Chart 1. The information is grouped by Number of government-assisted refugees who arrived in a location in a year (appearing as row headers), Retention rate, calculated using percent units of measure (appearing as column headers).

Number of government-assisted refugees who arrived in a location in a year	Retention rate
	percent
20	59.2
33	61.5
55	63.8
90	66.1
148	68.4
245	70.8
403	73.1
665	75.4
1097	77.7
1808	80.1
2981	82.4
4915	84.7
8103	87.0
Source: 2020 Longitudinal Immigration Database

Discussion and conclusion

This study analyzed the retention and secondary migration patterns of GARs. In particular, the study examined whether retention rates had increased among successive landing cohorts since the early 2000s. Moreover, the study aimed to understand the extent to which the sociodemographic characteristics of GARs, as well as the context of their intended destination, influenced their decision to undertake a secondary migration.

The study observed that the overall retention rate among GARs increased considerably among successive cohorts, while it declined among economic and family class immigrants. The increase in retention rates occurred despite changes in sociodemographic factors and the context of the intended destinations that, on their own, may have contributed to a decline in retention rates. Thus, the increase in the retention rate was related to factors that could not be examined in this study. For instance, some unobserved characteristics (e.g., experiences of violence, physical disabilities, mental illness and extended time in refugee camps) of more recent cohorts may increase their tendency to stay in their designated destination. Rising housing costs in major metropolitan areas would make it less affordable for GARs to move away from smaller urban areas. In resettling GARs, the government might have taken better consideration of their preferences. It is also possible that resettlement services have been improved to meet refugees’ needs in intended destinations.

The study found that regardless of the size of the intended destination, the rate of out-migration was highest in the year of arrival and the following year but drastically tapered off following this period. For example, it was observed that among GARs destined for medium-sized CMAs, about one-quarter had left by the first year after landing, but it took a subsequent nine years for an additional 11% to leave.

The results of the multivariate analysis showed that the number of GARs who resettled in the same community in the same year was strongly correlated with an increase in retention. However, this association was not linear. Simply put, the positive association became smaller as the number of GARs assigned to a city in a year increased beyond a few hundred.

As supported by the existing literature, this study found that the presence of co-ethnic communities was significantly associated with GARs’ retention rate in their intended destination. Lastly, while the mere presence of RAP SPOs did not necessarily increase retention rates, cities with RAP SPOs that have been in operation for more than 10 years were associated with high retention rates compared with cities that did not have RAP SPOs.

The findings in this study inform policy discussions in several ways. This study showed that about two-thirds of refugees who were assigned to medium-sized CMAs ultimately stayed, and 39% to 44% of those who were assigned to small CMAs and small urban areas stayed 10 years after immigration. These rates may not seem high, but they were similar to those of economic immigrants whose initial destinations were not assigned by the government. Moreover, it is particularly salient that even when GARs leave their intended destination, the majority do not choose to resettle in Canada’s large gateway cities. This study’s observation that the majority of secondary migration occurs within the landing year or the year after is an important consideration for improving GARs’ initial settlement experiences. In terms of future resettlement, the analysis about cluster resettlement is specifically integral when considering retention rates. The findings indicate that resettling a sizable number of GARs in the same community in the same year is strongly associated with increased retention, but this association starts to diminish when the size of the cluster resettlement increases beyond a few hundred.

 

References

Aydemir, A., & Robinson, C. (2008). Global labour markets, return, and onward migration. Canadian Journal of Economics, 41(4), 1285‒1311.

Derwing, T. M., & Krahn, H. (2008). Attracting and retaining immigrants outside the metropolis: Is the pie too small for everyone to have a piece? The case of Edmonton, Alberta. Journal of International Migration and Integration, 9(2), 185–202.

Hou, F. (2007). Changes in the initial destinations and re-distribution of Canada’s major immigrant groups: Re-examining the role of group affinity. International Migration Review, 41(3), 680‒705.

Hou, F. (2014). A general approach to effect decomposition. Social Science Quarterly, 95(3), 894‒904.

Hou, F., Schellenberg, G., & Berry, J. (2018). Patterns and determinants of immigrants’ sense of belonging to Canada and their source country. Ethnic and Racial Studies, 41(9), 1612‒1631.

Hyndman, J., Schuurman, N., & Fiedler, R. (2006). Size matters: Attracting immigrants to Canadian cities. Journal of International Migration and Integration, 7(1), 1‒25.

Immigration, Refugees and Citizenship Canada. (2011). Evaluation of Government Assisted Refugees (GAR) and Resettlement Assistance Program (RAP). Government of Canada.

Immigration, Refugees and Citizenship Canada. (2016a). Evaluation of the Resettlement Programs (GAR, PSR, BVOR and RAP). Government of Canada.

Immigration, Refugees and Citizenship Canada. (2016b). The Matching Centre. Government of Canada.

Immigration Refugees and Citizenship Canada. (2020). Departmental Plan 2020/2021. Government of Canada.

Kaida, L., Hou, F., & Stick, M. (2020). Are refugees more likely to leave initial destinations than economic immigrants? Recent evidence from Canadian longitudinal administrative data. Population Space and Place, 26(5):1-14.

Kaida, L., Stick, M., & Hou, F. (2021). Changes in selection policy and refugee welfare use in Canada. International Migration.

Krahn, H., Derwing, T. M., & Abu-Laban, B. (2005). The retention of newcomers in second- and third-tier Canadian cities. The International Migration Review, 39(4), 872–894.

Mood, C. (2010). Logistic regression: Why we cannot do what we think we can do, and what we can do about it. European Sociological Review, 26(1), 67–82.

Newbold, K. (1996). Internal migration of the foreign-born in Canada. The International Migration Review, 30(3), 728–747.

Nogle, J. (1994). Internal migration for recent immigrants to Canada. The International Migration Review, 28(1), 31–48.

Sherrell, K., Hyndman, J., & Preniqi, F. (2005). Sharing the wealth, spreading the “burden”? The settlement of Kosovar refugees in smaller British Columbia cities. Canadian Ethnic Studies, 37(3), 76–96.

Simich, L. (2003). Negotiating boundaries of refugee resettlement: A study of settlement patterns and social support. The Canadian Review of Sociology and Anthropology, 40(5), 575–591.

Simich, L., Beiser, M., & Mawani, F. (2002). Paved with good intentions: Canada’s refugee destining policy and paths of secondary migration. Canadian Public Policy, 28(4), 597–607.

Statistics Canada. (2020). Longitudinal Immigration Database (IMDB) Technical Report, 2018 (Diversity and Sociocultural Statistics, No. 24).

Steenbergen, M. R., & Jones, B. S. (2002). Modeling multilevel data structures. American Journal of Political Science, 46(1), 218–237.