Chapter 4. Family ties: How family reunification can impact migrant integration

This chapter investigates if migrants’ long-term integration outcomes are affected by delays in family reunification. The integration outcomes of both principal migrants and the spouses who reunite with them are considered. The chapter provides new empirical evidence for a range of OECD countries and discusses potential reasons why delays in family reunification influence integration outcomes such as wages, employment, and language proficiency. It also explores the effect that age at arrival can have on the integration outcomes of migrant children as well as the role played by the presence of migrants’ parents. The chapter concludes by highlighting implications for policies regulating family reunification.

This work was supported by Germany’s Federal Ministry for Family Affairs, Senior Citizens, Women and Youth.

    

The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.

Introduction

This chapter provides a first empirical examination of the links between family presence and migrants’ long-term integration outcomes in OECD countries. It investigates, firstly, how the integration of the principal migrant is affected by the presence of the spouse or of a parent, and secondly, how the integration of spouses and children is affected by delays in their arrival. The measures of integration considered include employment status, hours worked, wage levels and proficiency in the host-country language. Apart from analyses on migrants who arrived as children, all analyses focus on persons in married migrant couples.

A number of recent policy debates have questioned the role of family reunification in integration. In particular, this question has arisen in developing policy for family reunification in the context of the refugee surge in 2015/2016 but also in the intensified competition between countries to attract and retain high-skilled labour migrants. More generally, the question is pertinent to management of family migration, as it is the largest channel of migration to OECD countries. However, quantitative evidence is rare, especially on the effects of delays in the arrival of a migrant’s spouse. These delays could affect the integration of both principal migrants and the spouses themselves.

Main findings

• The vast majority of married migrants live with their spouse in the host country. The share of migrants whose spouse is absent remains below 20% in almost all OECD countries, and it falls with duration of stay.

• By contrast, delays in family reunification are frequent in OECD countries: only 54% of married migrants arrived in the same year as their spouse. Empirical evidence on how these delays affect integration is virtually non-existent.

• The evidence obtained on principal migrants is not clear-cut. On the one hand, principal migrants whose spouse arrived with some delay earn significantly lower wages than otherwise comparable principal migrants, after ten years or more in the host country. In the United States, one year of additional delay is associated with wages being 3% lower. In European OECD countries, it is associated with a lower probability of earning a wage above the median. This might partly reflect lower initial wages among migrants who struggled to bring their family quickly. Yet further results suggest that delays also cause lower wages, e.g. by delaying investments in education.

• On the other hand, principal migrants whose spouse arrives with delay exhibit slightly higher employment probabilities after ten years or more. The host-country language proficiency of principal migrants after ten years or more appears unaffected by delays in the spouse’s arrival. Migrants who live with their spouse exhibit roughly the same subjective well-being as migrants who live without their spouse.

• Delays also seem to affect the integration outcomes of the spouse who reunites with the principal migrant, especially in the case of female spouses. In European OECD countries and the United States, spouses who arrive with delay exhibit lower language proficiency after five years or more in the host country. In European OECD countries and Canada, delay is also associated with a lower employment probability.

• Integration outcomes of migrant children can be strongly affected by long delays. Children who arrive at pre-school age can have substantially more favourable integration outcomes as adults than children who arrive at school age, in terms of educational attainment, employment, wages and especially host-country language proficiency.

• In both European OECD countries and the United States, migrants whose parents live in the same household exhibit higher employment probability and longer working hours, especially when they have young children.

• Evidence from a policy change in Germany shows that the imposition of certain conditions, e.g. requiring spouses to reach basic language proficiency before arrival, can significantly lengthen the delays in family reunification. If the objective of such conditions is to ensure that spouses integrate well, then the delays caused by the conditions can undermine this objective.

Motivation and context

Migration to OECD countries has an important family dimension: many principal migrants have family members who either accompany them to the host country or reunite with them after some time. While the principal migrant holds a residence permit for employment, study or on humanitarian grounds, family members’ residence permits are based on the kinship link to the principal migrant, who therefore acts as sponsor (see Table 4.1 for terminology). If the principal migrant marries a person abroad while residing in the host country, the new spouse may be eligible to join the principal migrant. Taken together, admissions on family grounds have represented the largest migration channel to OECD countries in recent years, and family reunification accounts for a considerable part of this flow (Chaloff and Poeschel, 2017[1]).

The large flow of family migrants, many of whom arrive with some delay, raises the question of how this process affects integration outcomes. This notably concerns integration outcomes of principal migrants: how is their integration in the host country affected by the presence of their family and by initial periods of separation? It further concerns integration outcomes of family migrants themselves: how does delayed reunification affect the integration of sponsored spouses and children? If such effects exist, do they persist in the long-term?

Table 4.1. Forms of family migration
Subcategory	Definition
Family formation	A resident national or foreigner marries a foreigner and sponsors that individual for admission or for status change.
Accompanying family	Family members are admitted together with the principal migrant.
Family reunification	Family members migrate after the arrival of a principal migrant who sponsors their admission. The family ties predate the arrival of the principal migrant.
International adoption	A resident national or foreigner adopts a child of foreign nationality resident abroad.

Answers to these questions would help address several policy issues. In European OECD countries, it is not clear how to deal with large numbers of requests for family reunification made by recently arrived refugees. These sponsors are often not yet in a position to support their families in the destination country but may be exempted from income requirements, depending on their specific status. Next, policy makers in many OECD countries wonder how they can attract and retain highly-skilled labour migrants. The possibility for migrants to bring their families without delay appears to play a central role – the programmes most strongly geared towards highly-skilled labour migrants typically offer the most generous rules for family members. The integration of family migrants themselves is often challenging, and it is not yet well understood whether the timing of their arrival or preconditions for their language proficiency can be used as policy levers to improve their long-term integration outcomes.

Despite the relevance for migration and integration policy, empirical evidence on integration effects from family migration is largely absent. The few studies that approach the question typically focus on a particular aspect rather than an overall effect and take a qualitative approach based on very small samples, so that their results cannot be generalised (see below for a literature review). Quantitative approaches that cover several countries appear to be missing altogether. The paucity of evidence therefore contrasts with the prominence of recent debates on family reunification.

This chapter seeks to deliver some first insights, primarily based on large data sets that together cover most OECD countries (Box 4.1). These data include permanent and temporary migrants from all migrant categories. While the insights obtained might apply with some generality across countries, they should be treated with caution: due to data limitations, it is often not possible to distinguish causal links from mere correlation and to sufficiently explore alternative explanations. The analyses first examine how integration outcomes of a principal migrant vary with delays in the arrival of the spouse. Further analyses examine the effect of delay on integration outcomes of spouses and children. In each case, the analysis is limited to migrants who are married to another migrant (i.e. excluding mixed couples of one migrant and one native-born person) and the effect of delays is isolated from other factors relevant for migrants’ integration. The next section begins with some descriptive evidence obtained in this context.

Box 4.1. Data sources and sample selection

The main analyses draw on three micro-level data sets, the European Labour Force Survey 2010-17 (and its ad-hoc modules 2008 and 2014), the American Community Survey 2013-16, and the 2016 Canadian Census. These data sets offer a number of important advantages: they consist of large numbers of observations for recent years; they are representative of the entire population; they include data on several household members as well as migration-related variables; and they together cover many OECD countries. These data can be expected to provide a sufficient empirical basis for estimating effects of family presence on migrants’ integration across OECD countries. The household dimension is critical for the estimation approach, as it allows matching the data on the two spouses in married couples.

The micro data of the European Labour Force Survey (EU LFS) is produced by Eurostat, merging household survey data from the 28 member countries of the European Union and from three EFTA countries (Norway, Switzerland and Iceland). Analyses in this chapter only exclude non-OECD member countries from these data. For individuals aged 15 and above, the data include variables such as age, sex, marital status, education, labour force status, hours worked, citizenship, country of birth and duration of stay in the host country. The data on households is (partly) missing for Finland, Iceland, Norway and Switzerland, so that several analyses in this chapter cannot include these countries. In the case of Poland, a substantial part of the data on migrants have missing values.

In addition, ad-hoc modules are linked with the core variables of the EU LFS in the same year. The ad-hoc modules 2008 and 2014 oversampled migrants and include migration-specific variables such as the self-declared main reason for migration (employment, family, study or humanitarian reasons) and proficiency in the host-country language. All EU and EFTA member countries are covered, except Croatia and Iceland in 2008 and Denmark, Germany, Iceland, Ireland and the Netherlands in 2014. Analyses in this chapter cannot use 2008 data on Norway and Finland due to missing variables.

The American Community Survey (ACS) is conducted by the United States Census Bureau. On an annual basis, it provides representative data for the United States, based on a sample of more than 3.5 million households. Its variables include those mentioned above for the EU LFS as well as a variable for detailed wages. Information on job tenure is not available from either data set. Migration-related variables in the ACS include proficiency in English and language spoken at home, while information on reason of migration or migration category is not available. Strong overlap between the variables covered by the ACS and the EU LFS allows for largely parallel analyses in this chapter.

The 2016 Canadian Census was conducted by Statistics Canada. The public-use file on individuals includes records for 2.7% of the Canadian population, drawn from a larger sample that covers one-quarter of the population. In addition to the sociodemographic variables mentioned above for the EU LFS, the data provide information on proficiency in Canada’s official language, the language spoken at home, ethnicity, age at immigration and parents’ place of birth. A variable on migration category is based on administrative records from Immigration, Refugees and Citizenship Canada, and it allows distinguishing between economic migrants, family migrants and refugees. As only 2016 data are available, sample sizes are relatively small, so that this chapter can derive only few results for Canada.

For additional analyses, the chapter draws on the German Socio-Economic Panel (GSOEP) for the years 1984-2016. This data set includes a number of observations for the same person in different years. Such longitudinal data allow separating the effect of (time-variant) family presence from the role of (time-invariant) personal characteristics of the sponsor or the spouse, notably unobserved personal characteristics or behaviours that affect integration outcomes. The data set also covers a range of migration-related variables, such as language spoken at home and social contacts. Because the annual sample size is comparatively small, the data are pooled across all available years.

Most analyses in this chapter draw on samples of matched migrant spouses derived from the data sets above. The EU LFS, the ACS, the Canadian Census and the GSOEP collect data on all adult members (aged 15 and above) of the sampled households. A household identifier allows determining which persons live in the same household, and the relationship between them can be established through further variables. Using these household links, data sets were assembled that associate the two spouses in a married couple but also retain married persons whose spouse is not observed. For the regression analyses, only migrants married to another migrant are retained. As only the ACS includes information on the year of marriage, it is often not clear whether the migrant pair was already married at arrival, and approximations have to be used to distinguish family reunification from family formation (see Table 4.1 for definitions).

For both spouses in a migrant pair, the data sets include a selection of demographic and socio-economic variables. Where a spouse does not live in the same household, information on them is typically not available in the original data, which leads to missing values for one spouse. This concerns about 4% of observations on married persons in the data set derived from the ACS and 14% of observations derived from the EU LFS. While the individual married person is the unit of analysis throughout this chapter, most individuals can therefore be described further by information on their spouse.

Descriptive evidence on the presence of migrants’ families

This section documents the extent of family presence among migrants in OECD countries, provides estimates of the delays involved in family reunification and identifies some groups where separation from family members appears especially frequent. The section focuses throughout on migrants who are married to another migrant. For close to 95% of married migrants, the spouse is present in the same household (Figure 4.1). This share ranges from 66% in Lithuania and 81% in Finland to 98% in the United Kingdom.

The variation in spouse presence across countries likely reflects different factors whose contribution can vary from one country to another. Firstly, the refugee surge in 2015/2016 resulted in large numbers of principal migrants arriving in some countries without their spouses. Secondly, some countries have larger shares of intra-EU migrants who are more often temporary migrants and therefore less likely to bring their family. Thirdly, some European countries have adopted restrictive policies regarding family reunification in recent years, as documented in Chapter 1.

Figure 4.1. Married migrants by presence of the spouse, 2013-17

Married migrants aged 15-64

Note: Figures for the United States refer to 2013-16, are limited to persons who were married at the time of arrival, and only the presence of related children under 18 can be identified in this case. Migrants in mixed couples are not included nor are legally separated migrants, with the exception that migrants in mixed couples where the native-born spouse is absent cannot be excluded from the base of the percentage.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html, and the 2016 Canadian Census (Statistics Canada), https://www12.statcan.gc.ca/census-recensement/2016/dp-pd/index-eng.cfm.

 StatLink https://doi.org/10.1787/888933989133

Across all durations of stay, Table 4.2 shows that less than 20% of married migrants live separately from their spouse in almost all OECD countries. This percentage is below 10% in about two-thirds of the cases and stands at 6% on average across OECD countries for which data are available. The lowest shares are observed in Ireland and the United Kingdom (below 4% in both cases). However, two groups of OECD countries tend to exhibit substantial shares of married migrants whose spouse is absent: some countries in Central and Eastern Europe (notably Latvia, Lithuania and Slovenia) and three Nordic countries (Denmark, Finland and Sweden). In Finland and Sweden, the high shares might reflect relatively large numbers of recently arrived refugees.

Around half (54%) of married migrants across OECD countries arrived in the host country in the same year as their spouse (Table 4.2). This share differs considerably between the United States (67%) and European OECD countries (44%). The high share for the United States partly reflects the fact that the data are restricted to migrants who were already married when they arrived, a restriction that is not possible for European OECD countries. Married migrants in OECD countries who arrived earlier than their spouse have spent on average four of their first ten years in the host country without their spouse. This average delay is somewhat lower for migrants in the United States (3.3 years) than in European OECD countries (4.0 years) and ranges from 3 years in Ireland to 5.5 years in Greece.

The share of married migrants whose spouse is absent is initially significant in the United States (11% for up to one year of stay) but quickly falls below 8% as duration of stay rises (Panel A in Figure 4.2). At durations of 20 years or more, only 4% of married migrants in the United States do not live with their spouse. Comparatively low shares of married migrants in Canada (here including mixed couples) do not live with their spouse: from an initial value below 5%, the share falls with duration of stay to below 2% at durations of 20 years or more. In European OECD countries, the corresponding share remains at a higher level, between 6% and 10%.

Further results using the same data differentiate married migrants in migrant couples by region of origin. A similar ranking emerges for European OECD countries and the United States in terms of shares of married migrants who live without their spouse. In both European OECD countries and the United States, spouses are more often absent among married migrants from Africa (reaching rates of 11% and 9% of married migrants, respectively). This does not appear to be driven by comparatively short durations of stay among migrants from Africa. Spouses are rarely absent among married migrants from Europe (4% of migrants from Europe in the United States and 6% of migrants from EU/EFTA countries in European OECD countries). The second lowest incidence of spouses being absent is observed for migrants from Asia, again both in the United States (4%) and European OECD countries (6%).

European OECD countries and the United States also exhibit very similar shares of married migrants with own children in the household (Panel B in Figure 4.2). Initially, slightly more than half of married migrants live with their own children both in European OECD countries and in the United States, then both shares rise steadily over time and reach a peak at 15-19 years’ duration of stay (attaining 75% and 80%, respectively). Shares for Canada are substantially lower at all durations of stay and only have a weak tendency to rise over time.

Table 4.2. Indicators for presence of the migrant’s spouse, 2013-17
Married migrants aged 15-64
	Stock of married migrants (annual average in thousands)	Share of married migrants whose spouse is absent (%)	Share of married migrants whose spouse arrived accompanying (%)	If not accompanying: years in the first 10 that migrants spent without a spouse
Austria	553	6.6	39.4	3.6
Belgium	493	10.7	42.0	4.0
Canada	2 445	4.3
Czech Republic	76	6.9	72.0	4.1
Denmark	111	16.9	44.3	3.9
Estonia	30	9.8	47.4	4.2
Finland	32	19.4
France	2 103	5.7	37.5	4.4
Germany	4 732	7.0	55.4	3.6
Greece	324	4.4	43.9	5.5
Hungary	44	4.8	73.0	4.4
Ireland	222	3.7	57.3	3.0
Italy	2 176	13.0	21.9	5.3
Latvia	29	19.0	49.1	5.3
Lithuania	20	33.6	80.2	5.1
Luxembourg	82	10.6	61.8	3.9
Netherlands	488	4.7	34.3	3.7
Norway	207		41.5	3.9
Portugal	128	9.8	39.1	4.3
Sweden	410	16.1	37.4	4.7
Slovenia	67	15.2	13.3	4.9
Spain	1 904	6.0	31.8	4.4
United Kingdom	2 962	1.8	50.2	3.1
United States	14 347	4.6	66.7	3.3
European OECD	17 201	7.2	43.6	4.0
OECD (24)	51 195	6.3	53.9	3.9
Note: Figures for the United States refer to 2013-16 and, in columns three and four, are limited to persons who were married at the time of arrival. Figures for Canada refer to 2016. Migrants in mixed couples are not included, nor are legally separated migrants. Figures in column 2 are slightly overestimated because migrants in mixed couples where the native-born spouse is absent are counted in. Column 3 is calculated based on both spouses arriving in the same year and cases where one spouse is absent are not included. Due to data limitations, figures for European countries in column 3 are based on migrants with duration of residence up to ten years. For the calculation of column 4, delay equals the duration of stay of the observed spouse in cases where the spouse is absent. Some figures cannot be calculated for Finland and Norway due to missing data.
Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html, and the 2016 Canadian Census (Statistics Canada), https://www12.statcan.gc.ca/census-recensement/2016/dp-pd/index-eng.cfm.
StatLink https://doi.org/10.1787/888933989361

Figure 4.2. Presence of migrants’ spouses and children by duration of stay, 2013-17

Married migrants aged 15-64

Note: Figures for the United States refer to 2013-16, are limited to persons who were married at the time of arrival, and only the presence of related children under 18 can be identified in this case. In Panel A, the initial part of the series for European OECD countries likely reflects small sample sizes or cohort effects. The series for Canada include migrants in mixed couples.

Source: OECD Secretariat analysis based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, the Canadian Labour Force Survey (Statistics Canada), https://www150.statcan.gc.ca/n1/en/surveys/3701, and the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

 StatLink https://doi.org/10.1787/888933989798

Figure 4.3. Differences in spouses’ year of arrival, 2013-17

Distribution of delays (up to nine years) for married migrants aged 15-64

Note: Figures for the United States refer to 2013-16 and are limited to persons who were married at the time of arrival. Figures for European OECD countries are based on migrants with up to ten years of residence. Migrants in mixed couples are not included. Cases where the spouse is absent are not included.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, and the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

 StatLink https://doi.org/10.1787/888933989817

Husbands tend to arrive with less delay than wives

The distribution of delays in spouses’ arrivals differs substantially between European OECD countries and the United States, and between husbands and wives (Figure 4.3). In the United States, 81% of husbands and 61% of wives who arrive within ten years already arrive accompanying. The corresponding shares are substantially lower in European OECD countries: 67% of husbands and only 39% of wives. After the first year of delay, the difference between the arrival of husbands and wives vanishes. Overall, longer delays appear relatively frequent in European OECD countries compared with the United States.1

Review of existing literature

The literature offers very few quantitative results on the effects of family presence on integration, except for a few sociological studies with very limited sample sizes. Other relevant literature explores how family presence changes the behaviour of principal migrants but does not quantify the effect on integration outcomes. At the same time, measuring this effect has been identified as a gap in the literature (see Tate (2011[2]), Bonjour and Kraler (2015[3]) and Charsley et al. (2017[4])).

A number of studies have investigated family separation in the context of refugees. Miller et al. (2018[5]) surveyed 165 refugees who were resettled to Albuquerque in the United States. They identify family separation as a major source of stress and document significant links with measures of mental health. A joint report by the Refugee Council and Oxfam (Beaton, Musgrave and Liebl, 2018[6]) offers qualitative evidence on the cases of 44 families. It suggests that refugees who are unable to reunite with family members suffer from stress and anxiety, are unable to focus on language courses, are burdened with family tasks that had previously been shared and sometimes develop mental health issues. Rousseau, Mekki-Berrada and Moreau (2001[7]) find that family separation can compound the effects of an existing trauma, based on evidence from 113 refugees in Montreal. Using panel data on refugees in Germany, Walther et al. (2019[8]) find that family separation is associated with higher levels of stress and lower levels of well-being.

Mlati and Duarte (2005[9]) surveyed 50 recognised refugees in France who were awaiting the outcomes of family reunification procedures. They report that refugees who had already reunited with a part of their family were more likely to focus on integrating than those who had not yet been reunited. Caplan’s (2007[10]) meta-analysis of studies on recent Latino migrants in the United States offers more systematic evidence and goes beyond refugees. Its conclusions point to migrants’ inability to reunite with their family as the most frequently cited source of stress. However, none of these studies attempts to link the problems described to measures of integration.

An assessment by Canada’s IRCC (2014[11]) investigated the role of family members for integration outcomes through a survey of 2 000 migrants who sponsored family members between 2007 and 2011. Among sponsors of spouses (or partners), many indicated that their spouse helped them settle in Canada (43%) and work more hours (40%); two-thirds said that the spouse contributed to household income. Some literature is available on ways in which family presence or absence might affect integration outcomes. Gracia and Herrero (2004[12]) find in a general context that stress and depression undermine social integration. Where the absence of the family induces stress, it could therefore undermine the migrant’s social integration. In turn, the presence in particular of the spouse might reduce stress and stabilise the migrant’s situation through a change of risky behaviour (Muñoz-Laboy, Hirsch and Quispe-Lazaro, 2009[13]).

Olwig (2011[14]) reports for Scandinavia that family relations significantly help new migrants and refugees to establish themselves, and similar findings are provided by the literature on the role of migrant networks. Insofar as a family has more relatives and can develop larger networks than a single person, this points to a supportive effect. Family life itself can broaden migrants’ networks: Facchini et al. (2015[15]) find that the birth of a child, for instance, increases the likelihood that the migrant parent has a native-born friend.

On the other hand, a number of studies find that strong family links correlate with lower participation in civil society (Kim and Wilcox (2013[16]), Ginsborg (1995[17])). Yet there are also studies that find complementarity between close social contacts and interaction with society at large (Nannestad, Lind Haase Svendsen and Tinggaard Svendsen, 2008[18]). Bragg and Wong (2016[19]) point out that, where family reunification fails due to regulations, migrants’ sense of belonging to the destination country society can be undermined.

Other ways in which spouses might affect integration outcomes relate to their financial contribution to the migrant household. Kaida (2015[20]) finds that migrant women contribute significantly to the household income of recent migrants in Canada, especially for low-income households. The income of the spouse can therefore prevent poverty. In addition, the income of the spouse can allow migrants the possibility of enrolling in further education (either full- or part-time) and eventually securing better jobs (Boyd, 1989[21]; Creese, Dyck and McLaren, 2008[22]). Similarly, other family members such as migrants’ parents can help with childcare, so that migrants are freer to work or pursue an education (VanderPlaat, Ramos and Yoshida, 2012[23]).

While spouses are still waiting to reunify in the destination country, their choices might be affected by the uncertainty about if and when the reunification will happen, and by frustration with the wait. In the context of asylum seekers, some recent publications have documented that initial waiting times – for the conclusion of the asylum procedure or for admission to the labour market – can significantly affect subsequent integration outcomes (Hainmueller, Hangartner and Lawrence, 2016[24]; Marbach, Hainmueller and Hangartner, 2018[25]; Brenzel and Kosyakova, 2019[26]). Such findings might also apply to migrants waiting for family reunification.

Challenges of measuring the role of family in integration

Isolating the effect of family presence on integration involves many challenges and potential bias. The bias can be in either direction, leading to either an overestimated or underestimated effect. Table 4.3 and Table 4.4 list problems for measurement and empirical identification strategies, together with potential methodological solutions. Endogeneity is a frequent problem. In the context of family reunification, better integration outcomes associated with the presence of a spouse may simply reflect that only successfully integrated migrants are allowed to bring their spouses.

If, for example, family reunification is conditional on the income and housing situation of the principal migrant, then principal migrants whose integration outcomes are favourable will be more likely to bring their families. The resulting correlation of family presence and favourable integration outcomes could then be falsely interpreted as a positive effect of family presence on integration outcomes. If, on the other hand, the migrant eventually gives up employment after the family has arrived because the spouse has found employment, this endogenous change of behaviour would register as a negative effect of family presence on employment prospects. Due to such concerns about endogeneity, results from a comparison between migrants whose spouse is present and migrants whose spouse is absent have to be treated with caution. Moreover, as only a small proportion of married migrants does not live with their spouse, it might be an unrepresentative selection.

Table 4.3. Measurement problems for the estimated effect of family presence
Problem	Possible remedy
Most effects can only be discerned after some time	Focus on long durations of stay (ideally 10+ years)
Partners joining later might be formation of new families	Define as family formation after a cut-off duration of stay
The principal migrant cannot always be identified	Focus on migrants who arrive ahead of their family
Observed effects from spouses are the combined effect of their presence and their characteristics. Characteristics of partners who are absent are unobserved	Focus on delays in the spouse’s arrival rather than the presence/ absence of the spouse. Account for observed characteristics of partners
Benefit payments for family members might reduce incentives for employment but are typically unobserved	Account for family size or the number of children
Variables that capture social integration (such as community engagement, volunteering) are often missing	Draw on specific data sets for this information

If the effects of family presence materialise slowly over time, they need to be distinguished from the tendency that integration outcomes tend to improve with duration of stay, which has been documented across OECD countries (OECD, 2016[27]). It is therefore important to ensure that effects from any measure of family presence are not confounded with effects from duration of stay. This has implications for the specification of the econometric estimation (see Box 4.2).

Table 4.4. Identification problems for the estimated effect of family presence
Problem	Possible remedy
Family presence is endogenous (I): certain residence permits confer this right immediately	Account for migration category of the principal migrant
Family presence is endogenous (II): integrating well makes it more likely that conditions for family reunification are met relatively quickly	Focus on delays in the spouse’s arrival rather than the presence/ absence of the spouse. Focus on long durations of stay (10+ years). Consider refugees and other migrants who are exempt from conditions
Behaviour of the principal migrant may change endogenously: once conditions for family reunification are met, might switch to family work, devote more time to family or leisure, or invest more in education	Focus on delays in the spouse’s arrival rather than the presence/ absence of the spouse. Consider changes in enrolment in education, hours worked, and employment status of the partner
Migrants who plan to stay only temporarily might not invest in integration and do not bring their family	Focus on long durations of stay (10+ years)
Those who fail to meet conditions for family migration often leave again	Consider refugees and other migrants who are exempt from conditions

Focusing on long-term effects of delays in spouse arrival

In order to identify the effect of delays in family reunification on labour market outcomes of the principal migrant, the considerations in Table 4.3 and Table 4.4 suggest focusing on long-term effects. Therefore, the analysis below will relate the number of years between the arrival of the principal migrant and the arrival of the spouse to integration outcomes observed after at least five or ten years of stay. The focus on long-term effects is expected to limit bias from endogeneity: while short-term integration outcomes could often determine how soon family members arrive, this is unclear for integration outcomes long after the family has arrived. Links between long-term integration outcomes and family presence during the initial years might well indicate that the family situation in the crucial first years after arrival has long-term consequences for integration. Alternatively, long-term integration outcomes could reflect short-term ones. The analyses below will seek to disentangle these two possibilities by using instrumental variable estimation.

A restriction to first-arriving migrants ensures that estimated effects refer to principal migrants. Estimates in this context can depend strongly on whether only principal migrants or also family members are included, as family members exhibit substantially less favourable integration outcomes than principal migrants, even in the long-term (Chaloff and Poeschel, 2017[1]). Given that most data sets do not indicate the principal migrant, this chapter identifies principal migrants based on who arrives first. When spouses arrive at the same time, it is not clear which spouse is the principal migrant. Accordingly, analyses of how later arrival of the spouse affects the integration of principal migrants focus on first-arriving migrants only. Analyses on how spouses themselves are affected focus on second-arriving migrants only.

Box 4.2. Estimation methods in the main analyses

The main analyses investigate how delay in the arrival of the spouse, measured in years, affects either principal migrants or the spouses themselves. The same estimation methods are used but the included variables differ. Analyses on principal migrants include delay in the arrival of the spouse as an explanatory variable, together with characteristics of the principal migrant and some characteristics of the spouse. These variables are used to explain an integration outcome of the principal migrant (the dependent variable). Analyses on the spouse include delay in the spouse’s own arrival, together with the characteristics of the spouse and some characteristics of the principal migrant. In this case, the dependent variable is an integration outcome of the spouse.

The explanatory variables on the other person in the migrant pair typically include employment status, level of education and host-country language proficiency (where this information is available). Whenever a particular integration outcome of one person in the migrant pair is investigated, the corresponding outcome for the other person is also included as an explanatory variable. The selection of explanatory variables thus changes somewhat from one analysis to another.

The analyses thus relate a migrant’s integration outcome (the dependent variable $Y_i$) to a range of variables that describe the migrant and the migrant’s spouse (explanatory variables). Where the integration outcome is a binary variable, such as employment status (equal to one if employed and equal to zero otherwise), the regression analyses use a linear probability model. In this model, the probability that the dependent variable equals one is determined by explanatory variables in a linear way:

$\mathrm{P}\mathrm{r}\left(Y_i=1\right|X_i)={\beta }_0+\mathrm{ }{\beta }_1{d}_i+X_i^{\mathrm{\text{'}}}{\beta }_2+Z_i^{\text{'}}{\beta }_3+{\gamma }_n+{\epsilon }_i$

The explanatory variables include a constant with coefficient ${\beta }_0$, the variable for delay $d_i$ with coefficient ${\beta }_1$, a vector $X_i^{\text{'}}$ of variables on migrant i and a vector $Z_i^{\text{'}}$ of variables on the spouse of migrant i. The term ${\epsilon }_i$ allows for random disturbances in the empirical relation and analyses for European OECD countries include a fixed effect ${\gamma }_n$ for each country n, which captures institutional differences and labour market conditions to some extent. Where the integration outcome is a continuous variable, such as wage and hours worked, the regression analyses use a log-linear model with the same explanatory variables:

$Y_i=\mathrm{e}\mathrm{x}\mathrm{p}({\beta }_0+\mathrm{ }{\beta }_1{d}_i+X_i^{\mathrm{\text{'}}}{\beta }_2+Z_i^{\text{'}}{\beta }_3+{\gamma }_n+{\epsilon }_i)$

The linear probability model is estimated using the method of ordinary least squares, which can also be used for the log-linear model once it is linearised by taking logarithms, so that ${\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{ }Y}_i$ becomes the dependent variable. The results offer estimates for ${\beta }_0, {\beta }_1,{\beta }_2,\mathrm{ }{\beta }_3$ as well as ${\gamma }_n$. This chapter refers to them as estimated effects or associations. Positive (negative) estimates imply a positive (negative) association between the dependent variable and the respective explanatory variable, after accounting for the association between the dependent variable and the other explanatory variables. As linear and log-linear models are used in for all estimations, the estimates can be easily interpreted.

Explanatory variables that are highly correlated with each other can mimic each other’s role for the dependent variable, so that the effect of one explanatory variable might be ascribed to the other explanatory variable. Therefore, it is important that delay is not too highly correlated with other explanatory variables. Table 4.5 shows the correlation between years spent without a spouse present and duration of stay. Although the analyses are limited to migrants with at least ten years of stay (at least five years in the analyses on later arriving spouses), the correlation can still be high, especially for married migrant women.

Table 4.5. Correlation between duration of stay and years of delay in the spouse’s arrival
Correlation coefficients for first-arriving married migrants aged 15-64, after 10+ years of stay
	European Labour Force Survey (2013-17)	European Labour Force Survey (ad-hoc modules 2008/2014)	American Community Survey (2013-16)	Canadian Census (2016)
All	0.72	0.60	0.14	0.57
Men	0.55	0.55	0.09	0.57
Women	0.81	0.67	0.21	0.56
Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, the European Labour Force Survey (Eurostat) ad-hoc modules 2008 and 2014 on the labour market situation of migrants and their immediate descendants, http://ec.europa.eu/eurostat/statistics-explained/index.php/EU_labour_force_survey_-_ad_hoc_modules, the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html, and the 2016 Canadian Census (Statistics Canada), https://www12.statcan.gc.ca/census-recensement/2016/dp-pd/index-eng.cfm.
StatLink https://doi.org/10.1787/888933989380

Therefore, the analyses for principal migrants and spouses do not include duration of stay as a single continuous variable but rather as a series of fixed effects for each value of duration of stay. This way of including duration of stay avoids the problem of high correlation between explanatory variables. At the same time, it ensures that the estimation procedure compares observations with the same value for duration of stay. This basic matching technique therefore serves to isolate the estimation results better from the role of duration of stay. However, almost all reported results of the main analyses arise very similarly also in regressions that simply include duration of stay as a continuous variable. In these regressions, the coefficient for duration of stay is almost always statistically significant and indicates that integration outcomes tend to improve with duration of stay.

Estimates also appear strongly affected by the extent to which cases of family formation are included (see Table 4.1 for a definition). To ensure that a large majority of cases included in the estimation are family reunification rather than family formation, two further restrictions are made. Almost all cases of family reunification should occur within a few years of the principal migrant’s arrival, while a large share of family formation might occur after the principal migrant has already spent a number of years in the host country. To focus on family reunification, the analyses below are therefore limited to cases in which the spouse arrives within seven years of the principal migrant’s arrival. In addition, data used for the United States include information on migrants’ marital status at arrival, which is used to limit these analyses to migrants who were already married when they arrived.

The role of spouses in long-term integration outcomes

This section presents the empirical results on how principal migrants may be affected by delays in the arrival of their spouse. Box 4.2 above briefly presents the estimation methods that use micro data from large-scale surveys (Box 4.1). Estimations in this section account for a range of individual characteristics, notably age, gender, educational attainment, duration of stay and the region of origin. They are included both for the principal migrant and for the spouse, so that the estimated effect from delays in the spouse’s arrival should not be driven by the characteristics of the spouse. It should rather reflect unobserved, hard-to-measure aspects that may be associated with spouse presence: stability, relying on each other and sharing burdens, long-term planning and investment behaviour, an orientation towards communication and activities of the couple, changes in work intensity, changes in emotional well-being, etc. In order to ensure that presence of spouse is not confounded with a higher likelihood of children being present, separate variables for the presence of (young) children are always included. To focus on cases of family reunification rather than family formation, only delays up to seven years are considered.

Delay in spouse arrival is associated with lower wages for the principal migrant in the long-term

Migrants’ wages may be a particularly informative indicator of integration: higher wages can be attained over time as the result of growing work experience, investment in education and better matching between the migrants’ skills and the requirements of their jobs. Wages can therefore reflect long-term integration success better than other, more transitory, integration outcomes. Panel A of Figure 4.4 shows estimated effects of delays in the spouse’s arrival on gross wages earned by the principal migrant after at least ten years in the United States. The results suggest that each year of delay in the spouse’s arrival is associated with principal migrant’s wages being 3% lower in the long-term. The estimated effect arises similarly for male and female principal migrants. As explained below, this significant effect on wages likely reflects several factors.

The data for European OECD countries include information on net wages but only in terms of the decile reached in the wage distribution. The analysis for European OECD countries therefore examines the principal migrant’s probability of earning net wages above the median net wage, after at least ten years in the host country. As shown in Panel B of Figure 4.4, this probability decreases by 1.5 percentage points with each year of delay in the spouse’s arrival. This finding is driven by male principal migrants, for whom a decrease of the same magnitude is observed. By contrast, delays in the spouse’s arrival seem to leave the probability for female principal migrants unchanged.

Since the measures of wages differ between Panel A and Panel B of Figure 4.4, the results are not directly comparable. In particular, the estimates in Panel B do not necessarily suggest a weaker effect in European OECD countries: in this analysis, only changes that lift wages above the median wage level are taken into account, while other changes along the distribution of wages are neglected. For Canada, information on (rounded) wage levels is available from the 2016 census. However, an estimated effect obtained from these data was too small to be statistically significant.

Figure 4.4. Estimated long-term effects of delays in spouse arrival on principal migrants’ wages, 2010-17

First-arriving married migrants who are aged 15-64 and employed, after 10+ years of stay

Note: All reported results are statistically significant at the 5% significance level. Figures for the United States refer to 2013-16 and are limited to persons who were married at the time of arrival. Migrants in mixed couples are not included. Norway and Sweden are not included in Panel B due to a lack of information on wages.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, and the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

 StatLink https://doi.org/10.1787/888933989152

One possible explanation of the results in Figure 4.4 could be the principal migrant’s behaviour after the arrival of the spouse. While it is still unclear whether their spouse and family will join them, principal migrants cannot rely on staying in the host country in the long-term. This might undermine their incentives to make investments that are specific to the host country. Such investments include securing additional formal qualifications, making applications for the recognition of foreign qualifications, and building up specific work experience, all of which can raise wages in the long-term. Once the spouse has arrived, principal migrants might not only have more incentives for such investments but the spouse might also support them in the implementation. This phenomenon is well documented in the literature, known as the family investment hypothesis (see e.g. Long (1990[28]) or Cobb-Clark and Crossley (2004[29])).

However, an alternative explanation for the results in Figure 4.4 would regard delays in the spouse’s arrival as a consequence of low wages of the principal migrant. Legislation in a number of OECD countries ties family reunification to a sufficiently high income of the principal migrant or certain migrant categories (see Table A.1 in OECD (2017[30])). Principal migrants who initially have low wages or arrived as temporary migrants can therefore face higher hurdles for family reunification. Principal migrants who are not constrained by such legislation might also choose of their own volition not to bring their spouse to the host country until they have reached a certain level of income, as spouses often need to be supported, at least initially. With high initial wages, migrants might also decide more quickly to stay and therefore bring their family sooner. To the extent that wages are correlated over time, this could induce an empirical link between delays in the spouse’s arrival and the principal migrant’s wages being lower in the long-term.

Further evidence points to a causal effect on the principal migrant’s wages

To distinguish between these two competing explanations, techniques have to be used that resolve the problem of endogeneity by determining in which direction a correlation should be interpreted. The estimation for the United States is carried out again using an instrumental variable for the spouse’s delay. This estimation examines the association between the principal migrant’s wage and an instrumental variable that is related to the spouse’s delay but unaffected by the wage (and therefore not endogenous). If a significant association is found, this will provide a strong indication that the spouse’s delay causes the observed change in the principal migrant’s wage, rather than the reverse causality.

A suitable instrumental variable for the spouse’s delay is found within the same data set: the disability status of the spouse is empirically linked with the spouse’s delay while it is unrelated to the wage of the principal migrant.2 It appears that spouses with disabilities exhibit significantly higher delays, which might reflect the need for additional arrangements before a move can take place, with regards to facilities in the new location or issues regarding health insurance. It also seems plausible that the spouse’s disability status does not affect the principal migrant’s wage directly, but only indirectly through delaying the spouse’s arrival.

The instrumental variable estimation produces a statistically significant result for disability status of the spouse (at the 5% significance level). The estimate suggests that delay in the spouse’s arrival has a causal effect on the principal migrant’s wage, reducing it by about 9%. This estimated effect is larger than in Figure 4.4 partly because disability status is a binary variable, so that its effect corresponds to the cumulative effect of several years of delay.3 Overall, this estimation provides evidence that delay in the spouse’s arrival significantly reduces the wage of the principal migrant.

Estimated long-term effects on the principal migrant’s employment probability are positive but limited

Figure 4.5 shows the estimated effects of delays in the spouse’s arrival on the principal migrant’s employment probability after at least ten years in the host country. The results for European OECD countries, the United States and Canada all suggest that an additional year of delay is associated with a slight increase of principal migrants’ employment probability, on average by one percentage point or less. These consistently positive estimates could arise for a number of reasons. Principal migrants who find themselves in the host country without their spouse might often focus on work, which could reflect their choice or the necessity to make a living and prepare for the arrival of family members. This early drive towards employment could lead to long-term effects on principal migrants’ employability and their commitment to the labour market, and therefore the long-term effects shown in Figure 4.5. The long-term effects could be especially large for groups of principal migrants whose labour market participation tends to be comparatively low, so that some work experience in the host-country has a comparatively large impact. This interpretation would align with the larger effect estimated for women in the United States.

Figure 4.5. Estimated long-term effects of delays in spouse arrival on principal migrants’ employment probability, 2010-17

First-arriving married migrants who are aged 15-64 and employed, after 10+ years of stay

Note: Reported results are statistically significant at the 5% significance level, except the results for “all” in the United States and in European OECD countries (10% significance level), as well as the result for “men” in Canada. Figures for the United States refer to 2013-16 and are limited to persons who were married at the time of arrival. Figures for Canada refer to 2016. Migrants in mixed couples are not included.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html, and the 2016 Canadian Census (Statistics Canada), https://www12.statcan.gc.ca/census-recensement/2016/dp-pd/index-eng.cfm.

 StatLink https://doi.org/10.1787/888933989171

The results in Figure 4.5 might also reflect the behaviour of migrants who come to OECD countries primarily to work, giving less priority to family reunification. This includes labour migrants who expect to stay for a limited time only and therefore initially choose not to exert their right to family reunification. They might arrange for family reunification only after their stay turns out to be longer. Similarly, migrants who arrive irregularly do not have a right to family reunification but might gain this right once their situation in the host country is regularised. Therefore, work as initial motive for migration might be empirically associated with comparatively high delays in family reunification.

Given that employment probabilities of principal migrants tend to be high, the magnitude of the estimated effects in Figure 4.5 appears limited overall, with the exception of women in the United States. Current employment status does not necessarily capture migrants’ long-term integration success well – migrants might be temporarily out of employment for a variety of reasons, from adverse economic conditions to unobserved individual circumstances. Therefore, delays in the spouse’s arrival at the beginning of the stay in the host country can be expected to have rather small effects on principal migrants’ employment status after ten years or more. Weekly hours worked, as a measure for employment intensity, were also considered. However, statistically significant results did not emerge for this integration outcome. This could reflect that hours worked tend to vary more strongly than current employment status, further loosening the empirical link with delays in the spouse’s arrival.

Box 4.3. Well-being of principal migrants whose spouse is absent

How absence of the spouse affects the well-being of the principal migrant is an important question that cannot be addressed using the main data sources (see Box 4.1). As pointed out in the review of the existing literature, it is often suspected that the absence of the spouse causes stress and suffering to the principal migrant, to an extent that might undermine their integration. The Gallup World Poll provides a range of variables that allow investigating individual well-being. The sample is limited to married migrants in OECD countries, and principal migrants whose spouse is absent are identified as married migrants who live alone (the size of their household equals one). After these restrictions, the sample sizes for most variables are still considerably larger than in the existing literature.

Table 4.6 shows how the various measures of well-being differ between principal migrants whose spouse is absent and all married migrants in OECD countries. Most results are very similar across groups. This notably includes the same incidence of stress. Principal migrants whose spouse is absent only appear to have a somewhat less positive attitude to the future. Overall, these findings suggest that migrants’ well-being is not affected by the absence of their spouse. While asylum seekers and refugees whose spouse is still in a conflict zone often experience stress and anxiety (Miller et al., 2018[5]; Walther et al., 2019[8]), this group is either not well covered in the data or is too small to have an impact on the average.

Table 4.6. Well-being of principal migrant by presence of spouse, 2009-18
Married migrants in OECD countries (all ages)
Statement	Applies and spouse is absent (%)	Applies and spouse is present (%)	Total replies of yes or no (absent/present)
Well-being (index <30)	11	10	440/14897
Well-being yesterday (index >50)	79	79	440/14897
Positive attitude (index <30)	41	39	444/15289
Positive attitude to future (index >50)	41	45	444/15289
Experienced stress yesterday	35	35	438/14833
Experienced happiness yesterday	79	78	218/6804
Personal health (index <30)	12	9	440/14897
Health problems comp. to others of same age	25	22	432/14691
Satisfied with health	81	81	179/5432
Note: Responses are used as unweighted frequencies. Missing responses and “do not know” are not counted towards the base of the shares
Source: Gallup World Poll 2009-18, https://www.gallup.com/analytics/232838/world-poll.aspx.
StatLink https://doi.org/10.1787/888933989399

Some limited differences appear when health measures are considered (Table 4.6): principal migrants whose spouse is absent appear to suffer somewhat more frequently from bad health, for instance, although they report the same satisfaction with their health as married migrants reunited with their spouse. While further analyses would be required before drawing a conclusion, a negative effect of spouse absence on health would seem consistent with the literature: the so-called “protective effect of marriage” has been widely documented (see e.g. Rendall et al. (2011[31]), Robards et al. (2012[32])), and this effect is probably weaker when spouses do not live together.

Effects of delays on principal migrants’ language proficiency are not detected

In all data sets used in this chapter (see Box 4.1), the principal migrant’s proficiency in the host-country language has been considered as an integration outcome that could be affected by delays in the spouse’s arrival. While statistically significant correlations can arise in this context, the estimated effects remain small and are highly unstable across different estimations. Ultimately, neither a positive nor a negative link has emerged from these analyses. To some extent, this might reflect measurement problems associated with language proficiency: the data typically include it as a binary variable, so that small changes in language proficiency over time are difficult to trace. In addition, many observations have to be excluded in this context, such as those for migrants who are native speakers of the host-country language.

The literature offers only a few insights into how the presence of family members might affect migrants’ proficiency in the host-country language. A recurrent finding is that the two spouses often exhibit similar levels of language proficiency. For example, Chiswick and Miller (2002[33]) documented a strong correlation between the level of proficiency in English for principal migrants and their spouses arriving in Australia. Such correlations also emerge in the data sets used in this chapter: after accounting for the correlation of language proficiency with individual variables such as educational attainment, age and employment status, a strong positive correlation still arises with the language proficiency of the spouse.

Some recent findings further suggest that a lack of exposure to the host-country language can undermine its acquisition. From a survey of the literature, Saleh AlHammadi (2016[34]) concludes that migrants are less likely to reach proficiency in the host-country language in the presence of a large community of similar migrants. Danzer and Yaman (2016[35]) find that the language proficiency of migrants in Germany is negatively affected when they live in areas where persons from the same ethnicity are concentrated. They attribute this finding to limited contact with the native-born.

Medium and long-term effects of delayed arrival on spouses

Building on the empirical approaches used in the previous section, this section investigates how delays in the arrival of spouses are linked with their own integration outcomes. Delays could affect the integration of the spouses themselves for a number of reasons, including the disruptive effects of being separated from their partner for years, limited influence on how life in the host country is set up, or changes in the distribution of roles within the couple. As a result, spouses might encounter greater difficulties with integration or have a lower inclination towards integration when they eventually arrive, potentially with long-term consequences. However, such factors again need to be distinguished from alternative explanations: delays could signal existing problems or hesitations that subsequently also affect spouses’ integration outcomes.

While the same data and definitions as before are used in this investigation, the focus is now exclusively on married migrants who arrive second and who are therefore not the principal migrant. In order to prevent results from being driven by family formation instead of family reunification, again only delays up to seven years are considered. The main difference concerns the time factor: the analyses below consider spouses’ integration outcomes after at least five years of stay in the host country, rather than ten years, in order to use a larger sample.4

The results appear driven by married migrant women (Figure 4.6): in European OECD countries, an additional year of delay is associated with a decrease in their employment probability by two percentage points. The decrease found on average is somewhat lower. For Canada, the analysis suggests an average decrease in the employment probability of one percentage point. In the United States, an additional year of delay is associated with a slight increase in employment probability (below one percentage point), which also appears driven by migrant women.

Figure 4.6. Estimated long-term effects of delays in spouse arrival on their integration outcomes, 2010-17

Second-arriving married migrants who are aged 15-64, after 5+ years of stay, in percentage points

Note: All reported results are statistically significant at the 5% significance level except the result on language proficiency for “all” in the United States and the result for Canada (at 10% significance level). Figures for the United States refer to 2013-16 and are limited to persons who were married at the time of arrival. Results for language proficiency in European OECD countries refer to 2008/2014. Figures for Canada refer to 2016. Migrants in mixed couples are not included. Migrants in European OECD countries whose native language coincides with the host-country language are excluded from the analysis.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview including ad-hoc modules 2008 and 2014 on the labour market situation of migrants and their immediate descendants, http://ec.europa.eu/eurostat/statistics-explained/index.php/EU_labour_force_survey_-_ad_hoc_modules, the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html, and the 2016 Canadian Census (Statistics Canada), https://www12.statcan.gc.ca/census-recensement/2016/dp-pd/index-eng.cfm.

 StatLink https://doi.org/10.1787/888933989190

In terms of proficiency in the host-country language, results for both the United States and European OECD countries suggest that delays are associated with a decrease in the probability of being proficient, by one-half of a percentage point and two percentage points, respectively. This finding again arises very similarly for migrant women but not for migrant men. However, there are few observations on men, since a large majority of second-arriving migrants are women. Therefore, it cannot be ruled out that the results for women might extend similarly to men.

The estimated decreases in the employment probability and the probability of being proficient in the host-country language could be a consequence of the delays in arrival. Principal migrants might have a first-mover advantage, so that arrangements in the host-country primarily reflects their needs and supports their employment. The later-arriving spouse may have to take many parameters as given, so that they face relatively limited choices of jobs or training opportunities. For example, the location in the host country might have been determined entirely by the employment of the principal migrant, and this choice might not be easily reversed.

Similarly, if one of the two spouses needs to focus on raising children, this role will more likely fall on the later-arriving spouse, given that the principal migrant has often already secured employment in the host country. For the later-arriving spouse, such initial circumstances could have medium and long-term effects because they accumulate less relevant work experience that would qualify them for future employment or maintain their skills from previous jobs. A lack of work experience or training opportunities might also affect their acquisition of the host-country language.

However, the reverse causality is also possible: spouses who find the transition to the host country more difficult might arrive with longer delays. For example, spouses with lower proficiency in the host-country language might choose to arrive later. Where spouses search for a job as a precondition for their arrival, more employable spouses would find a job sooner, which could generate the observed decreases in employment probability associated with longer delays. While the estimation accounts for a range of individual characteristics, unobserved characteristics such as previous work experience in the origin country could be the drivers in this context.

The path of integration can differ considerably for late-arriving spouses

Most likely, both the delay itself and the characteristics of the later-arriving spouses play a role in their integration, leading to integration paths that can differ widely between spouses who arrive soon and those who arrive later. Based on German longitudinal data, Figure 4.7 shows the evolution of spouses’ employment probability with duration of stay. Apart from the initial level, the focus on changes observed for the same person ensures that influences from most individual characteristics cancel out. The initial level appears as a strong determinant of the spouse’s integration path in terms of employment. It likely reflects various characteristics of the spouse, whose role is in this case not accounted for. Conditions for family reunification often favour high-skilled labour migrants, for example, and their spouses tend to have relatively high educational attainment themselves. Similarly, criteria for labour market access may be more generous for spouses of high-skilled labour migrants.

In line with the results above, employment probability is consistently lower for spouses who arrive with longer delays (Figure 4.7). Spouses who arrive with 2-5 years delay only exhibit about half the initial employment probability (at 1-5 years of duration of stay) that is observed for spouses arriving with one-year delay (34% compared with 62%). The subsequent evolution of the employment probability does not offset the initial difference.

Figure 4.7. Evolution of spouse employment probability by delay in their arrival, Germany, 1984-2016

Average within-person changes of second-arriving married migrants aged 15-64, over years of residence in the host country

Note: Apart from initial (average) levels, changes are the average differences in employment status (0 or 1) for the same migrant observed at different years of residence. Migrants are not observed in all years of residence.

Source: OECD Secretariat analyses using the German Socio-Economic Panel, https://www.diw.de/en/soep.

 StatLink https://doi.org/10.1787/888933989323

Similarly, spouses arriving with greater delay exhibit a substantially lower initial probability of being proficient in the host-country language (Figure 4.8). They also exhibit lower growth in this probability as duration of stay increases, compared with spouses who arrived with one-year delay. In terms of monthly wages (not shown), there is hardly any difference in the initial levels of the two groups (about EUR 1 400), but spouses with greater delay exhibit somewhat lower wage growth. A larger difference arises for wages of women, whose initial wage is significantly lower in the group with 2-5 years delay (EUR 1 000 compared with EUR 1 400).

Strongly different integration paths of spouses who arrive with substantial delay could explain earlier findings on poor integration outcomes of family migrants in comparison to principal migrants. Chaloff and Poeschel (2017[1]) document that the labour market integration of adult family migrants is roughly as slow as that of humanitarian migrants. Liebig and Tronstad (2018[36]) emphasise the triple disadvantage encountered by wives of humanitarian migrants who often also lack work experience. Lochmann et al. (2018[37]) find that adult family migrants in France benefit less from language courses than other migrants. While the barriers to the integration of spouses are not fully understood, the circumstances of their arrival could play a role.

Figure 4.8. Evolution of spouse proficiency in host-country language by delay in spouse arrival, Germany, 1984-2016

Average within-person changes of second-arriving married migrant women aged 15-64,over years of residence in the host country

Note: Apart from initial (average) levels, changes are the average differences in proficiency in the host-country language (0 or 1) for the same migrant observed at different years of residence. Migrants are not observed in all years of residence.

Source: OECD Secretariat analyses using the German Socio-Economic Panel, https://www.diw.de/en/soep.

 StatLink https://doi.org/10.1787/888933989342

Box 4.4. Conditions for family reunification can cause delays: Evidence from Germany

This section investigates the link between conditions for family reunification and delays in the arrival of the spouse. The evidence is based on a quasi-natural experiment in Germany, where a pre-arrival language requirement for most non-EU spouses was introduced in 2007. Other conditions imposed in OECD countries include requirements for principal migrants’ income, their housing situation, or the length of their residence in the host country. If such conditions contribute to delays, they will indirectly contribute to the effects of delays that are documented in this chapter.

One study from Norway investigates how family reunification was affected when income requirements were extended to migrants admitted on humanitarian grounds (Bratsberg and Raaum, 2010[38]). For the affected principal migrants, the study finds that the change caused a decline in family reunifications by 21 percentage points, increased the employment rate by 4-8 percentage points, significantly increased earnings and reduced receipt of social assistance. Schmidt et al. (2009[39]) document that, following the introduction of a minimum age (24 years) for reunification with a spouse in Denmark, the number of reunifications decreased especially in the age group 20-23 and the age at marriage rose.

However, there does not seem to be an econometric evaluation of how delays are linked to conditions for family reunification. To provide such an evaluation here, data from the German Socio-Economic Panel are used. In late 2007, new legislation changed the conditions for family reunification: prior to their arrival, spouses need to demonstrate basic skills in German (level A1 in the Common European Framework of Reference) and had to be at least 18 years old (previously 16 years). However, this requirement only applies when neither the principal migrant nor the spouse are citizens of an EU/EFTA country or citizens of certain non-EU countries (who can travel visa-free to Germany): Andorra, Australia, Brazil, Canada, El Salvador, Honduras, Israel, Japan, Korea, Monaco, New Zealand, San Marino and the United States.

This set-up with groups of countries allows for an evaluation of the causal effect of the policy change on delays, using the method of difference-in-difference (DID). After accounting for some observed factors that influence delays, a difference in delays remains between non-EU nationalities that were affected by the law change and those that were unaffected. Provided other changes over time apply similarly to both groups, a change in that difference after 2007 can be related to the law change. Nationals of countries where many refugees originated are excluded from the analysis, as requirements often did not apply to refugees. As applications for family reunification were normally processed in a few months, there is no major problem of rules overlapping around the time of the change.

Figure 4.9. Average delays before and after a policy change in Germany, 1984-2016

Conditional average delays in years incurred by married migrants aged 18-64

Note: Conditional averages were obtained as a linear prediction from a regression analysis of delays, which accounted for gender, age, indicators for high and for medium-level education, employment status, wage (in logarithms), duration of stay in the host country, and an indicator for affected nationalities. The assignment to time periods is based on the year of the spouse’s arrival.

Source: OECD Secretariat analyses using the German Socio-Economic Panel, https://www.diw.de/en/soep.

 StatLink https://doi.org/10.1787/888933989209

Figure 4.9 shows the main result of the analysis, based on observations on 1 800 individuals, 800 of whom arrived after 2007 (including 460 who were affected by the law change). After accounting for some observed factors, the affected nationalities exhibited a somewhat higher average delay relative to the non-affected nationalities already before the law change: 4.2 years compared with 3.8 years. This difference increased by 0.4 years (about five months) after 2007, which is likely driven by the law change. In 2008-16, average delays of both groups are lower than in 1984-2007, which may be a statistical artefact: given the end of the observation period in 2016 and relatively high numbers of individuals who arrived after 2007, relatively many short delays are observed after 2007 while many long delays were still running in 2016.

Regression analyses of delays before and after 2007 suggest that average delays for affected nationalities even increased by 1.5-2 years. While these results are significant and other results of the same regressions are plausible, the statistical artefact likely biases the estimate of the effect. However, the basic result that the gap in delays widened significantly after 2007 appears robust: it still materialises when observations are matched on duration of stay or on duration of the marriage, and after collapsing observations on the same individuals in several years into a single observation in either period (up to 2007 and after 2007). This analysis therefore provides evidence that stricter conditions for family reunification can lead to significant increases of the delays in reunification.

Age at time of immigration and long-term integration outcomes of migrant children

This section highlights that the integration of migrant children depends on how quickly they join their parent(s) in the destination country. Cross-country evidence indicates that migrant children who spend their early years in the destination country achieve substantially better integration outcomes, compared with migrant children who only arrive when they are close to adulthood. This notably applies to educational attainment: Figure 4.10 shows the baseline proficiency gap between foreign-born and native-born 15-year-olds in reading, mathematics and science. In most OECD countries, this gap is considerably smaller for foreign-born students who arrived before the age of 12 than for foreign-born students who arrived later. The reverse only occurs in a few OECD countries where the gap between foreign-born and native-born 15-year-olds is comparatively low.

Figure 4.10. Differences in baseline academic proficiency of 15-year-old migrants and native-born, by age at arrival, 2015

Difference between migrant and native-born students in the percentage of students attaining baseline academic proficiency, in percentage points

Note: Students who attain baseline academic proficiency are students who reach at least PISA proficiency level two in all three PISA core subjects – math, reading and science. The age of students can range from 15 years and three months to 16 years and two months.

Source: (OECD, 2018[40]) based on OECD PISA 2015 Database, http://www.oecd.org/pisa/data/2015database.

 StatLink https://doi.org/10.1787/888933989228

In recent years, a number of in-depth studies have found significant effects of children’s age at arrival on integration outcomes in various OECD countries. Most of these studies derive their results from cross-section data by linking the exact age at arrival to integration outcomes later in life, while accounting for other factors. Some studies – Böhlmark (2008[41]), Åslund, Böhlmark and Skans (2015[42]), Hermansen (2017[43]) and Lemmermann and Riphahn (2018[44]) – can draw on data for siblings who arrived at different ages, which allows them to give their results a strong causal interpretation.

Most of these studies focus on education outcomes by age at arrival. For migrants in Canada and the United States, Schaafsma and Sweetman (2001[45]), Gonzalez (2003[46]) and Chiswick and DebBurman (2004[47]) find that those who arrived as young children stay in school longer than those who arrived as teenagers, attaining a higher total number of years of schooling. Similarly, early arrival is linked to reaching a higher educational level in the United States (Myers, Gao and Emeka, 2009[48]; Lee and Edmonston, 2011[49]), Norway (Hermansen, 2017[43]), Germany (Lemmermann and Riphahn, 2018[44]) and the United Kingdom (Aoki and Santiago, 2018[50]). Böhlmark (2008[41]) finds for Sweden that arriving later than at age nine has adverse effects on average grades in school. Beck et al. (2012[51]) document a higher probability of dropping out from high school for those arriving in the United States after age eight, and Cohen-Goldner and Epstein (2014[52]) document a similar pattern in Israel.

Several of the same studies further find that arriving later adversely affects measures of proficiency in the host-country language. The estimated effects can be large: for example, Myers et al. (2009[48]) conclude that, compared to migrants who arrive as teenagers, those who arrive before the age of six are six times as likely to reach the end of high school and almost fourteen times as likely to attain a high proficiency in the host-country language. Language proficiency may well be one of the key drivers behind the link between age at arrival and educational outcomes. This aligns with the critical ages found in several studies, of nine years and younger: Bleakley and Chin (2010[53]) refer to the critical period of language acquisition, and Beck et al. (2012[51]) argue that the important transition from “learning to read” to “reading to learn” can be disrupted by migration. Based on test results for children of Hispanic immigrant parents in the United States, Mukhopadhyay (2018[54]) reports that the acquisition of English appears to be undermined from age six by an already existing knowledge of Spanish. However, Myers et al. (2009[48]) as well as Lee and Edmonston (2011[49]) do not find evidence of a critical age.

Arriving at a later age also appears to have an impact on migrants’ employment outcomes and wages. Schaafsma and Sweetman (2001[45]) and Gonzalez (2003[46]) find lower returns to education for those who arrived later, which the latter study attributes to the greater share of schooling received abroad. Through the link with proficiency in the host-country language, age at arrival can affect employment outcomes, as reported by Guven and Islam (2015[55]) for migrants in Australia. In addition, Hermansen (2017[43]) estimates an effect of age at arrival on receipt of social assistance in Norway. Using historical data, Alexander and Ward (2018[56]) document a link between age at arrival and wages for brothers who migrated from Europe to the United States between 1892 and 1924. Recently, a younger age at arrival has also been linked with higher levels of social integration, notably intermarriage and residential proximity (Bleakley and Chin (2010[53]) and Åslund et al. (2015[42])).

Findings of adverse long-term effects from arriving at later age generalise

The broad data sets used in this chapter allow for an examination of whether key findings of this literature generalise beyond specific countries and migration cohorts. To this end, the standard methods used in the literature are also applied here. Analyses that use age at arrival as an explanatory variable suffer from strong correlations between explanatory variables, notably age and duration of stay. This problem is overcome by focusing on gaps between native-born and foreign-born persons, after accounting for age (see Schaafsma and Sweetman (2001[45]), for example). Where language proficiency is analysed, however, a simpler solution is proposed here: once a migrant has reached adulthood, age as such is unlikely to matter for language proficiency, as long as duration of stay is still accounted for. Then correlation among explanatory variables can be avoided by dropping age from the analysis.

The analysis focuses on adult migrants who arrived as children, comparing those who arrived aged 0-6 to those who arrived aged 7-15. Migrants who arrived at later ages are excluded because they represent a very heterogeneous and probably more strongly selected group. While only some results are obtained (Figure 4.11, Panel A), they align with findings in the literature and provide some evidence on wages, which has received little attention in previous studies. In the United States, the probability of being proficient in English is 20 percentage points higher for migrants who arrived aged 0-6, compared with migrants who arrived aged 7-15. The gap between the employment probability of native-born persons and migrants is significantly smaller (around five percentage points) for migrants who arrived aged 0-6 rather than aged 7-15. Similarly, the gap between wages of native-born persons and wages of migrants is somewhat smaller in their case (by 2%).

For European OECD countries, the only statistically significant result concerns migrants’ probability to attain a high education level (i.e. tertiary education). Clarke (2016[57]) emphasises that estimated effects of age at arrival on educational attainment will be biased substantially if parents’ education is not accounted for, as is the case in virtually all existing studies. The result in Panel B of Figure 4.11, however, was obtained using the ad-hoc modules 2009 and 2014 of the European Labour Force Survey, which include this information on parents’ educational attainment (also when parents are not present in the household). Because education typically continues well beyond age 15, the analysis was also restricted to persons aged 25-64. The resulting estimate suggests that the gap between the probability of native-born persons to attain a high education level and the corresponding probability of migrants is substantially lower for migrants who arrived aged 0-6 (by nine percentage points), compared to migrants who arrived aged 7-15.

Results for Canada compare migrants who arrived aged 0-4 to migrants who arrived aged 10-14, due to a different grouping used in the Canadian data. Migrants who arrived aged 0-4 exhibit a slightly higher probability of being proficient in one of Canada’s official languages (Figure 4.11, Panel A). The gap with native-born persons in terms of employment probability is significantly smaller for migrants who arrived aged 0-4 than for migrants who arrived aged 10-14 (Figure 4.11, Panel A). At six percentage points, the latter result is closely in line with the result for European OECD countries.

Figure 4.11. Estimated effects of arrival in early childhood on migrants’ long-term integration outcomes, 2013-16

Migrants aged 15-64 who arrived in the host country at age 0-6, compared to those arriving at age 7-15

Note: All reported results are statistically significant at the 5% significance level. Figures for the United States refer to 2013-16. The result for probability of high education in European OECD countries refers to 2009/2014 and are limited to persons aged 25-64. The result for Canada compares age at arrival 0-4 years to 10-14 years. Migrants in European OECD countries whose native language coincides with the host-country language are excluded from the analysis.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview including ad-hoc modules 2009 and 2014, http://ec.europa.eu/eurostat/statistics-explained/index.php/EU_labour_force_survey_-_ad_hoc_modules, and the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

 StatLink https://doi.org/10.1787/888933989247

Box 4.5. Effects of family separation on those left behind

While the analyses in this chapter focus on the role of migrants’ families in integration in destination countries, family separation is also likely to have effects on the situation of those family members who stay behind in origin countries. These effects can come to matter for destination countries when the family eventually reunifies in the destination country: for example, family members whose development in the origin country has suffered due to family separation may face greater challenges of integrating into the destination country. Therefore, family admission policies should take the effects of family separation in origin countries into account.

For a number of reasons, it is difficult to establish which effects on family members who stayed behind were caused by family separation (see Antman (2013[58]) for an overview). In general, the situation of the family in the origin country may be the cause for migration, not a consequence of migration. Where a stressful situation in the origin country caused one of the family members to seek employment or refuge abroad, for example, effects of this situation on those staying behind could be mistaken for effects of family separation.

In addition, effects from family separation need to be distinguished from effects of remittances or, if the migrant is unable to send remittances, from the family’s loss of income. For spouses who stay behind, remittances have the well-documented effect that spouses decrease their labour supply, possibly associated with an increase in non-market work. After family reunification, the spouses’ resulting lack of work experience might then undermine integration into the labour market in the destination country.

With regards to children staying behind, studies have frequently reported a positive effect of migration on their schooling, especially for girls. For example, Hanson and Woodruff (2003[59]) estimate that daughters of low-educated Mexican parents who emigrated to the United States obtain 0.2-0.9 years of additional schooling. Again, much of this may be a result of remittances. Isolating the effect of parents’ absence from remittances, Cortes (2015[60]) finds that absence of mothers implies greater risks for the school performance of children left behind in the Philippines than absence of fathers. Analysing internal migration in China, Zhao et al. (2018[61]) similarly find negative effects of parents’ absence on children’s cognitive performance in a survey, especially in cases where mothers leave children behind in rural areas.

The frequently reported improvement in girls’ schooling is not necessarily linked to remittances but could also reflect a shift of family decision making from migrant fathers to mothers who stay behind. Using evidence on expenditures of households in Mexico, (Antman, 2011[62]) documents that households whose head is still in the United States spend a lower share of their resources on boys relative to girls than household whose head has returned from the United States.

A number of recent studies have examined effects on the emotional well-being of those staying behind. Some studies report evidence of lower well-being that is likely linked to separation. Based on survey data of families staying behind in Mexico, Silver (2014[63]) finds increasing feelings of loneliness and depression, especially among women. Using the same data, Nobles et al. (2015[64]) likewise report that wives who are left behind with children exhibit higher levels of stress, measured as incidence of sadness, crying and sleep disorders. Graham et al. (2015[65]) report corresponding findings for wives who are left behind with children in Indonesia, the Philippines and Viet Nam.

Drawing on cross-country survey evidence from the Gallup World Poll, Ivlevs et al. (2019[66]) conclude that having household members abroad (which cannot be specified further) is associated with a higher general life satisfaction but also a greater risk of experiencing stress or depression. Remittances are found to further increase general life satisfaction, which hints that positive effects on well-being could reflect expectations of a better financial situation or greater opportunities for the family in the future.

Mazzucato et al. (2015[67]) can investigate the psychological well-being of children and youth left behind (ages 11-21), based on survey evidence from Ghana, Nigeria and Angola. For the latter two countries, they find that children with at least one parent abroad exhibit lower psychological well-being, and in all countries, a change of primary caregiver is associated with lower psychological well-being. For children left behind in Moldova, Gassmann et al. (2013[68]) do not find effects from migration as such but children in households with return migrants exhibit higher emotional well-being.

The role of parents in migrant integration outcomes

In many OECD countries, migrants may be joined by their parents through family reunification (see Table 3.3 in Chaloff and Poeschel (2017[1])). At the same time, almost nothing is known about the effects that parents’ presence has on migrants’ integration. One of the rare studies in this context finds that parents of migrants in Canada, especially mothers, help with housework and family obligations (VanderPlaat et al., (2012[23])). These findings align with qualitative evidence from the United States (Treas and Mazumdar, 2004[69]). Results from the survey by the IRCC (2014[11]) similarly point to the role of migrants’ parents for childcare, allowing migrants to work more hours. In addition, the survey indicates that migrants’ parents can help with obtaining educational qualifications. This section therefore investigates if the presence of parents affects integration outcomes of migrants, especially migrants with young children.

The same data sets are used as in the analysis of delays in the arrival of the spouse (Box 4.1) except the Canadian data, which do not include information on parents. The available data limit the analysis to a simpler approach: comparisons of households in which at least one parent of the migrant is currently present, to households without parents. These comparisons only estimate effects from parents’ presence in the same household. Figure 4.12 shows that migrants’ mothers are most likely to be present in the same household. In almost all European OECD countries, less than 3% of married migrants live with a parent in the same household. While migrants’ parents can also play an important role when they live in another household nearby, the data only record their presence in the same household.

The analyses in this section again only consider effects on married migrants. In a second step, the analysis focuses on the subset of households where young children are present. The data for European OECD countries record the presence of children under 15 in the household. Those for the United States record the presence of related children under 18. In many households with young children, migrants might face childcare obligations. In these cases, the effect of the presence of migrants’ parents might be different.

Figure 4.12. Presence of migrants’ parents in the same household, European OECD countries, 2013-17

Married migrants aged 15-64

Note: Information on the presence of parents in the same household is missing for married persons in Germany.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview.

 StatLink https://doi.org/10.1787/888933989266

By taking care of grandchildren, migrants’ parents can facilitate employment

From a simple comparison between migrant households with and without parents present, evidence suggests that migrants’ parents have a positive effect on the employment probability of migrants with young children (Figure 4.13). Employment rates of migrants with parents present are higher than without parents in all countries except the Czech Republic. The differences between the employment rates of the two groups of households is six percentage points on average but reaches 14 percentage points in the Netherlands.

Figure 4.13. Employment rates of migrants with children by presence of parents, 2013-17

Married migrants aged 15-64 who live with spouse and children under 15

Note: Countries are selected based on sample sizes and availability of information on the presence of parents. Figures for the United States refer to 2013 and 2015/2016 and to households with children under 18. Migrants in mixed couples and cases where the spouse is absent are not included. Information on the presence of parents in the same household is missing for married persons in Germany.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview and the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

 StatLink https://doi.org/10.1787/888933989285

Figure 4.14 reports results of all analyses for the United States and for European OECD countries. The results in Panel A are based on all married migrants, while those in Panel B are based on married migrants with children under 15 only. As these analyses can draw on a much larger number of observations than the analyses for principal migrants and spouses above, more results are obtained here. However, the results should still be treated with caution because they might reflect correlation rather than causation.5

The presence of migrants’ parents is associated with significantly higher employment probabilities, especially for migrant women (Figure 4.14, Panel A). In both European OECD countries and the United States, migrant women’s employment probability is six percentage points higher than for comparable migrant women whose parents are not present in the household. Migrants also appear to work more hours per week when parents are present, in both European OECD countries and the United States. Hours worked appear higher by between 1% and 4%, where the largest effects again arise for women. This aligns with the result, obtained for the United States, that the wages of migrant women in particular are higher when parents are present (by 3%). In the United States, the presence of migrants’ parents is also associated with a slightly lower probability (1-2 percentage points) of being proficient in the host-country language.

Figure 4.14. Estimated effects of presence of migrants’ parents on migrants’ integration outcomes, 2013-17

Note: All reported results are statistically significant at the 5% significance level. Migrants’ parents are considered present if at least one of the couple’s parents lives in the same household. Young children are defined as children under 15 in European OECD countries and children under 18 in the United States. Figures for the United States refer to 2013 and 2015/2016 and are limited to persons who were married at the time of arrival. Wages refer to gross wage levels in the United States and the probability of earning a net wage above the median wage in European OECD countries. Migrants in mixed couples and cases where the spouse is absent are not included.

Source: OECD Secretariat analyses based on the European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview, and the American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

 StatLink https://doi.org/10.1787/888933989304

Almost all of these results appear stronger for the subset of households with children under 15 (Figure 4.14, Panel B), especially the results for women. In these households, the presence of migrants’ parents is associated with an increase in migrant women’s employment probability by 7-8 percentage points, in their hours worked by 3%-6%, and in their wages by 5%. The finding that estimated effects are larger for households with young children suggests that migrants’ parents often help with childcare and other housework so that more migrants can take up employment and increase their hours worked. Various forms of parental support might also explain why higher employment probabilities, more hours worked and higher wages arise in Panel A.

Conclusions and policy implications

This chapter has investigated how delays in family reunification affect integration outcomes of principal migrants and later-arriving spouses or children. Except for effects from age at arrival on migrant children, the existing literature hardly offers any conclusive results on these questions. Empirical results are indeed difficult to obtain due to data limitations and serious methodological challenges. The approach taken in this chapter uses large-scale survey data sets in which both spouses in migrant couples can be observed. In order to separate causal effects of delays from spurious correlations, a range of other factors is accounted for and typically only long-term effects are considered.

The results obtained in this chapter suggest that delays in family reunification may have adverse consequences for migrant integration in the long-term. After ten years or more in the host country, principal migrants whose spouse joined them after some delay earn significantly lower wages than otherwise comparable principal migrants. Spouses who arrive with delay are themselves less likely to be proficient in the host-country language after five or more years of residence. Migrant children who arrived at pre-school age later appear better integrated than migrants who arrived in school age, especially in terms of language proficiency. As all analyses take duration of stay in the host-country into account, these results do not simply reflect a shorter duration of stay following delays or arrival at a later age.

Certain results, however, also point to certain potential benefits of delays in family reunification for employment probabilities. Principal migrants who initially spent a longer time in the country without their spouse are more likely to be employed after ten or more years in the host country. Spouses who arrive with delay in the United States appear slightly more likely to be employed after five years or more. However, these results are found only in specific contexts and do not hold across countries and gender.

The results have implications for migration policy: conditions imposed on family reunification can contribute to delays, as this chapter demonstrates for the case of Germany. In 2017, almost all OECD countries applied a condition based on principal migrants’ income in one way or another; 25 required demonstrating adequate housing; 14 required a minimum residence period for the principal migrant; and five applied pre-arrival language requirements for spouses (see Table A.1 in OECD (2017[30])). Many migrant families meet these conditions only after a number of years, so that family reunification is delayed. If the objective of the conditions and procedures is to ensure that spouses integrate well, then the evidence in this chapter does not support this expectation.

In addition, policy makers in OECD countries should carefully consider the role of family reunification for their countries’ ability to attract and retain high-skilled migrants. The conditions and procedures for family reunification as well as the conditions for labour market access of family members matter for a country’s attractiveness to high-skilled migrants with families (Tuccio, 2019[70]). Earlier analyses have shown that high-skilled migrants appear to stay substantially longer if their spouse also finds employment in the host country (OECD, 2016[71]). In a number of OECD countries, the conditions for family reunification are therefore more generous for high-skilled migrants than for other migrants, or they are waived altogether.

Comparatively strong results are obtained in this chapter for the effect of age at arrival on integration outcomes of migrants’ children, in line with existing literature. These results imply that avoiding long delays is especially important for the integration prospects of migrants’ children. Conditions and procedures for family migration can be designed in a way that encourages migrants to quickly bring their children to the host country (OECD, 2017[30]). Efforts to reduce delays in family reunification could also include systematically informing migrants of the possibilities for family reunification, ensuring that the associated conditions and procedures are transparent, and accelerating the procedures.

With regards to public debates on migration policy, the findings in this chapter suggest that statements about generally positive or generally negative effects of family presence are over-simplified. Family presence can play a positive role for some integration outcomes and a negative role for others. In many cases, the magnitude of these effects may be relatively small, and effects might differ substantially between groups of migrants and across host countries. Bearing this in mind, careful empirical studies should be undertaken to explore the role of family presence in each particular national context and for different admission categories of principal migrants.

References

[56] Alexander, R. and Z. Ward (2018), “Age at Arrival and Assimilation During the Age of Mass Migration”, The Journal of Economic History, Vol. 78/03, pp. 904-937, https://doi.org/10.1017/S0022050718000335.

[58] Antman, F. (2013), “The impact of migration on family left behind”, in Constant, A. and K. Zimmermann (eds.), International Handbook on the Economics of Migration, Edward Elgar, Cheltenham (UK) and Northampton, MA.

[62] Antman, F. (2011), “International Migration and Gender Discrimination among Children Left Behind”, American Economic Review: Papers & Proceedings, Vol. 101/3, pp. 645-649, https://doi.org/10.1257/aer.101.3.645.

[50] Aoki, Y. and L. Santiago (2018), “Speak better, do better? Education and health of migrants in the UK”, Labour Economics, Vol. 52, pp. 1-17, https://doi.org/10.1016/J.LABECO.2018.03.003.

[42] Åslund, O., A. Böhlmark and O. Skans (2015), “Childhood and family experiences and the social integration of young migrants”, Labour Economics, Vol. 35, pp. 135-144, https://doi.org/10.1016/J.LABECO.2015.05.004.

[6] Beaton, E., A. Musgrave and J. Liebl (2018), Safe but not settled: The impact of family separation on refugees in the UK, Refugee Council and Oxfam, https://www.refugeecouncil.org.uk/assets/0004/2515/Safe_but_not_settled.pdf.

[51] Beck, A., M. Corak and M. Tienda (2012), “Age at Immigration and the Adult Attainments of Child Migrants to the United States”, The ANNALS of the American Academy of Political and Social Science, Vol. 643/1, pp. 134-159, https://doi.org/10.1177/0002716212442665.

[53] Bleakley, H. and A. Chin (2010), “Age at Arrival, English Proficiency, and Social Assimilation Among US Immigrants”, American Economic Journal: Applied Economics, Vol. 2/1, pp. 165-192, https://doi.org/10.1257/app.2.1.165.

[41] Böhlmark, A. (2008), “Age at immigration and school performance: A siblings analysis using Swedish register data”, Labour Economics, Vol. 15/6, pp. 1366-1387, https://doi.org/10.1016/j.labeco.2007.12.004.

[3] Bonjour, S. and A. Kraler (2015), “Introduction: Family Migration as an Integration Issue? Policy Perspectives and Academic Insights”, Journal of Family Issues, Vol. 36/11, pp. 1407-1432, https://doi.org/10.1177/0192513X14557490.

[21] Boyd, M. (1989), “Family and Personal Networks in International Migration: Recent Developments and New Agendas”, International Migration Review, Vol. 23/3, p. 638, https://doi.org/10.2307/2546433.

[19] Bragg, B. and L. Wong (2016), ““Cancelled Dreams”: Family Reunification and Shifting Canadian Immigration Policy”, Journal of Immigrant & Refugee Studies, Vol. 14/1, pp. 46-65, https://doi.org/10.1080/15562948.2015.1011364.

[38] Bratsberg, B. and O. Raaum (2010), “Effekter av krav om forsørgelsesevne ved familiegjenforening”, No. 4/2010, Ragnar Frisch Centre for Economic Research, https://www.frisch.uio.no/publikasjoner/pdf/rapp10_04.pdf.

[26] Brenzel, H. and Y. Kosyakova (2019), “Längere Asylverfahren verzögern Integration und Spracherwerb”, IAB-Kurzbericht, No. 6/2019, http://doku.iab.de/kurzber/2019/kb0619.pdf.

[10] Caplan, S. (2007), “Latinos, Acculturation, and Acculturative Stress: A Dimensional Concept Analysis”, Policy, Politics, & Nursing Practice, Vol. 8/2, pp. 93-106, https://doi.org/10.1177/1527154407301751.

[1] Chaloff, J. and F. Poeschel (2017), “A portrait of family migration in OECD countries”, in International Migration Outlook 2017, OECD Publishing, Paris, https://dx.doi.org/10.1787/migr_outlook-2017-6-en.

[4] Charsley, K., M. Bolognani and S. Spencer (2017), “Marriage migration and integration: Interrogating assumptions in academic and policy debates”, Ethnicities, Vol. 17/4, pp. 469-490, https://doi.org/10.1177/1468796816677329.

[47] Chiswick, B. and N. DebBurman (2004), “Educational attainment: analysis by immigrant generation”, Economics of Education Review, Vol. 23/4, pp. 361-379, https://doi.org/10.1016/J.ECONEDUREV.2003.09.002.

[33] Chiswick, B. and P. Miller (2002), “Immigrant earnings: Language skills, linguistic concentrations and the business cycle”, Journal of Population Economics, Vol. 15/1, pp. 31-57, https://doi.org/10.1007/PL00003838.

[57] Clarke, A. (2016), “Age at Immigration and the Educational Attainment of Foreign-born Children in the United States: The Confounding Effects of Parental Education”, International Migration Review, https://doi.org/10.1111/imre.12294.

[29] Cobb-Clark, D. and T. Crossley (2004), “Revisiting the family investment hypothesis”, Labour Economics, Vol. 11/3, pp. 373-393, https://doi.org/10.1016/J.LABECO.2003.05.002.

[52] Cohen Goldner, S. and G. Epstein (2014), “Age at immigration and high school dropouts”, IZA Journal of Migration, Vol. 3/1, https://doi.org/10.1186/s40176-014-0019-2.

[60] Cortes, P. (2015), “The Feminization of International Migration and its Effects on the Children Left Behind: Evidence from the Philippines”, World Development, Vol. 65, pp. 62-78, https://doi.org/10.1016/J.WORLDDEV.2013.10.021.

[22] Creese, G., I. Dyck and A. McLaren (2008), “The ‘Flexible’ Immigrant? Human Capital Discourse, the Family Household and Labour Market Strategies”, Journal of International Migration and Integration / Revue de l’integration et de la migration internationale, Vol. 9/3, pp. 269-288, https://doi.org/10.1007/s12134-008-0061-0.

[35] Danzer, A. and F. Yaman (2016), “Ethnic concentration and language fluency of immigrants: Evidence from the guest-worker placement in Germany”, Journal of Economic Behavior & Organization, Vol. 131, pp. 151-165, https://doi.org/10.1016/J.JEBO.2016.08.012.

[15] Facchini, G., E. Patacchini and M. Steinhardt (2015), “Migration, Friendship Ties, and Cultural Assimilation”, The Scandinavian Journal of Economics, Vol. 117/2, pp. 619-649, https://doi.org/10.1111/sjoe.12096.

[68] Gassmann, F. et al. (2013), “The impact of migration on children left behind in Moldova”, UNU-MERIT Working Papers, http://collections.unu.edu/view/UNU:87.

[17] Ginsborg, P. (1995), “Italian Political Culture in Historical Perspective”, Modern Italy, Vol. 1/01, pp. 3-17, https://doi.org/10.1080/13532949508454754.

[46] Gonzalez, A. (2003), “The education and wages of immigrant children: the impact of age at arrival”, Economics of Education Review, Vol. 22/2, pp. 203-212, https://doi.org/10.1016/S0272-7757(02)00004-3.

[12] Gracia, E. and J. Herrero (2004), “Determinants of social integration in the community: an exploratory analysis of personal, interpersonal and situational variables”, Journal of Community & Applied Social Psychology, Vol. 14/1, pp. 1-15, https://doi.org/10.1002/casp.746.

[65] Graham, E., L. Jordan and B. Yeoh (2015), “Parental migration and the mental health of those who stay behind to care for children in South-East Asia”, Social Science & Medicine, Vol. 132, pp. 225-235, https://doi.org/10.1016/J.SOCSCIMED.2014.10.060.

[55] Guven, C. and A. Islam (2015), “Age at Migration, Language Proficiency, and Socioeconomic Outcomes: Evidence From Australia”, Demography, Vol. 52/2, pp. 513-542, https://doi.org/10.1007/s13524-015-0373-6.

[24] Hainmueller, J., D. Hangartner and D. Lawrence (2016), “When lives are put on hold: Lengthy asylum processes decrease employment among refugees”, Science Advances, Vol. 2/8, p. e1600432, https://doi.org/10.1126/sciadv.1600432.

[59] Hanson, G. and C. Woodruff (2003), “Emigration and Educational Attainment in Mexico”, http://www.childmigration.net/files/Hanson_2003.pdf (accessed on 29 January 2019).

[43] Hermansen, A. (2017), “Age at Arrival and Life Chances Among Childhood Immigrants”, Demography, Vol. 54/1, pp. 201-229, https://doi.org/10.1007/s13524-016-0535-1.

[11] IRCC - Immigration, Réfugiés et Citoyenneté Canada (2014), Evaluation of the Family Reunification Program, Research and Evaluation Branch, https://www.canada.ca/content/dam/ircc/migration/ircc/english/pdf/pub/e4-2013-frp.pdf.

[66] Ivlevs, A., M. Nikolova and C. Graham (2019), “Emigration, remittances, and the subjective well-being of those staying behind”, Journal of Population Economics, Vol. 32/1, pp. 113-151, https://doi.org/10.1007/s00148-018-0718-8.

[20] Kaida, L. (2015), “Ethnic Variations in Immigrant Poverty Exit and Female Employment: The Missing Link”, Demography, Vol. 52/2, pp. 485-511, https://doi.org/10.1007/s13524-015-0371-8.

[16] Kim, Y. and W. Wilcox (2013), “Bonding alone: Familism, religion, and secular civic participation”, Social Science Research, Vol. 42/1, pp. 31-45, https://doi.org/10.1016/J.SSRESEARCH.2012.08.001.

[49] Lee, S. and B. Edmonston (2011), “Age-at-Arrival’s Effects on Asian Immigrants’ Socioeconomic Outcomes in Canada and the U.S.”, International Migration Review, Vol. 45/3, pp. 527-561, https://doi.org/10.1111/j.1747-7379.2011.00857.x.

[44] Lemmermann, D. and R. Riphahn (2018), “The causal effect of age at migration on youth educational attainment”, Economics of Education Review, Vol. 63, pp. 78-99, https://doi.org/10.1016/J.ECONEDUREV.2017.11.001.

[36] Liebig, T. and K. Tronstad (2018), “Triple Disadvantage? : A first overview of the integration of refugee women”, OECD Social, Employment and Migration Working Papers, No. 216, OECD Publishing, Paris, https://dx.doi.org/10.1787/3f3a9612-en.

[37] Lochmann, A., H. Rapoport and B. Speciale (2018), “The Effect of Language Training on Immigrants’ Economic Integration: Empirical Evidence from France”, IZA Discussion Paper, No. 11331, http://ftp.iza.org/dp11331.pdf (accessed on 7 January 2019).

[28] Long, J. (1990), “The Effect of Americanization on Earnings: Some Evidence for Women”, Journal of Political Economy, Vol. 88/3, pp. 620-629, https://doi.org/10.1086/260892.

[25] Marbach, M., J. Hainmueller and D. Hangartner (2018), “The long-term impact of employment bans on the economic integration of refugees”, Science Advances, Vol. 4/9, p. eaap9519, https://doi.org/10.1126/sciadv.aap9519.

[67] Mazzucato, V. et al. (2015), “International parental migration and the psychological well-being of children in Ghana, Nigeria, and Angola”, Social Science & Medicine, Vol. 132, pp. 215-224, https://doi.org/10.1016/J.SOCSCIMED.2014.10.058.

[5] Miller, A. et al. (2018), “Understanding the mental health consequences of family separation for refugees: Implications for policy and practice”, American Journal of Orthopsychiatry, Vol. 88/1, pp. 26-37, https://doi.org/10.1037/ort0000272.

[9] Mlati, F. and C. Duarte (2005), “L’unité familiale, enjeux et impacts sur l’insertion professionnelle des réfugiés statutaires”, in Le regroupement familial des réfugiés, France Terre d’Asile, Paris, http://www.france-terre-asile.org/images/stories/publications/pdf/CduS_06-vweb-finale.pdf.

[54] Mukhopadhyay, S. (2018), “Language Assimilation and Performance in Achievement Tests among Immigrant Children: Evidence from a Field Experiment”, IZA Discussion Paper, No. 11524, http://ftp.iza.org/dp11524.pdf.

[13] Muñoz-Laboy, M., J. Hirsch and A. Quispe-Lazaro (2009), “Loneliness as a Sexual Risk Factor for Male Mexican Migrant Workers”, American Journal of Public Health, Vol. 99/5, pp. 802-810, https://doi.org/10.2105/AJPH.2007.122283.

[48] Myers, D., X. Gao and A. Emeka (2009), “The Gradient of Immigrant Age-at-Arrival Effects on Socioeconomic Outcomes in the U.S.”, International Migration Review, Vol. 43/1, pp. 205-229, https://doi.org/10.1111/j.1747-7379.2008.01153.x.

[18] Nannestad, P., G. Lind Haase Svendsen and G. Tinggaard Svendsen (2008), “Bridge Over Troubled Water? Migration and Social Capital”, Journal of Ethnic and Migration Studies, Vol. 34/4, pp. 607-631, https://doi.org/10.1080/13691830801961621.

[64] Nobles, J., L. Rubalcava and G. Teruel (2015), “After spouses depart: Emotional wellbeing among nonmigrant Mexican mothers”, Social Science & Medicine, Vol. 132, pp. 236-244, https://doi.org/10.1016/J.SOCSCIMED.2014.11.009.

[40] OECD (2018), The Resilience of Students with an Immigrant Background: Factors that Shape Well-being, OECD Reviews of Migrant Education, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264292093-en.

[30] OECD (2017), Making Integration Work: Family Migrants, Making Integration Work, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264279520-en.

[71] OECD (2016), “Attracting and retaining skilled migrants and international students in the Netherlands”, in Recruiting Immigrant Workers: The Netherlands 2016, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264259249-9-en.

[27] OECD (2016), “Labour market outcomes of recent migrants and integration policies in OECD countries”, in International Migration Outlook 2016, OECD Publishing, Paris, https://dx.doi.org/10.1787/migr_outlook-2016-5-en.

[14] Olwig, K. (2011), “‘Integration’: Migrants and Refugees between Scandinavian Welfare Societies and Family Relations”, Journal of Ethnic and Migration Studies, Vol. 37/2, pp. 179-196, https://doi.org/10.1080/1369183X.2010.521327.

[31] Rendall, M. et al. (2011), “The Protective Effect of Marriage for Survival: A Review and Update”, Demography, Vol. 48/2, pp. 481-506, https://doi.org/10.1007/s13524-011-0032-5.

[32] Robards, J. et al. (2012), “Marital status, health and mortality”, Maturitas, Vol. 73/4, pp. 295-299, https://doi.org/10.1016/J.MATURITAS.2012.08.007.

[7] Rousseau, C., A. Mekki-Berrada and S. Moreau (2001), “Trauma and Extended Separation from Family among Latin American and African Refugees in Montreal”, Psychiatry: Interpersonal and Biological Processes, Vol. 64/1, pp. 40-59, https://doi.org/10.1521/psyc.64.1.40.18238.

[34] Saleh AlHammadi, F. (2016), “Psycholinguistic determinants of immigrant second language acquisition”, https://doi.org/10.1016/j.lingua.2016.03.001.

[45] Schaafsma, J. and A. Sweetman (2001), “Immigrant earnings: age at immigration matters”, Canadian Journal of Economics/Revue Canadienne d`Economique, Vol. 34/4, pp. 1066-1099, https://doi.org/10.1111/0008-4085.00113.

[39] Schmidt, G. et al. (2009), Ændrede familiesammenføringsregler - Hvad har de nye regler betydet for pardannelsesmønstret blandt etniske minoriteter?, SFI, Kopenhagen, https://pure.vive.dk/ws/files/257874/0928_familiesammenfoering.pdf.

[63] Silver, A. (2014), “Families Across Borders: The Emotional Impacts of Migration on Origin Families”, International Migration, Vol. 52/3, pp. 194-220, https://doi.org/10.1111/j.1468-2435.2010.00672.x.

[2] Tate, E. (2011), “Family Separation and Reunification of Newcomers in Toronto: What does the literature say?”, https://www.semanticscholar.org/paper/Family-Separation-and-Reunification-of-Newcomers-in/c64ddad31a370ddbd9a6fa25ce2fe7056f4380e7.

[69] Treas, J. and S. Mazumdar (2004), “Kinkeeping and Caregiving: Contributions of Older People in Immigrant Families”, Journal of Comparative Family Studies, Vol. 35, pp. 105-122, https://doi.org/10.2307/41603919.

[70] Tuccio, M. (2019), “Measuring and assessing talent attractiveness in OECD countries”, OECD Social, Employment and Migration Working Papers, No. 229, OECD Publishing, Paris, https://dx.doi.org/10.1787/b4e677ca-en.

[23] VanderPlaat, M., H. Ramos and Y. Yoshida (2012), “What do Sponsored Parents and Grandparents Contribute?”, Canadian Ethnic Studies, Vol. 44/3, pp. 79-96, https://doi.org/10.1353/ces.2013.0006.

[8] Walther, L. et al. (2019), “Living conditions and the mental health and well-being of refugees: Evidence from a large-scale German panel study”, SOEP Papers on Multidisciplinary Panel Data Research, No. 1029, DIW, https://www.econstor.eu/handle/10419/195293.

[61] Zhao, Q. et al. (2018), “China s Migrant and Left-behind Children: Correlation of Parental Migration on Health, Cognitive and Non-cognitive Outcomes”, https://ageconsearch.umn.edu/record/277164/files/1174.pdf.

Database references

American Community Survey (U.S. Census Bureau), https://www.census.gov/programs-surveys/acs/about.html.

Canadian Census (Statistics Canada), https://www12.statcan.gc.ca/census-recensement/2016/dp-pd/index-eng.cfm.

Canadian Labour Force Survey (Statistics Canada), https://www150.statcan.gc.ca/n1/en/surveys/3701.

European Labour Force Survey (Eurostat), http://ec.europa.eu/eurostat/web/lfs/overview

European Labour Force Survey (Eurostat) ad-hoc modules 2008 and 2014 on the labour market situation of migrants and their immediate descendants, http://ec.europa.eu/eurostat/statistics-explained/index.php/EU_labour_force_survey_-_ad_hoc_modules.

Gallup World Poll, https://www.gallup.com/analytics/232838/world-poll.aspx.

German Socio-Economic Panel (DIW Berlin), https://www.diw.de/en/soep.

Notes

← 1. The difference might arise because the data for European OECD countries cannot be limited to migrants who were married at the time of arrival in the host country. It should also be noted that results in Figure 4.3 are not directly comparable with results in Panel A of Figure 4.2: migrants whose spouse is absent or arrived with more than 9 years delay are not included in Figure 4.3.

← 2. While the correlation between the spouse’s delay and disability status is limited, (correlation coefficient of 0.12), it does not appear to be a weak instrument: it is highly significant at the first stage of the estimation (with an F-test statistic of 90) and passes tests designed to identify weak instrumental variables. The results are based on a substantial number of observations as almost 8% of the spouses identified in the American Community Survey indicate having a disability.

← 3. Another reason for the difference could be bias, which often arises in instrumental variable estimation. However, various diagnostics suggest that disability status performs well as an instrumental variable (see the previous note).

← 4. Most reported results arise similarly also for durations of stay of at least 10 years, but with somewhat less statistical reliability.

← 5. Since migrants’ spouses are present in all considered households and their characteristics are included, the results should not be driven by the presence of the spouse.