Fertility Adaptation of Child Migrants to Canada

Abstract

This study analyzes the fertility behavior of immigrant women arriving to Canada before age 19 using the 20 per cent sample of the Canadian Census from 1991 through 2006. Findings show that fertility increases with age at immigration, and is particularly high for those immigrating in their late teens. This pattern prevails regardless of the country of origin or whether the mother tongue of the migrant is an official language in Canada or not. We do not find a ‘critical age’ at which the behavior of migrants with and without official mother tongue start to diverge by more, even though the fertility of migrants without official mother tongue is always higher on average. Formal education matters as the fertility of immigrants who arrived to Canada before adulthood and graduated from college is similar to that of their native peers regardless of their age of arrival. However, the fertility of those with less than tertiary education increasingly diverges with age at migration from similarly educated Canadians.

1. Introduction and Background

In recent years, immigrant flows to Western OECD countries have increased considerably. With rising dependency ratios due to low fertility and the aging of the baby boom generation, immigration is an important factor for sustaining current levels of public services and maintaining population growth (Belanger 2003, 2005). In this context, a significant body of research has examined whether there are differences in fertility between immigrants and the native born and whether these tend to disappear over the medium to long term. (U.N. 2000; Sobotka, 2008). Moreover, since childbearing decisions are related to education and labor market participation choices, the analysis of immigrant fertility differentials sheds light on the socioeconomic integration of immigrants, particularly women. . Although a relative high fertility rate among first generation immigrants may help sustain population growth, it can also hinder the economic success of female immigrants and negatively impact the economic wellbeing of immigrant families and the human capital investments on their children.

Canada, with one the largest share of immigrants across Western nations, – around 20 per cent of the population was born abroad in the 2006 Census -- is an interesting case to study immigrant fertility. Unlike in other major immigrant recipient countries (except recently Australia), the Canadian government has developed policies that target educated immigrants and reward knowledge in local languages. Immigrants to Canada are admitted through three main categories (skilled, family and refugees) and skilled immigrants currently account for around 60 per cent of all new admissions. This produces an average immigrant profile that is somewhat relatively more educated and closer to natives than that of other countries, such as the US. These policy differences are prone to have important consequences for assimilation in different dimensions, including fertility.

The analysis of migrant/native fertility differences requires an understanding of the forces that shape the fertility choices of migrants. In their seminal paper, Goldstein and Goldstein (1981) identify three important mechanisms to explain the differential fertility behavior of migrants: selection, disruption, and adaptation. The selection hypothesis posits that individuals who migrate differ systematically from non-migrants in their countries of origin, and this selectivity may explain their subsequent fertility patterns (Kahn 1988; Forste and Tienda 1996; Sobotka 2008). The second mechanism draws attention to the (actual and anticipated) short-term disruption of fertility at the time of migration (Goldstein and Goldstein 1981; Ram and George 1990; Stephen and Bean 1992; Blau 1992; Kahn 1994; Ng and Nault 1997; Toulemon 2004). Finally the adaptation hypothesis posits that as migrants settle in their new environment their fertility norms and expectations begin to resemble those of the native population (Goldstein and Goldstein 1981; Stephen and Bean 1992; Alba and Nee 1997). Their interaction with Canadians and the circumstances they face in Canada, both in terms of opportunities and costs, ultimately determine their fertility decisions. Immigrant women facing better labor market prospects in Canada than in their countries of origin may decide to reduce/postpone fertility in order to work (Galor and Weil 1996). Absence of informal child care provided by relatives and the need to resort to more expensive forms of formal daycare increases the incentives to trade off children for work (see Carter 2000 for Mexicans in the US) and to devote more resources to the rearing of each child rather than increasing their offspring (Becker 1981).

All three mechanisms are likely to shape the fertility of women who migrate as adults, and the independent role of each can be difficult to isolate and assess. Some of these forces can coexist: it may be possible to observe an initial drop in fertility because of disruption at the time of immigration, followed by a subsequent rise in fertility, but also a gradual adaptation towards the fertility levels of the host country. These hypotheses have been tested in the literature, with none of them being conclusively accepted or rejected (Kulu 2005).

Adaptation and the role of age at migration

Research so far has focused primarily on the fertility outcomes of adult migrants. The study discussed in this paper analyzed the fertility of immigrant women who arrived in Canada before the age of 19 relative to that of the native born using data from four Canadian Censuses from 1991 to 2006. For individuals who migrated as children, the link between migration and subsequent behavior should be less directly affected by any short-term, disruptive effects of the migration experience and less biased by selection and simultaneity bias at the individual level. Child immigrants are likely to arrive with their parents and, as a result, their ‘decision’ to immigrate is plausibly independent of other variables affecting fertility. Thus, by studying child migrants we are able to focus in understanding the adaptation process in the receiving country and the role that age at migration plays on it. Related economics literature has also recognized age at immigration as a decisive factor to explain the differential process of adaptation of immigrants in many socio-economic dimensions (Chiswick 1991; Picot et al. 2005, 2007; Ferrer, Green and Riddell 2006).

Another reason to focus on women who migrated as children and their age at immigration is that there might be critical ages at which individuals learn a particular behavior or skill, such as the local language, that are crucial for future social and economic outcomes (Schaafsma and Sweetman 2001; Bleakley and Chin 2010). Fluency in the language of the destination country has long been recognized to play a key role in immigrant’s outcomes and degree of adaptation (see e.g. Kossoudji 1988; Dustmann 1994; Schaafsma and Sweetman 2001; Chiswick and Miller 2001, 2004; Dustman and van Soest 2002; Bleakley and Chin, 2004). Studies by Leslie and Lindley (2001) and Dustmann and Fabbri (2003), for example, find that lack of fluency in English has a detrimental impact on the employment and earnings of ethnic minority men and women in Britain. In the case of fertility, a non-official mother tongue may impact the ability of the child-migrant to access local cultural cues through school and peers to form her fertility preferences. Previous analyses have found greater English fluency to be associated with lower fertility in the US (Sorenson 1988; Swicegood et al. 1988; Bleakley and Chin 2010). It is well documented that fluency in the language of the destination country decreases with age at immigration among other things (Chiswick 1991; Stevens, 1992, 1999; Espenshade and Fu 1997; Espinosa and Massey 1997; Akresh 2007) and should be greater for those migrants who arrive as children. In their paper, Bleakley and Chin (2010) show that the outcomes of immigrants from non-English speaking countries systematically differ from those of other migrants only among those arriving after the ‘critical period’ of 8 to 9 years of age. In the case of fertility, there may be an additional mechanism for age at arrival to matter if cultural norms regarding reproductive behavior adopted at a particular age (for instance, the onset of puberty) are difficult to modify later on.

Finally, the extent of adaptation may hinge on the distance between the norms governing fertility in both the destination and source country. Those norms or attitudes are part of an individual’s cultural background. Accounting for heterogeneity of origins in the migrant population has been proven relevant in explaining variation in fertility outcomes in different contexts (Kahn 1994; Andersson, 2004; Coleman 2006; Guinnane et al. 2006; Parrado and Morgan 2008; Georgiadis and Manning, 2011, among others). The majority of these studies report substantial differences in fertility by ethnicity, although these tend to shrink (or disappear) among the second generation of migrants Two studies by Georgiadis and Manning (2011) and Coleman and Dubuc (2010), for example, explore the relatively slower fertility assimilation of Pakistanis and Bangladeshis in Britain as compared to that of Chinese, Indians and Black Caribbean whose fertility is at or below the UK national average. In Canada, fertility studies show that up to 1980 Canadian immigrants had lower fertility rates than the Canadian born (Kalbach 1970), but the trend has since reversed. A study by Belanger and Gilbert (2003) suggests that the increase in the share of Canadian immigrants from areas with traditionally high fertility rates such as the Middle East, Southern Asia and Latin America is likely responsible for this change in fertility patterns. The literature findings underscore the importance of taking into account the heterogeneity of the foreign-born population and analyzing whether, once indicators for place of origin are included in the empirical analysis, adaptation occurs in a similar way for individuals moving from different places.

The study presented in this paper took advantage of the large sample sizes in the Canadian Census to explore, first, how the fertility adaptation of child migrants is explained by differences in age at migration. If there are critical ages at which integration becomes more difficult, we should find sharp discontinuities in fertility with age at immigration. Further, in keeping with the literature, the analysis looked at whether immigrants from different areas showed different fertility adaptation processes; that is, whether the relation between age at migration and relative fertility varied by place of birth. Third, the study explored whether language is an intervening variable in the process of adaptation of child migrants. In particular, whether women who do not have an official mother tongue and who arrive after a certain critical age, experience difficulties in language acquisition that impair their assimilation of cultural norms and, as a result, exhibit fertility patterns that diverge sharply from those of other migrants. If language fluency is a major channel through which age at immigration affects fertility, the fertility of immigrants whose mother tongue is either English or French should be similar to that of Canadian born regardless of age at arrival. Moreover, very young immigrants, arriving before the age of five (or even in early elementary school), should behave similarly to Canadian-born regardless of whether their mother tongue is an official one or not, as they are unlikely to encounter language barriers in their adaptation to local fertility behaviour. Fourth, the analysis turned to education as another plausible channel that potentially explains the extent of a child immigrant’s fertility adaptation. Although the study introduced indicators for educational attainment in the initial specifications, that model assumed that the influence of education on fertility was the same regardless of the age at immigration. However, an early age at immigration affords better integration into the school system and more exposure to the norms of the receiving country and hence it is likely to influence educational attainment and, as a result, fertility. If age at immigration is related to fertility mostly (or in part) through its impact on educational attainment; the fertility behavior of an immigrant should mimic that of similarly educated native born Canadians regardless of age at migration.

The findings of the study show that fertility of migrants relative to natives rises continuously with age at immigration, with a steeper increase for those arriving in their late teens, even after indicators of educational attainment and region of birth are included in the model. This increasing profile is similar when estimated separately for immigrants with different cultural backgrounds, although the fertility levels vary by country of origin. Results suggest that the relationship between age at immigration and fertility is mostly unrelated to the individual’s proficiency of an official language -contrary to what previous literature has found in other margins of immigrant assimilation (Bleakely and Chin 2010), even though migrants without official mother tongue have higher fertility than other migrants, net of any region of origin differences. Finally, the profile differs by educational achievement and age at migration is particularly salient for those without tertiary education.

Section 2 describes the data and empirical strategy used in the study. Section 3 presents the results of the study on childbearing patterns of child migrants and explores different channels through which age at migration could be related to fertility behavior. The final section concludes with a general discussion of the findings.

2. Data and Empirical Approach

The study presented in this paper focused on the total number of children born to women in Canada aged 16 to 45, conditional on their migration status as well as on a set of additional independent variables. Because the dependent variable is a discrete count variable, we estimated it using a Poisson model that took the form:

$$lnE(F_i)={\sum }_{j=0.5}^{18+}({\beta }_j{A}_{ji})+\alpha X_{ii}$$ (1)

Where F_i is the fertility measure of woman i; A_ji is a set of indicator variables, one for each age at immigration from under age 1 to age 18, which constitutes the focus of our research, as well as one indicator for those arriving after age 18; X is a vector of demographic variables that varies somewhat across specifications. As a result Canadian-born women are the reference group. The estimated Poisson models always included a rural indicator and dummies for the province of residence; and five-year group indicators for the current age of the woman to allow for differences in fertility across the life cycle. In addition they contained five-year group indicators for birth cohorts to account for the fact that women born at different times might have different attitudes toward fertility and face different constraints regarding access to family planning, or government services, among others. Finally indicators for census years were also added, to account for specific survey year effects. Since we observe respondents of different ages, we controlled for their exposure time to fertility (defined as age minus 15 years) in the models. We decided not to include income measures, first, because it is not realistic to regard them as exogenous from fertility choices and, second, because of lack of longitudinal information on income preceding births. Including education in the analysis (see below) to some extent helps us to control for potential income. Although we do not report estimates for all control variables in the paper, they show consistent results across all specifications and are available upon request.

A first set of Poisson models encompassed, first, the basic covariates in equation (1) and then included both indicators for highest educational attainment (less than high school, high school, non-university post-secondary education and university education) and place of birth sequentially. Inclusion of these indicators produced estimates of the effect of age at immigration on fertility that are net of the educational or regional composition of the sample. However, the underlying assumption is that education and place of birth have a uniform effect on the fertility of immigrants regardless their age at immigration. Next, to study whether the fertility adaptation profile varied across regions of origin, a set of interactions of the age at migration indicators with place of birth was added to the basic model.

The last two models in the study explored whether the two channels discussed in the previous section, language fluency and schooling, play a role mediate the process of adaptation of child migrants. The Canadian census asks whether the individual speaks one of the official languages, but does not provide any measure of fluency. Because the ability to speak an official language is one of the criteria the Canadian government uses in the selection of immigrants the variable is mostly uninformative. As a result the Canadian immigration literature has typically relied on mother tongue, the language that individuals first understood and can still understand, as a proxy for fluency. The advantage of mother tongue relative to self-reported fluency is that it is a fairly exogenous variable. Further, the distance between mother tongue and the official language in the destination country is generally related to labor market and education outcomes as well as to migration flows (Adsera and Chiswick 2007; Adsera and Pytlikova 2012). We created the indicator ‘official mother tongue’ (OMT) to control for whether a woman’s mother tongue is the prevalent official language in her province of residence. Although Canada recognizes two official languages, French is used mostly in Quebec and English elsewhere. Hence, the OMT indicator was interacted with the age at immigration indicators to estimate the fertility profiles of immigrants with and without an OMT to compare them with the fertility of Canadian born, both those with an OMT and those without it.

The last model in the study added to the basic equation (1) interactions of an indicator of university degree with the age at immigration indicators. These produced estimates of the fertility profiles by age at migration of immigrants with and without university education separately and compared them with similarly educated Canadian-born women.

Data

The data employed in the study come from the confidential files of the Canadian Census of Population (20 per cent sample) for the years 1991, 1996, 2001, and 2006. Confidential census data provide large samples that allow multiple robustness analysis of the estimates. In addition they include more detailed information on individuals, as well as a richer categorization of relationships among members of the household than in the generally available data. With this detailed information, we were able to compute the number of children of each woman living in a household. From each census year, we selected all non-aboriginal women between 16 and 45 years of age and gathered information on age, education, marital status, number of live-born children (available in the 1991 Census), number of children living in the household, province of residence and immigrant status. In addition, for immigrant women we collected their year of immigration, age at immigration and country of birth. We grouped individuals by country of origin into 20 relatively homogenous groups (for a detailed listing see Adsera and Ferrer, 2011). Age at immigration refers to the age at which the respondent became a permanent resident. Most child immigrants enter the country as permanent residents, so this should not affect our sample greatly. To reduce computing time to a reasonable length, for each census year we selected all immigrants plus a 20 per cent random sample of domestic born individuals and weighted observations accordingly. The four censuses were then pooled together, resulting in approximately 1,800,000 observations.

In general, vital statistics are the most accurate source of information on fertility records in developed countries. However, since they contain little additional information about each individual and the household, they are inadequate for an in-depth analysis of fertility behavior. An alternative method is to indirectly estimate fertility from survey information such as Census data, which typically report the number of children living in the household. The method, known as the “own children method”, exploits the fact that the vast majority of young children live with their mother at the time of the census. Since the date of birth of both mother and children is known, it is possible to reconstruct each woman’s fertility history. Belanger and Gilbert (2003) show that estimated fertility differentials for immigrants and domestic born individuals for the period 1996–2001 using both methods are not very sizeable – with a downward bias of the census for women younger than 30 and an upward bias for those aged beyond 30.

Following this line of research, we used the number of children living in the household as our measure of fertility in the analysis. To the extent that some children may not live with their mothers, our dependent variable may be measured with some error. To reduce this problem, we restricted our sample to relatively young women (up to 45 years of age) whose children are more likely to live at home. Still, there are several caveats to the measure. First, the census questionnaire asks respondents to include children in joint custody who live most of the time in a household as household members. Therefore, our sample excludes all the children who are living only with their father. To the extent that young children are far more likely to live with their mothers, even after marriage disruption, this is not too important for our analysis. Second, it may be difficult to properly capture the very early childbearing of older women in the sample as some of their children may have already left home. That should be a concern particularly if the departure of children from the household in their late teens or early adulthood (e.g., attending college far from home, earlier marriage or cohabitation) occurs at a differential rate between immigrants and the Canadian born.

To assess the importance of the potential bias introduced by our dependent variable, we undertook three types of robustness exercises. First, we used the information on the total number of children ever born to a woman, available in the 1991 census, to re-estimate the models and compare the results with those obtained for the 1991 Census using our fertility measure. Second, we re-estimated the models using a restricted sample of women up to age 40. This reduces the likelihood that some children have already left home but it misses late childbearing, which in turn may be differentially important among groups (e.g., according to education, country of origin, etc). In this regard, Vezina and Turcotte (2009), after comparing data from the Canada Census and from the General Social Survey, note that there is no appreciable bias in the characteristics (including immigrant status) of the fraction of women aged 40 to 44 who have a child aged five or over based on whether some of the children live with them or not. Third, we re-estimated the models restricting the children included in our fertility measure to those aged 18 and under. Differences in the results when using this definition of the dependent variable would indicate that the bias introduced by the own children method is important. The overall pattern of the results and the estimated coefficients were robust across these different samples and specifications. These are available upon request.

Table 1 presents sample statistics separately for Canadian born and for immigrants. The first two columns correspond to the sample means over the 1991, 1996, 2001, and 2006 censuses pooled together. In order to provide a sense of the temporal variation in the data, the table also includes statistics for 1991 and 2006 separately. Around 50 per cent of immigrants arrived in Canada as adults, past the age of upper secondary schooling. The peak age of entry is between 20 and 30 years of age. The fraction of recent immigrants over the whole pool has increased in the latest Censuses and more recent immigrants arrive at a slightly older age than in the past. Still a substantial proportion of migrants, around 50 per cent in the 1991 census and 45 per cent in the 2006 census arrived before adulthood and they are the focus of the study reported here. Further, immigrants are increasingly arriving from countries in Asia and Africa rather than from old traditional source countries in Europe. Consequently, the fraction of immigrants with a non-official mother tongue, or that of immigrants using a non-official language at home, has steadily increased over the period 1991–2006. The focus of Canadian immigration policies on skills have attracted highly educated immigrants. On average, 45% of immigrants have post-secondary (college or more) education, versus 37% of the Canadian born, a trend that has increased over the sample years. These characteristics of the immigrant population, which are likely to have an impact on fertility behaviour, are well documented in the Canadian literature of immigration (Ferrer and Riddell 2008).

Table 1.

Descriptive statistics of characteristics and control variables for the sample of Canadian born (CB) and immigrants (IMM) used in the study of fertility, Canadian Census 1991–2006

	1991–2006		1991		2006
	CB	IMM	CB	IMM	CB	IMM
Number of children	0.84	1.10	0.89	1.18	0.77	1.03
Age	30.41	32.93	30.08	32.91	30.31	33.04
Education
Less than HS	0.25	0.22	0.30	0.29	0.19	0.14
High school	0.28	0.26	0.30	0.28	0.27	0.24
Trades (less than 1 year)	0.09	0.08	0.09	0.08	0.11	0.08
Non-university post-secondary	0.19	0.16	0.17	0.15	0.19	0.15
University-BA	0.16	0.23	0.12	0.16	0.20	0.31
Graduates	0.02	0.06	0.02	0.04	0.03	0.08
Marital Status
Married (+ common law)	0.54	0.64	0.59	0.67	0.50	0.63
Mixed couples	0.12	0.20	0.11	0.23	0.12	0.18
Years since migration	--	13.39	--	14.61	--	12.92
Arrived 0 to 5 years ago		0.27		0.25		0.28
Arrived 6 to 10 years ago		0.20		0.14		0.20
Arrived 11 to 15 years ago		0.17		0.16		0.20
Arrived 16 to 20 years ago		0.14		0.19		0.14
Arrived more than 20 years ago		0.23		0.27		0.19
Age at Immigration	--	19.56	--	18.32	--	20.14
Between 0 and 5 years of age		0.13		0.16		0.12
Between 6 and 11 years of age		0.13		0.13		0.14
Between 12 and 16 years of age		0.12		0.11		0.12
Between 17 and 19 years of age		0.09		0.10		0.08
Between 20 and 45 years of age		0.53		0.50		0.55
Country of origin
US		0.05		0.06		0.04
Central and South America		0.15		0.15		0.14
Europe		0.30		0.41		0.23
Middle East		0.06		0.06		0.07
South Asia		0.12		0.11		0.15
Rest of Asia		0.26		0.18		0.29
Africa		0.06		0.05		0.08
Pacific		0.01		0.01		0.01
Non official Mother tongue	0.05	0.68	0.04	0.60	0.05	0.74
Non official Language at home	0.01	0.42	0.01	0.34	0.01	0.46
Observations	914,386	920,940	204,170	198,090	242,596	260,790

Source: Canadian Census of Population (20 per cent sample) for the years 1991, 1996, 2001, and 2006.

3. Results

This section presents the results from the Poisson regressions estimated in the study. Findings are shown graphically in the form of incident rate ratios (IRRs). These should be interpreted as fertility rates of a particular group of immigrant females relative to those of otherwise similar Canadian-born females in the reference group. Coefficient estimates and their p-values can be found in the electronic appendix.

Fertility and Age at Immigration

As noted, the focus of the study are women who migrated to Canada before adulthood (18 years of age or younger) to better isolate fertility patterns that reflect adaptation to the rules and expectations that govern fertility behavior in the destination country. Figure 1 shows the fertility of immigrant women by age at migration relative to Canadian women separately for three specifications. A first baseline model with single age at migration indicators (up to age 18) as in equation (1) only included the basic controls. A second model added indicators of educational attainment because education both reflects successful adaptation to the host culture and influences fertility outcomes. A third model encompassed indicators for place of birth to account for the fact that fertility is greatly influenced by norms or attitudes that are part of an individual’s cultural background (Kahn 1994; Andersson 2004; Guinnane et al. 2006; Parrado and Morgan 2008; Coleman and Dubuc 2010; Georgiadis and Manning 2011; among others). Immigrants from the US were the reference immigrant group for this regression. Since their country is relatively close to Canada in terms of language and institutions, it is plausible to assume that their fertility choices are quite similar to those of the native born. In this model we estimated relative fertility by the age of arrival net of the influence of cultural preferences for fertility at origin and of the effect of the changing composition of the source country of immigration over census years_[0] on the fertility level. This specification implicitly assumes that the effect of age at immigration is the same across places of birth. We explore alternative possibilities later.

Figure 1.

Incidence Rate Ratios (IRRs) from a Poisson regression of the number of children in the household of women who migrated as children by age at immigration relative to Canadian born women, Canadian Census 1991–2006.

Note: All models include controls for age, birth cohort, location of residence and census year. Additional controls are mentioned in the caption. In the model with place of birth, the US is the reference group shown in graph. Estimates of the Poisson regressions are available in Table 1A of online appendix.

Source: As for Table 1.

Holding current age and birth cohort constant, fertility generally increases with age at immigration. In the basic model, the fertility rate of individuals migrating up to age 6 is either somewhat lower or indistinguishable from that of natives. The average immigrant to Canada arriving between 7 and 11 years of age shows a differential of 3 to 4 per cent with respect to the native born. This differential increases to about 10 per cent for those arriving during early adolescence (ages 12 to 15), and for late adolescents rises sharply from 22 per cent for those who arrive at age 16 to 33 per cent for the 18 year olds. Adding education to the model changed fertility differences somewhat by slightly flattening the fertility profile of late teen arrivals. Once place of birth indicators were included, the estimated relative fertilities for most ages of arrival were significantly different from 1 (the baseline for native born). An interesting feature of the findings in Figure 1 is that while there is no evidence of a clear discontinuity by age at immigration in the fertility of immigrants, there are slight steps at years of arrival that roughly correspond to milestones in the schooling system (elementary school, junior high school and senior high school). Further, fertility increases more gradually with age at immigration until the early teens, and then proceeds at a faster pace during the late teens. This result does not seem to be driven by a higher probability of intermarriage among those arriving younger. In separate regressions, not shown here, we observed that immigrants in mixed marriages (married to native-born Canadians) showed similar increasing fertility with age at immigration. The observed peak in fertility for those arriving in the late teens could also be attributed in part to the phenomena of “child brides” present in some European countries. However, there is a general perception that this phenomenon is much less relevant in Canada, a country that receives fewer immigrants from cultures favoring arranged marriages and has a more selective immigration policy regime than most European countries (Adsera and Ferrer 2011).

A natural question to ask is whether observed differences in total fertility are driven in part by differences in the timing of births. Our focus on child migrants minimizes the concern of shifts in timing due to temporary disruption of fertility among immigrant women (Andersson 2004). However, if child migrants have their children much earlier in life than the Canadian born, estimates may portray a higher fertility among immigrants just because they are closer to completed fertility. Ideally the study should focus on women who have already completed fertility. However this exercise would be unfruitful with Census data since, as explained before, the Census does not include a complete fertility history of women and the use of the “own children method” would not be valid in older women. To get a closer measure of this the full model was re-estimated with a sample restricted to women aged between 35 and 45. Results in this restricted sample (available upon request) are very similar to the ones with the whole sample, with fertility rising with age at arrival and peaking for the late teen arrivals. This suggests that any difference in the timing of birth is not likely to be driving results in Figure 1.

Estimates that included place of birth are very similar to those without it, suggesting that results are not driven by compositional variation in the origin of migrants arriving at different ages. Nonetheless, in Table 2, the estimated IRRs by region of origin (relative to the reference group of US migrants) from this specification expose large differences in fertility levels among groups. The fertility rate at each age-at migration of immigrants from Mexico is 40 per cent higher than that of US migrants. Other immigrant groups with high fertility rates are those from Central America, Africa (except South Africa), the Middle East, and Southern Asia. Their age-at-immigration profiles are shifted upward in relation to that of US migrants with more exacerbated positive peaks. In contrast, the profiles are flatter and shift down in relation to the US born particularly for Eastern European immigrants and those from elsewhere in Asia. Chinese immigrants exhibit the lowest fertility rate of all groups, only 0.73 times that of the US born.

Table 2.

Incidence Rate Ratios (IRRs) of place of birth indicators from the Poisson regression of the number of children in the household in figure 1, Canadian Census 1991–2006.

	IRR	p-value
US	reference
Caribe	0.89	(0.00)
Mexico	1.40	(0.00)
Central America	1.25	(0.00)
South America	0.94	(0.00)
Europe	0.93	(0.00)
Eastern Europe	0.88	(0.00)
UK-Ireland	0.91	(0.00)
Southern Europe	1.09	(0.00)
Middle East	1.24	(0.00)
China	0.73	(0.00)
North East Asia	0.89	(0.00)
South Eastern Asia	0.81	(0.00)
Southern Asia	1.13	(0.00)
North Africa	1.20	(0.00)
Central Africa	1.20	(0.00)
West Africa	1.11	(0.00)
Southern Africa	0.94	(0.00)
Eastern Africa	1.00	0.88)
Pacific	0.97	(0.00)

Note: The model includes controls for age at migration, education, age, birth cohort, location of residence and census year. The US is the reference group. Estimates of the complete Poisson regression are available in Table 1A of online appendix.

Source: As for Table 1.

Leaving differences in levels by place of origin aside, we are interested in whether the fertility of each immigrant group follows the same pattern by age at migration as that of the average migrant To analyze whether child immigrants arriving from certain regions adapt to greater extent to the rules that guide fertility behavior in the host country than others with similar years of exposure, we included in the model a set of interaction dummies with place of birth and age at immigration. For ease of exposure, we categorized age at immigration into five groups corresponding roughly to different steps in the school system: those immigrating as pre-school children (aged five or less at the time of immigration), those immigrating as elementary school children (aged 6 to 11 at the time of immigration), those immigrating in their early teens (ages 12 to 15) who are likely to enter in junior high school and those immigrating in the late teens (between ages 16 to 18) who are still highly likely to receive some form of Canadian schooling, and finally those immigrating after 18 years of age. This categorization was intended to avoid mixing in the same group children entering Canada at different school levels, as this might affect the type of interactions they experience (Khmelkov and Hallinan 1999). We also grouped places of birth into broader areas (US-Europe, Middle East, South Asia, Other Asia and the Pacific, Africa, and South America - including Central America) to facilitate graphic representation.

Results in Figure 2 reveal that immigrants from all origins portray a similar increasing pattern of fertility by age at immigration that peaks in the late teens. The difference across groups is one of levels of fertility. Immigrants from the Middle East display high fertility rates even if arriving before age six (approximately 7 per cent higher than the native born), while the fertility rate of those arriving between ages zero and five is relatively low among Africans and South Asians (around 15 per cent lower than the native born) and even lower for immigrants from the Rest of Asia and the Pacific (around 30 per cent lower than the native born). Immigrants from the Middle East and South Asia show very steep profiles with age at immigration (for these groups the fertility of the eldest arrival group is 70 per cent points higher than that of the youngest arrival group). Immigrants from Africa and the Rest of Asia and the Pacific also show an increasing fertility profile (the fertility of the eldest arrival group being 41 and 32 per cent points higher than that of the youngest arrival group, respectively). The increase is more moderate for American immigrants.

Figure 2.

Incidence Rate Ratios (IRRs) from a Poisson regression of the number of children in the household of women who migrated as children grouped by age at migration and place of birth relative to Canadian born women, Canadian Census 1991–2006.

Note: Controls for age, education, birth cohort, location of residence and census year are included in the model. Estimates of the Poisson regression are available in Table 2A of the online appendix. Rest of Asia includes the Pacific

Source: As for Table 1.

These patterns suggest that, fertility adaptation is a relatively smooth process for US-Europeans, South Americans and Middle Easterners in the usual sense; the younger they arrive to the country, the more similar their fertility is with respect to the native born. Immigrants from Africa and South Asia show a somewhat different pattern, as the fertility levels of those arriving very young are significantly lower than those of Canadian born even though they rise subsequently for immigrants arriving later. However, among those from Rest of Asia and the Pacific fertility just reaches Canadian-born levels for those arriving in their late teens. Taken literally it is puzzling to talk about adaptation for these groups, as the fertility of their very young immigrants is significantly lower than the fertility of Canadian born. We believe that this is in part due to age and educational compositional differences across groups in the sample. Among those who arrived between zero and five years of age, the share who are still under 25 years of age - and far from completed fertility when observed in the Census - ranges from around 28% for US/European migrants, to around 55% among African and South American and up to two thirds for those from “Rest of Asia and the Pacific” and Middle East. Although the models controlled for age, they may not completely account for the multiple interactions of age and place of birth that explain this pattern. As age indicators only account for the average effect for all immigrants, it is possible that some groups of immigrants – in particular those who arrived before middle school from Rest of Asia and the Pacific and who are still disproportionately young, have different preferences for fertility and schooling than the average native-born. Given the sample size, it is not possible to add to this model a complete set of interactions of age, schooling and area of origin to ascertain such pattern. However, when we restricted the sample either to women aged 35 and older or to the birth cohort 1961–1966, for whom we observe most of their reproductive cycle, results show the expected adaptation pattern regardless of the place of birth. The fertility of immigrants arriving between zero and five years of age is relatively similar to that of the native born (only slightly lower for those from Rest of Asia and the Pacific), and increases above it with age at immigration for all groups (Results are available upon request). Whether the currently-young child migrants from these areas will follow the same pattern as previous cohorts remains open to speculation. However, judging by the behaviour of older cohorts, it is possible that the younger cohorts are just delaying fertility to a greater extent than the native born.

Age at immigration and Language Fluency

Figure 3 presents the fertility of immigrants by age at migration, with and without official mother tongue (OMT) separately, relative to Canadian born women with no official mother tongue. Migrants from the US are the reference group depicted. The graph also shows that the fertility rate of natives with an OMT is 5 per cent higher than that of other Canadians. This is consistent with previous findings in the literature that second generation immigrants in Canada have on average lower fertility levels than first generation immigrants and third generation Canadians (Ferrer and Adsera 2010).

Figure 3.

Incidence Rate Ratios (IRRs) from a Poisson regression of the number of children in the household by mother tongue of Canadian women and those who migrated as children by age at immigration, relative to Canadian born women with no official mother tongue, Canadian Census 1991–2006.

Note: Controls for age, birth cohort, education, location of residence, census year and place of birth (reference: USA) are included in the model. Estimates of the Poisson regression are available in Table 3A of the online appendix.

Source: As for Table 1.

Immigrants without an OMT have higher fertility rates than any other group, even if they arrived at very young ages. The fertility of migrants from the US with an OMT and who migrated before age five is similar to that of Canadian born with an OMT. The relative fertility of both types of immigrants, with and without an OMT, increases with age at immigration, although the profile is steeper for those without an OMT (particularly among those who arrived as teens). Unlike previous studies on immigrant outcomes in the US (Bleakly and Chin 2010), we do not find any ‘critical age’ (e.g. middle school) at which the behavior of migrants with and without OMT start to diverge sharply. Nonetheless, since there are persistent_[0] differences between the two groups, language fluency, proxied by the first language the immigrant spoke as a child, clearly plays a mediating role (presumably by enabling easy access to local norms for those fluent in an official language). However, the fact that the age at migration profile is increasing even among those with an OMT suggests that it is not the main mechanism in the adaptation of the child migrants to Canadian fertility patterns. This finding is robust to considering separately the effect of speaking French in Quebec and English in the rest of Canada.

Age at Immigration and Education

Figure 4 displays the fertility of immigrants with and without university education relative to that of native-born Canadians without university education. Immigrants without university education follow the same increasing pattern by age at migration shown in Figure 1. Their fertility is significantly higher than that of natives with the same educational attainment even among those who migrated at very young ages. By contrast, the fertility of university educated immigrants is fairly independent of their age at migration and similar to that of university educated Canadian-born, except for a slight increase among those who arrived at 18. These results are robust to broader specifications of post-secondary education that include college or other forms of post-secondary. Further, when the model was estimated with three levels of schooling - less than high school, high school and college –fertility differentials with respect to natives by age at migration emerged mainly for those without completed secondary education. To ensure that age-composition effects were not driving these results, we re-estimated this model using either only the 1961–66 birth cohort or the subsample of women over 35, and obtained the same patterns. In addition, the estimated age at migration profiles were relatively flat for the subsample of the university educated regardless of the place of birth

Figure 4.

Incidence Rate Ratios (IRRs) from a Poisson regression of the number of children in the household by education of Canadian women and those who migrated as children by age at immigration, relative to Canadian born women with no university education, Canadian Census 1991–2006.

Note: Controls for age, birth cohort, education, location of residence, census year and place of birth (reference: USA) are included. Estimated from Poison regression are available in Table 4A of the online appendix.

Source: As for Table 1.

These findings suggest that, for child immigrants, educational achievement and the degree of fertility adaptation to native patterns are highly associated. However, since these are reduced form estimates, they fail to take into account the joint nature of schooling and fertility decisions and how preferences regarding these variables are transmitted inter-generationally. No causal statements can be made since unobserved heterogeneity may be driving both education and fertility choices

4. Summary and Discussion

The fertility of immigrant women is shaped by complex forces, and influenced by the migration process itself in different ways (Goldstein and Goldstein 1981). In contrast to previous research that had centered primarily on the fertility outcomes of adult migrants, this study focused on the behavior of immigrant women arriving to Canada before age 19. The fertility of those who migrate before adulthood should hardly be affected by disruption around the time of migration (Ram and George 1990; Ng and Nault 1997) or by any interrelatedness between migration and family formation that may boost fertility immediately after arrival, as suggested by the family formation hypothesis (Anderson 2004; Sobotka 2008). Moreover, selection bias at the individual level among child migrants should be less of a concern. Hence, although we cannot entirely discount the selection hypothesis, the adaptation hypothesis is clearly the most salient mechanism for this group. Child immigrants have had more time than adult immigrants to internalize the host country’s norms regulating fertility behavior, since their socialization during childhood takes place mostly in the destination country (Hervitz 1985). Of course, selection may still play a role at the household or family level. The Canadian immigration system, which awards more points to the highly educated, might have an indirect influence on child immigrants’ educational achievement through parental education (Corak 2008). In this regard Jasso (2004) shows that the parental choice of immigration among Mexicans to the US is partly influenced by their children’s future prospects. Parents who value education more might be more likely to migrate than others and to choose to move while their children are young enough to benefit from the educational opportunities available at destination.

We hypothesized that the younger immigrants are when they arrive to Canada, the closer their fertility patterns are to the ones of native born. To be sure we find a positive relationship between relative fertility and age at immigration, with a steeper increase for those arriving in their late teens. This profile is similar among immigrants coming from different regions of the world, although their actual fertility levels vary. A puzzling result that conflicts with the expected patterns of adaptation is that immigrants from Asia and Africa have significantly lower fertility than the Canadian born if arriving very young and, in the case of those from the Rest of Asia and the Pacific, only reach Canadian born levels of fertility if arriving during late adolescence. When we examined the fertility patterns of older women from those regions, for whom we observe most of their full reproductive cycle, results show the expected adaptation pattern regardless of the place of birth. It appears that recent immigrants from these areas who arrived very young might be following a very distinct pattern that involves higher levels of schooling and later fertility than similarly aged native born. With the data in hand it is not possible to know whether this will result in low quantum by the end of their fertile years. One reason to speculate that their completed fertility might ultimately be lower is their exceptionally high educational attainment. Since the educational attainment of recent migrants from Rest of Asia and the Pacific who arrived young is well over the average, it might that the appropriate benchmark to use when measuring their relative fertility is highly educated native born (with an estimated relative fertility of 0.77 compared to less educated natives, see Figure 4) instead of all natives. In that scenario the estimated patterns of relative fertility by age at migration for those from Rest of Asia and Pacific in this study would conform to the expectations of the adaptation hypothesis.

In exploring plausible mechanisms through which adaptation might take place, we assessed the influence of language fluency and educational attainment. Unlike previous studies on other immigrant outcomes (Bleakly and Chin 2010), we do not find a sharp discontinuity in immigrant fertility behavior relative to natives among migrants arriving before or after the age of entry into middle school and whose mother tongue is different to that of the receiving country. Their relative fertility increases rather continuously with arrival age. Moreover, fertility of immigrants with an official mother tongue also increasingly differs from that of natives by age of migration, though it remains lower than that of immigrants without an official mother tongue for all ages of arrival. Among educational groups, the pattern of increasing fertility differentials with age of arrival is apparent only among less educated immigrants. University educated young immigrants behave like natives regardless of their age of arrival and this result is independent of the region of birth. Findings of the study are robust to controlling for language spoken at home, intermarriage, and language of the spouse, among other things.

Some questions are left unanswered to unveil the mechanism through which fertility adaptation occurs. We find evidence that schooling rather than language fluency to be more related to the extent of fertility adaptation, but the data do not allow us to account for the role of educational choices. Hence, unobserved heterogeneity may determine both education and fertility choices: women who have low preferences for child bearing may choose high levels of education. This type of selection problem is common to research on fertility decisions of women as most variables that determine fertility are likely to be choice variables themselves. Conventional treatments of selection bias rely on longitudinal data, not available to us. Use of instrumental variable is also limited because these estimators are yet not computationally- developed to use with count models including large number of variables. Future research should focus on the endogeneity of education and how it is tied up with fertility and labour market outcomes of immigrant children.