1Following the article by Hélène Vézina et al. on the relation between ancestral origins and reported ethnic origins in Canada published in this same issue, this second article shows how an individual’s reported ethnic affiliation is liable to vary over time. Using record linkage between the 2001 and 2006 censuses, Éric Caron-Malenfant, Simon Coulombe, Eric Guimond, Chantal Grondin and André Lebel reveal the extent to which individuals’ self-reported identity changes within the space of five years. Between the two censuses, this ethnic mobility accounted for much of the increase in populations of North American Indians, and of Métis in particular, to the detriment of non-Aboriginal populations. This article shows how these ethnic categories are appropriated and used by different populations, and reveals the instability of reported ethnic affiliation over the life course.
2The population growth of Aboriginal peoples in Canada has been considerable in recent years. Despite the near absence of international immigration and mortality higher than the Canadian average (Wilkins et al., 2008), the number of persons who self-identify as Aboriginal (North American Indian, Métis or Inuit, see Section II.1) rose by 45% between the 1996 and 2006 censuses, far exceeding the population growth observed for the rest of the population (8% according to Statistics Canada, 2008). While Aboriginal peoples have a higher fertility rate than the rest of the population (Ram, 2004; Statistics Canada, 2011), this factor alone cannot explain this exceptional growth. Looking at specific Aboriginal groups, the Métis population nearly doubled between 1996 and 2006 (up 91%), whereas population growth among North American Indians and Inuit was 29% and 26%, respectively.
3It is generally acknowledged that the proportion of Aboriginal population growth in Canada that cannot be explained by demographic components is not due to measurement errors or variations in the quality of enumeration between censuses, but rather the fact that individuals change the identity they self-report in censuses over the course of their life. This phenomenon of intragenerational ethnic mobility has contributed significantly to the “demographic explosion” of Aboriginal populations since 1986 at least (Guimond et al., 2007; Guimond, 2009; Statistics Canada, 2011).
4Ethnicity and identity are not fixed, immutable attributes, but rather social constructs with a certain degree of fluidity, and the boundaries of concepts used to define them (e.g. Aboriginal, Métis) are permeable. Recognizing the existence of transfers from one group to another over the course of life has prompted many authors to advocate taking this component into account alongside fertility, mortality and migration to produce estimations and projections about the groups affected, whether they are Aboriginal or not (Guimond et al., 2007; Goldman, 2009; Perez and Hirschman, 2009). This is the reasoning behind Statistics Canada’s decision to incorporate this component into scenarios of its most recent population projections by Aboriginal identity (Statistics Canada, 2011).
5In the absence of longitudinal data allowing for a direct estimation of the phenomenon in Canada, previous studies had to rely on indirect, or residual, estimations. Following cohorts from one census to the next, they were able to measure net population gains and losses through changes in self-reporting of Aboriginal identity, but the direction and magnitude of the various flows remain unknown and the characteristics associated with these changes cannot be accurately analysed. Without such information, it is difficult to comprehend the ongoing changes in Aboriginal populations, not only in terms of size, but also in terms of various characteristics. For example, we know that Aboriginal populations in Canada differ in a number of ways from non-Aboriginal populations. They are younger and represent a higher share of the population in the western and northern regions of the country (Figure 1); they also have less education on average (Statistics Canada, 2008; Hébert et al., 2008). These populations are ageing, growing rapidly in areas where they were traditionally underrepresented, and making progress in terms of education. To what extent are these changes linked to growth of Aboriginal populations through ethnic mobility, to the detriment of non-Aboriginal populations, themselves older, metropolitan and educated? This question is relevant for the development, implementation and assessment of public policy concerning Aboriginal peoples.
Proportion of persons with an Aboriginal identity by census division, Canada, 2006
Proportion of persons with an Aboriginal identity by census division, Canada, 2006Note: A census division is defined as a “group of neighbouring municipalities joined together for the purpose of regional planning and managing common services” (Statistics Canada, 2007). In 2006, there were 288 census divisions in Canada.
6This article presents the results of an analysis of the intragenerational ethnic mobility of Aboriginal peoples using a new data source which allows a direct estimation of the phenomenon in Canada for the first time: a record linkage between the 2001 and 2006 census 20% sample microdatabases.  The paper aims to answer two main questions:
- What was the extent of ethnic mobility flows between Métis, North American Indian and non-Aboriginal populations between 2001 and 2006?
- What sociodemographic characteristics were associated with these flows? What are the consequences of ethnic mobility for the sociodemographic composition of these groups?
I – Literature review
7To answer the two main questions put forth by this paper, the literature on ethnic mobility needs to be reviewed in two stages: the first stage covers research on the magnitude of ethnic mobility among Aboriginal populations, and the second stage addresses the potential factors or characteristics associated with it.
1 – Magnitude of ethnic mobility among Aboriginal people
8Studies conducted on Canadian data have shown that growth in the Métis and North American Indian populations has been largely driven by intragenerational ethnic mobility in recent decades. Using a residual method  to observe that growth among people reporting a Métis identity and an Aboriginal ancestry exceeded the theoretical maximum natural increase of 5.5% per year,  Guimond (1999) concluded that an important part of the growth of this population between 1986 and 1996 could be attributed to intragenerational ethnic mobility. The same is true for North American Indians from 1986 to 1991, but this factor had the opposite effect from 1991 to 1996 for this group. Using a similar method, Lebel et al. (2011) concluded that intragenerational ethnic mobility accounted for the largest portion of demographic growth among the Métis and a significant share of growth among North American Indians from 1996 to 2006. In both studies, intragenerational ethnic mobility had a negligible effect on the dynamics of the Inuit population.
9Intragenerational ethnic mobility was also observed among Aboriginal populations in other Anglo-Saxon countries – sometimes as a growth factor, other times as a factor of decline – depending on the groups and periods. Using an indirect method, Passel (1996) estimated that ethnic mobility in the United States was a major growth factor during each intercensal period from 1960 to 1990 among American Indians, Eskimos and the Aleut, with an inflow higher than natural growth from 1970 to 1980. Studying the period from 1980 to 2006, Perez and Hirschman (2009) obtained similar results for 1980 to 1990 among these three populations, but they hypothesize that these groups sustained net population losses as a result of intragenerational ethnic mobility from 1990 to 2000 and from 2000 to 2006. These results can be partly explained by limitations in data comparability from one period to the next.
10In Australia, from 1991 to 1996, demographic factors accounted for barely half of the growth of indigenous populations (Aboriginal and Torres Strait Islanders), leaving a large portion unexplained, likely resulting from ethnic mobility (Ross, 1999). From 2006 to 2011, just under one third of the population growth cannot be explained by demographic components and may consequently be a result of the increasing propensity to self-identify as Aboriginal (Australian Bureau of Statistics, 2013). Another Australian study that focused not on Aboriginal identity, like Ross and the Australian Bureau of Statistics, but rather on ethnicity of ancestors obtained from a separate question on the census, shows a considerable decrease in the propensity to report indigenous origins (Aboriginal and Torres Strait Islanders) from 1986 to 2001. The author attributes this to changes in the census questionnaire (Khoo, 2006). 
11Lastly, analytical work in New Zealand supporting the assumptions of ethnic projections made by Statistics New Zealand (2010) shows that from 1966 to 2006, the Maori population generally experienced negative inter-ethnic exchanges with other groups, even though this may at times have played in its favour, as was shown by the results of a linkage between the 2001 and 2006 censuses (Brown and Gray, 2009). However, the very modest (+5,900) net gains of the Maori population from 2001 to 2006 estimated by the latter study concealed much larger inflows and outflows (which together totalled 65,000 transfers over this period).
2 – A few factors likely to be associated with ethnic mobility
12If reporting errors and changes to measurement tools and concepts are excluded,  the demographic literature identifies various factors that may contribute to changes in self-reported ethnicity over the life course. Ethnic background plays a unique role, as it is discussed in most studies on ethnic mobility. Having ancestors who belonged to different ethnocultural groups, and thus having a certain degree of mixed ancestry, forces individuals to choose and predisposes them to ethnic mobility (Guimond et al., 2007). A longitudinal study conducted in New Zealand by Carter et al. (2009) established that more than half of respondents who reported having more than one ethnicity during an interview for the first wave of the Survey of Family, Income and Employment (SoFIE) had changed their response at least once during the two subsequent waves by adding or removing ethnic backgrounds from their initial response, or by providing a completely different response. Similarly, Passel (1996) shows that only 22% of individuals who had reported having American Indian ancestors when asked about their ethnic origins self-identified as American Indians when asked about their race in the 1990 United States Census. Probably mostly of mixed origins, individuals of American Indian origin who do not self-identify as being of American Indian race make up a “pool of ‘potential’ American Indians”, which, according to Passel, might suggest potential continuing ethnic mobility in subsequent periods under favourable circumstances. Ethnic ancestry often gives rise to the assumption that ethnic mobility might be more frequent in certain places of residence. Guimond (2009) suggests that urban centres in Canada would be more likely to experience such mobility owing to greater propensity of mixed ancestry. Passel (1996) suggests the same hypothesis for the U.S. after noting that ethnic mobility has been more beneficial to American Indians living in states where their population is less concentrated.
13Guimond (1999) and Guimond et al. (2007) identify two other types of factors that, although mostly contextual, may produce lasting changes in the composition of the population: social factors and events likely to alter the perception of the groups making up society in either a positive or negative way, and thereby the sense of pride individuals may feel in their affiliation;  and legislative changes when, for instance, such changes lead to the legal recognition of a previously unrecognized group or when they broaden or narrow the legal definitions of specific groups, such as Registered Indians in Canada (see Section II.1). These authors provide examples of social factors (Royal Commission on Aboriginal Peoples, 1991 to 1996) and policy and legislative changes (1985 amendment of the Indian Act) that may have contributed to ethnic mobility in Canada during the period under study (1986 to 2001).
14Both ethnic background and social and legislative factors play a role in defining identity from an individual standpoint, which itself is related to the age of the individuals and to the circumstances and situations to which they are exposed. A number of studies suggest that identity is more fluid in younger than in older individuals. Ethnic mobility at a young age is exacerbated by the fact that individuals respond to the census themselves for the first time. This inverse relationship between age and intragenerational ethnic mobility was particularly observed in Canada (Lebel et al., 2011) and in New Zealand (Carter et al., 2009; Brown and Gray, 2009). In the United States, Perez and Hirschman (2009) observe, as per their expectations, strong net ethnic mobility not only at younger ages, but also in individuals aged 45 and older in certain groups, a result they suspect is linked to assumptions from their model on mortality and the differences in the quality of enumeration by age from one census to the next.
15Identity choices might also be linked to the level of education, as self-identification preferences may be sensitive to the environments to which individuals are exposed as they enter adulthood, in particular. This hypothesis is proposed by Eschbach et al. (1998), but no measurement is provided to verify it directly. Interested in the impact of ethnic mobility on education, these authors used indirect methods to determine that ethnic mobility contributed to higher education levels among Indians and Alaskan natives in the United States from 1970 to 1980. However, this was not the case from 1980 to 1990, when ethnic mobility was lower than in the previous decade. In Canada, Siggner (2003) and Guimond et al. (2007) also suggest that the quick rise in education levels among Aboriginal people from 1986 to 1996 may have benefited from ethnic mobility. In the absence of longitudinal data, however, like Eschbach et al. (1998), they were unable to distinguish to what extent this ethnic mobility contribution to higher education levels among Aboriginal people could be attributable to a differential propensity of ethnic mobility between education levels as compared to differences in composition by education level of the source and destination populations. The longitudinal study conducted by Carter et al. (2009) found no significant link between the probability of ethnic transfer and education in New Zealand.
II – Concepts, data and methods
1 – Concepts
16In Canada, the term Aboriginal generally signifies people who were present on the North American territory when the first Europeans arrived, and their descendents (Indian and Northern Affairs Canada, 2002). The Constitution Act, 1982 identifies three separate Aboriginal peoples: Indians, Inuit and Métis. The term “Indian” applies to all individuals recognized as Indian under the Indian Act (i.e. Status Indian registered in the Indian Register), and any individual who is not legally recognized but who self-identifies as Indian (non-Status Indian). Indians are often referred to as First Nations. The Inuit are Aboriginal people who traditionally live in the northern regions, in the Canadian Arctic. Inuit means “a people” in Inuktitut. The term “Eskimo” which once designated the Inuit is considered pejorative and is no longer used in Canada. The concept of Métis originally applied to children born in the Canadian prairies to Amerindian women and French fur traders. Today, more than one definition of Métis exists; at some times, it designates the descendents of the first prairies Métis, and at others, it designates all people who have both Aboriginal and non-Aboriginal ancestors.
17The evolving nature and complexity of the concept of aboriginality in Canada remains well-reflected by the absence of a consensus on a single definition (Guimond and Robitaille, 2009). In fact, many definitions co-exist. In this paper, three concepts are used, all derived from the information collected from the 2001 and 2006 censuses: Aboriginal ancestry, Aboriginal identity and Registered Indian status. Population counts for each of these definitions in 2001 and 2006 are presented in Table 1.
- Aboriginal ancestry: this first concept is derived from information obtained by means of a census question about the ethnic or cultural group(s) to which a person’s ancestors belonged. In 2006, the population with at least one Aboriginal ancestor, or Aboriginal ancestry population, numbered 1,678,235, about half a million more that the population with an Aboriginal identity. Among this population, 1,047,815 individuals, or 60%, of the respondents with Aboriginal ancestors, declared multiple ancestries (with or without non-Aboriginal ancestry), a proportion higher than in the population without Aboriginal ancestry (about 40%).
- Aboriginal identity: refers first and foremost to those persons who reported, through a separate question on the census, identifying with at least one Aboriginal group, that is, North American Indian, Métis or Inuit. In 2006, out of 31,241,030 Canadians, the Canadian Census enumerated 1,172,790 persons with at least one Aboriginal identity, including 698,025 North American Indians, 389,780 Métis and 50,480 Inuit. The Aboriginal identity population also includes a small number of persons who provided multiple Aboriginal identity responses (7,740) or Aboriginal responses not included elsewhere (26,760).
- Registered Indian status: this is a legal concept that, as mentioned earlier, stems from the Indian Act of Canada. It was initially created under the first version of the Indian Act (1876) to determine the right of residency on Indian reserves. Since 1985, registration is established on the basis of the status of the parents and grandparents (Clatworthy, 2005). In the 2006 Census, 623,780 persons self-identified as being Registered Indians, a vast majority of whom reported North American Indian identity.
Aboriginal ancestry, Aboriginal identity and Registered Indian populations in Canada, 2001 and 2006*
Aboriginal ancestry, Aboriginal identity and Registered Indian populations in Canada, 2001 and 2006** Includes those who identified themselves as Registered Indians and/or band members without identifying themselves as North American Indian, Métis or Inuit in the Aboriginal identity question (Statistics Canada, 2007).
18Aboriginal identity was chosen as the key Aboriginal indicator in this paper, because the other indicators available in the Canadian census do not specifically target individual membership or they do not cover all of the Aboriginal populations. Strictly speaking, Aboriginal ancestry refers to the ethnocultural membership of ancestors, whereas Registered Indian status identifies a subset of Aboriginal populations only. Moreover, the census question on Aboriginal identity, unlike the question on Aboriginal ancestry, was not changed between 2001 and 2006, making the information gleaned perfectly comparable for the period at issue.
19In its broadest sense, ethnic mobility is defined as changes in how ethnic affiliation is reported by individuals and among families (Guimond et al., 2007). The literature distinguishes between two types of ethnic mobility. First, intergenerational ethnic mobility is defined as a change in ethnic affiliation between parents and their children. Intragenerational ethnic mobility occurs when the same individual changes affiliation in the course of his or her lifetime (Boucher et al., 2009; Guimond et al., 2007; Brown et al., 2010). This paper focuses exclusively on intragenerational ethnic mobility by Aboriginal identity.
20Given the low population numbers involved and the limitations of the database used (next section), Inuit and multiple Aboriginal identity responses were excluded from the analysis. For the purposes of this analysis, which is to assess the separate effect of Registered Indian status and affiliation to one of the three Aboriginal identity groups, the few individuals who provided “Aboriginal responses not included elsewhere” were considered as not having an Aboriginal identity. Consequently, intragenerational ethnic mobility is limited here to changes that occur between North American Indian identity (single response), Métis identity (single response) and non-Aboriginal identity. Individuals with no Aboriginal identity may therefore be Registered or non-Registered Indians,  just like North American Indians and the Métis.
2 – Database
21The database used is the product of a record linkage between the microdata files from the long-form census questionnaires of the 2001 and 2006 Censuses of Canada. Made up of a sample of 20% of households representative of the non-institutional population whose usual place of residence is Canada, these microdata files include information not only on Aboriginal populations, but also on various socioeconomic characteristics.
22In theory, the probability of a person appearing in two 20% samples is 1 in 25. However, since the censuses were five years apart, the linked file excludes individuals who, during the 2001-2006 period, had left the country, had died or had been admitted to an institution. Similarly, the file also excludes individuals who were born, immigrated to Canada, were admitted as non-permanent residents or moved from the institutional population to the non-institutional population between 2001 and 2006. Moreover, since the names of respondents were unavailable in the microdata file for the 2001 Census, linking options were limited and led to other exclusions.
23Deterministic record linkage was performed in two separate and complementary steps. First, census records from 2001 and 2006 were linked using a method based on the linkage of pairs of individuals residing together at the time of each census. This method, widely used in coverage studies of the Canadian census, was chosen for its very low rate of false linkages. More specifically, for each pair of individuals formed within one household,  a linkage key was created on the basis of the date of birth and sex of both members. Where one of these pairs of individuals was identified in each of the two censuses, a “household link” was established and the individuals in this household were linked by date of birth and gender. To limit the possibility of false linkages, households could only be linked if they were in the same census division in 2001 and 2006.  In the second step, individuals were linked based on the link between dwellings in the 2001 census and those in the 2006 census using addresses. One benefit of this second step is that, contrary to the first step, it allowed for linkages of people living alone or with different people in 2001 and 2006, on the condition that they did not change address between the two censuses. The linkage of individuals residing at the same address in 2001 and 2006 was made by date of birth and gender. In total, 649,200 records were linked. The resulting database was then reweighted by Aboriginal identity, sex, age group, place of residence and marital status for the 2001 and 2006 censuses, adjusted to account for census net undercoverage.
24Because place of residence was one of the linkage variables taken into account, the file excludes most migrants, the exception being pairs of individuals who migrated together within the same census division. This may be of importance and may constitute a limitation if geographic mobility and ethnic mobility are related. Moreover, the linkage excludes persons who responded to census form 2D in 2001 or 2006. Form 2D, a variant of the main long questionnaire (2B), is distributed on Indian reserves and in the northern regions.  Although the exclusion of Indian reserves reduced the number of North American Indians by about half, their numbers remain sufficiently large for the purposes of this analysis. Work by Guimond (2009) also suggests that populations living on a reserve are less likely to experience ethnic mobility because of their homogeneity in terms of Aboriginal identity. In 2006, 88% of individuals living on an Indian reserve self-identified as North American Indians. This homogeneity reflects the fact that residency on Indian reserves is generally restricted to North American Indians. Given that the exclusion of northern regions significantly reduces the already low number of Inuit, this population was also excluded from the analysis. As we saw earlier, the net ethnic mobility of this population was very low. Lastly, the foreign-born population and members of a visible minority group  were also excluded in order to focus the analysis on the population that is most likely to make an ethnic transfer involving an Aboriginal identity. After these exclusions, the linked database consists of 436,200 records, that is, 427,200 individuals with non-Aboriginal identity, 4,500 North American Indians and 4,500 Métis in 2001.
25For validation purposes, we compared the net ethnic mobility obtained from the linked file with that estimated by Statistics Canada (2011) using a residual method applied to 2001 and 2006 census data. Statistics Canada’s residual analysis also excluded Indian reserves, territories (in the north), foreign-born individuals, and members of a visible minority group; it covered a target population similar to the one from the linked file, except in the case of internal migrants who are not excluded from the residual analysis. The results obtained from the two methods are similar: net population gains through ethnic mobility were 27% for Métis and 10% for North American Indians using the residual method, and 29% and 9%, respectively, using the linked file (Section III.1).
3 – Analytical methods
26The methodological approach relies on an analytical scheme similar to that sometimes used to study geographic mobility, which is why some tools and terms typically integral to these types of studies have been borrowed here. The analysis of characteristics associated with ethnic mobility revolves around the concept of migration flows between identity groups or specific ethnic transfers. Insofar as the characteristics of individuals making an ethnic transfer are likely to be linked to both the source (here in 2001) and the destination (in 2006) identity, we assume that each possible migration flow is in itself a distinct phenomenon and that, therefore, a transfer from Métis to North American Indian is a phenomenon distinct from a Métis to non-Aboriginal transfer. This is why the relationship between individual characteristics is determined here on the basis of not merely making an ethnic transfer, but rather making a specific transfer. More specifically, the analysis of characteristics associated with ethnic mobility focuses on four specific ethnic transfers between the three analysed groups: from non-Aboriginal to Métis, from non-Aboriginal to North American Indian, from Métis to non-Aboriginal and from North American Indian to non-Aboriginal. Transfers from Métis to North American Indian and vice-versa are not included in this aspect of the analysis due to their low occurrence.
27A logistic regression model was applied to each of these four specific transfers. These models are intended to assess the probability for a person in a given group (e.g. non-Aboriginal identity) of transferring to the specific destination group (e.g. Métis) based on certain characteristics at the beginning of the period (2001): Aboriginal ancestry, Registered or non-Registered Indian status, province/region of residence, living in a census metropolitan area (CMA),  education, age group and sex. To determine whether ethnic mobility is connected to the complexity of ethnic background, as the literature suggests, ancestry is defined on the basis of the following three categories: no Aboriginal ancestry, Aboriginal ancestry only, and mixed Aboriginal and non-Aboriginal ancestry. Since the geographic distribution of Aboriginal people differs from that of the non-Aboriginal population, this analysis focuses on the province/region of residence (as higher concentrations of Aboriginal people live in the Prairies  and in British Columbia) and on whether respondents live in a major metropolitan centre in Canada (as there are proportionally fewer Aboriginal people in these areas). Lastly, since the models (and analysis of characteristics) factor in the level of education, they were estimated only for the population aged 15 and over. However, the first part of the analysis, which looks at the magnitude of ethnic mobility, takes all ages into account.
III – Results of analysis
1 – Magnitude of intragenerational ethnic mobility
28Table 2 shows net intragenerational ethnic mobility from 2001 to 2006 estimated based on the linked file as well as the population counts studied in 2001 and 2006. The net variation in populations arising from intragenerational ethnic mobility from 2001 to 2006 was positive among North American Indians and higher still among Métis, but negative among non-Aboriginal persons. It contributed to the growth of the Métis population by 85,500 individuals, or approximately 29%, in five years. This is far ahead of the other factors,  which led to a population increase of just 17,500 individuals. The North American Indian population rose by about 31,900 people, or close to 9%, due to ethnic mobility, a component that also accounts for the majority of demographic growth here. The non-Aboriginal population, for its part, lost 117,400 people as a result of intragenerational ethnic mobility, limiting its growth to 403,700 individuals from 2001 to 2006. Because the foreign-born population is not counted, the contribution to this growth from international migration is excluded from the calculation.
Population and sources of growth among the Métis, North American Indians and non-Aboriginal people, 2001 to 2006, Canada
Population and sources of growth among the Métis, North American Indians and non-Aboriginal people, 2001 to 2006, CanadaNote: Excluding institutions, territories, Indian reserves, foreign-born population and members of a visible minority group.
29Although this data suggests, as have previous studies, that the non-Aboriginal population is a net contributor to the growth of Métis and North American Indian populations, it does not indicate whether or not the ethnic transfers are unidirectional, from non-Aboriginal to Aboriginal identity. Table 3, which cross-tabulates Aboriginal identity between 2001 and 2006, presents the multidirectional nature of ethnic mobility. In fact, about 21% of the 2001 North American Indian population and 26% of the 2001 Métis population transferred to another group during the study period, among which most self-reported a non-Aboriginal identity in 2006. Among the non-Aboriginal population in 2001, only 1% reported being Métis or North American Indian in 2006.
Aboriginal identity in 2001 and 2006, Canada
Aboriginal identity in 2001 and 2006, CanadaNote: Excluding institutions, territories, Indian reserves, foreign-born population and members of visible minority groups.
30Given the significantly higher propensity for ethnic mobility among Aboriginal people compared to non-Aboriginal persons, the net population gains experienced by the Métis and North American Indians through ethnic mobility are essentially a reflection of demographic imbalance between Aboriginal and non-Aboriginal populations. An outflow rate of just 1% among the non-Aboriginal population, which accounts for 97% of the study population, translates into approximately 220,000 ethnic transfers, or 60% of all transfers, 140,200 of which were to Métis identity and 79,500 to North American Indian identity. The exchanges between the two Aboriginal groups largely offset each other, meaning that their population gains essentially stem from their exchanges with the non-Aboriginal population.
2 – Relationships between ethnic mobility and sociodemographic characteristics
31Ethnic mobility is, therefore, a multidirectional phenomenon. What are the associated characteristics and do they vary by transfer type? Table 4 shows the ethnic transfer rates among the population aged 15 and over for the four main flows by selected characteristics.
32Specific ethnic transfers differ from one another with a number of the selected characteristics. First, having at least one Aboriginal ancestor is associated with low mobility among Aboriginal people and high mobility among non-Aboriginal people. This relationship is more pronounced among individuals with only Aboriginal ancestry for each of the ethnic transfer types. For example, of the individuals who reported no Aboriginal identity, but only Aboriginal ancestry in 2001, 14.2% self-identified as Métis in 2006. By comparison, this proportion was 11.6% among those who reported no Aboriginal identity and a mixed ancestry, and 0.4% among those who reported no Aboriginal ancestry in 2001. The other variable directly associated with Aboriginal affiliation, namely Registered Indian status, has the same relationship with ethnic mobility, as Registered Indian status is a factor of mobility among the non-Aboriginal population but a factor related to the maintenance of identity among Aboriginal people.
33The province/region of residence affects ethnic mobility differently, depending on the type of transfer. In regions where Aboriginal people are least concentrated – Atlantic Canada/Quebec and Ontario – those who reported Aboriginal identity in 2001 had a greater propensity to transfer to the non-Aboriginal identity group than in other regions. The opposite is true for transfers between non-Aboriginal people and the Métis, but the results for transfers between non-Aboriginal people and North American Indians are less clear. The differences between mobility rates of those who do and those who do not live in major metropolitan areas have typically been less significant than between provinces or regions, even though the results for transfers between North American Indians and non-Aboriginal people and between non-Aboriginal people and the Métis also suggest a link between ethnic mobility and the geographic concentration of Aboriginal people.
34The role of other variables in ethnic transfers is less clear. Ethnic mobility rates decrease with age for non-Aboriginal people, increase with age for North American Indians and do not really vary with age for the Métis. Ethnic mobility rates vary inversely with the level of education among the non-Aboriginal population, but appear to have no link among Aboriginal people. Lastly, ethnic mobility does not vary by sex among the non-Aboriginal population, but is slightly higher among Aboriginal males.
35The logistic regressions shown in Table 5 are indicative of the effect of each individual variable analysed up to now while controlling for that of the others. The results confirm most of the findings of the descriptive analysis. Having at least one ancestor from a different group than the one with which the individual self-identifies – a sign of a certain degree of complexity or mixing of ethnic backgrounds – is a strong determinant of ethnic mobility. Individuals in this situation, whether or not they are Aboriginal, show a greater propensity for ethnic transfers, as the literature suggests. These results are reinforced by those relating to Registered Indian status, which the multivariate analysis also confirms. Together, they show that the convergence of responses to questions used to define Aboriginal populations is a strong predictor of ethnic mobility, with ancestry and Registered Indian status each maintaining a distinct effect of ethnic mobility in the same direction.
Ethnic transfer rate* between 2001 and 2006 by various characteristics and Aboriginal identity in 2001 and 2006, population aged 15 and over in 2001, Canada
Ethnic transfer rate* between 2001 and 2006 by various characteristics and Aboriginal identity in 2001 and 2006, population aged 15 and over in 2001, Canada* The ethnic transfer rate (or ethnic mobility rate) represents, for each specific ethnic transfer, the percentage of the population at the start of the period that has made the transfer in question.
Note: Excluding institutions, territories, Indian reserves, foreign-born population and members of a visible minority group
36The multivariate analysis also confirms and clarifies the relationship between the place of residence and ethnic mobility, even though some exceptions remain. Generally speaking, living in one of the regions with the lowest Aboriginal concentration (Atlantic Canada-Quebec, Ontario, CMA) corresponds to higher ethnic mobility among the Métis and North American Indians and to lower mobility among the non-Aboriginal population, whereas the opposite is true in other regions. These results show that the relationship between the place of residence in Canada and ethnic mobility is not based solely on the composition of regions by ethnic ancestry, (a link suggested by the literature), as the effect of region remains significant when the effect of ancestry is controlled. Variations in the draw of identity groups involve other factors and mechanisms that might be clarified by assimilation theories. These results suggest that faster growth by way of ethnic mobility among Aboriginal groups in regions where they have low representation, observed during previous periods in Canada (Guimond, 2009) and in the United States (Passel, 1996), was not necessarily the result of more favourable probabilities of transfers than anywhere else. It is possible that it was rather due to a greater imbalance in population size, where low ethnic mobility rates among the non-Aboriginal population translate into a number of transfers which represent a relatively high share of destination group populations.
37The multivariate analysis also confirms that the propensity for ethnic mobility decreases with age among the non-Aboriginal population, which is consistent with the research findings discussed previously. This inverse relationship between age and mobility is not true for North American Indians and the Métis, however, as it does not play a statistically significant part in any age group. This difference between the composition of Aboriginal inflows and outflows by age might have contributed to decreasing net gains by age estimated from previous studies for the Métis and North American Indians by way of the residual method.
38The relationship between education and ethnic mobility observed in the descriptive analysis remains unchanged when controlling variables considered in the regression models. Table 5 shows that, net of other factors, non-Aboriginal people with the lowest level of education had the greatest propensity for ethnic mobility, a sign that non-Aboriginal individuals making an ethnic transfer resemble Aboriginal people in this respect more than they do the non-Aboriginal population, as Aboriginal people are on average less educated. Among Aboriginal people, the relationship between education and ethnic mobility is not statistically significant. These results are therefore only partially consistent with those obtained by Carter et al. (2009).
39In terms of the relationship between sex and ethnic mobility – whereby being a man increases the probability of transfer from North American Indian to non-Aboriginal but decreases the chances of an opposite transfer – it could possibly be linked to the differences between men and women with regards to labour market participation among North American Indians. Based on the assumption that certain aspects of the labour market may have an influence on perceptions of identity, this link could explain the differences between men and women (Table 5). However, the data presented here cannot be used to confirm this.
Odds ratios for changing Aboriginal identity between 2001 and 2006, population aged 15 and over in 2001, Canada
Odds ratios for changing Aboriginal identity between 2001 and 2006, population aged 15 and over in 2001, CanadaNote: Excluding institutions, territories, Indian reserves, foreign-born population and members of a visible minority group.
Significance levels: * 0.05; ** 0.01; *** 0.001; ns = not significantly different from the reference category at the 0.05 level.
40In total, the net population gains experienced by North American Indians and the Métis from 2001 to 2006 through intragenerational ethnic mobility conceal ethnic transfer flows, which differ from one another not only in their magnitude but also in their characteristics. These characteristics are generally different, and sometimes opposite, among Aboriginal and non-Aboriginal individuals. However, it is not possible to measure the impact that ethnic mobility may have on the composition of the groups solely on the basis of ethnic transfer probabilities. This impact depends not only on the composition of transfer flows, but also on the differences between the composition of transfer flows and the composition of the populations subject to ethnic mobility.
41Table 6 shows the distribution of the Métis and North American Indian populations defined on the basis of self-reporting in 2001 and 2006, by certain characteristics in 2006. These results are for the population aged 15 and over. The distribution corresponding to the identity reported in 2001 shows the numbers that would have been if no ethnic transfer had occurred; the distribution corresponding to the identity reported in 2006 shows the actual numbers following the transfers that occurred during the five-year period. The difference between the two distributions shows the net effect of ethnic mobility on the composition of both populations.
Métis and North American Indian populations by reported identity in 2001 and 2006, and selected characteristics in 2006, population aged 15 and over in 2001, Canada
Métis and North American Indian populations by reported identity in 2001 and 2006, and selected characteristics in 2006, population aged 15 and over in 2001, CanadaNote: Excluding institutions, territories, Indian reserves, foreign-born population and members of a visible minority group.
42The changes in composition by education and age as a result of ethnic mobility are altogether modest (Table 6). Without ethnic mobility, the Métis population would have been slightly less educated (28.1% of individuals without a high school diploma instead of 25.8%) and slightly younger in 2006. Among North American Indians, by contrast, ethnic mobility decreased the average level of education very slightly, as well as the proportion of persons aged 60 and over. In theory, one would have expected ethnic mobility to have a greater effect on level of education and ageing of Aboriginal populations, that are less educated and younger than the non-Aboriginal counterpart, source of population gains through intragenerational ethnic mobility. However, we have seen that non-Aboriginal individuals who make an ethnic transfer to Métis and North American Indian identities are younger and less educated than those who do not, which is not true for those making the opposite transfer.
43Ethnic mobility has further altered the geographic distribution of Aboriginal people. It has especially contributed to increasing the proportion of Aboriginal people living in Atlantic Canada/Quebec (up from 10.7% to 13.7% among the Métis and from 13.0% to 15.0% among North American Indians) and to decreasing the proportion of those living in the Prairies (49.8% of the Métis compared to 53.6% without ethnic mobility and 28.8% compared to 30.2% among North American Indians). It also helped raise, albeit modestly, the proportion of Métis individuals living in a CMA, the effect of ethnic mobility in this respect having been neutral among North American Indians. Generally, intragenerational ethnic mobility has contributed most significantly to Aboriginal demographic growth in regions with the lowest concentrations of Aboriginal people. These results underscore the importance of taking into account both the mobility rate and the population size to fully comprehend the demographic effects of this phenomenon. Low rates may in fact result in considerable population gains for small populations, such as Aboriginal people living in Atlantic Canada and Quebec.
44The purpose of this article was to assess the magnitude of intragenerational ethnic mobility in Canada from 2001 to 2006 between the Métis, North American Indians and the non-Aboriginal population living off Indian reserves and territories, as well as its relationship with certain sociodemographic characteristics. Relying on a linkage between microdata from the 2001 and 2006 censuses, the analysis has shown that ethnic transfers are not unidirectional, but rather multidirectional, with each group having experienced both gains and losses in its exchanges with the other groups, the net effect being a population increase for the Métis and North American Indians and a decrease for the non-Aboriginal population. The analysis also revealed relationships between certain characteristics and ethnic mobility. For Aboriginal people, generally speaking, reporting ancestry other than Aboriginal, not having Registered Indian status and living in a region with lower Aboriginal representation are factors associated with higher mobility to the non-Aboriginal group. For the non-Aboriginal population, having Aboriginal ancestry, having Registered Indian status, living in a region with a higher Aboriginal concentration, having a lower level of education or being younger all contribute to an increased probability of an ethnic transfer to an Aboriginal group. Lastly, the paper showed that ethnic mobility helped alter the composition of populations involved, with the geographic distribution of Aboriginal people in particular having changed as a result of intragenerational ethnic mobility.
45This study has a number of limitations. The first relates to the fact that the database used excludes a majority of persons who changed their place of residence during the period under study. It is possible that identity changes during the course of life are linked to changes in residence, motivated by education, work or a change in marital or family status. In this case, this study might have underestimated intragenerational ethnic mobility among groups most susceptible to geographic migration, such as young people. It might also have underestimated its impact on the composition of Aboriginal populations, particularly with respect to place of residence. It would be worthwhile for future studies to assess the significance of this relationship between ethnic mobility and migration, which was not possible here.
46Another limitation stems from the fact that the analysis focused on just one five-year period. Insofar as contextual factors (social, political or legal) can contribute to intragenerational ethnic mobility, as Guimond suggests, it is possible that certain aspects of the phenomenon might be presented in a different light from one period to another. It is possible, for example, that Métis ethnic mobility between 2001 and 2006 was influenced by the 2003 Supreme Court ruling in Powley, which both recognized certain Métis rights and proposed parameters to be considered in determining who belongs to this group. Since 2006, other legislative and social events involving Aboriginal people could have had an impact on the way individuals report their identity, perhaps differently than between 2001 and 2006. For example, Bill C-3 (2011), which granted tens of thousands of individuals the right to obtain Registered Indian status, or the highly publicized “Idle No More” movement,  in Canada which might have changed the perception groups have of one another or of themselves. The literature shows that ethnic mobility reversals are possible, as appears to have occurred in the United States since the 1990s and among North American Indians in Canada from 1991 to 1996. We therefore cannot rule out the fact that the five years between 2001 and 2006 were marked by certain specificities that affected the magnitude and/or characteristics of intragenerational ethnic mobility.
47For the first time, this article provides a direct estimation of intragenerational ethnic mobility in Canada. Although it answered a number of questions, particularly with respect to the role ancestry plays in the choice to self-identify as a member of one Aboriginal group or another, it raises many more. How do we explain the differences in the propensity for ethnic transfer based on level of education between Aboriginal and non-Aboriginal populations? Are the differences between men and women in terms of ethnic mobility from or to the North American Indian group tied to the labour market? What characteristics explain the regional differences observed in this regard? What is the proportion of permanent ethnic transfers? Since intragenerational ethnic mobility is a major component of changes Aboriginal populations have undergone in Canada for at least the past two decades, it would be worthwhile to conduct other longitudinal studies – for other periods or based on data sources comprising more than two time points – in order to broaden our knowledge of the populations that report different identities over the course of their life.
AcknowledgementsThe authors wish to thank Laurent Martel, Johanne Denis, Patrice Dion, Jean-Dominique Morency, Nora Bohnert and the three anonymous evaluators from Population for their comments and suggestions, as well as Aboriginal Affairs and Northern Development Canada for its financial support of this project. They are also grateful to Carol D’Aoust for her technical support.
A preliminary version of this paper was presented at the 2012 Population Association of America conference in San Francisco.
The views expressed in this paper are those of the authors and do not necessarily reflect those of Statistics Canada or Aboriginal Affairs and Northern Development Canada.
Aboriginal Affairs and Northern Development Canada.
Correspondence: Éric Caron-Malenfant, Statistics Canada, Demography Division, 1708 Main Building, 150 Tunney’s Pasture Driveway, Ottawa, K1A 0T6, Canada, tel.: 613-951-3073, email: firstname.lastname@example.org
This linked database was created by Statistics Canada to study the effect of the response mode (Internet versus paper) in the 2006 Census (Grondin and Sun, 2008). Permission to use it to analyse ethnic mobility was obtained as part of the development and validation of Population Projections by Aboriginal Identity in Canada, 2006 to 2031 (Statistics Canada, 2011), to better understand ethnic mobility and assess the feasibility of using such a database to develop projection parameters.
This method, which follows cohorts of individuals from one census to another, is used to estimate the remaining share of growth among the groups under study taking into account demographic components (migration, fertility, mortality and sometimes variations in the quality of enumeration from one census to the next). For a detailed description and discussion of two variants of this indirect method applied to ethnic transfers and the analysis of their sensitivity to certain assumptions, see Anderson and Silver (1983) as well as Perez and Hirschman (2009).
This annual rate of 5.5% stems from the combination – which the author (Guimond, 1999) believes probably never occurred – of the highest crude birth rate and the lowest crude mortality rate historically observed (Pressat, 1979).
In 1986, “Aboriginal” was given as an example in the question on ethnic group of ancestors, which was no longer the case in the 2001 questionnaire. The author indicates that of the individuals who self-identify as Aboriginal when answering the question on identity, the proportion who report Aboriginal ancestry dropped from 81% to 25% between 1986 and 2001.
Changes to census questions (wording, examples given, instructions to respondents) may affect the way in which individuals report their identity and ethnicity, as Ryder (1955) already suggested in his analysis of changes in the population sizes of ethnic origin groups in Canada in the first half of the twentieth century.
Note that during the focus groups organized by the Australian Bureau of Statistics, held to better understand the reasons that might lead to choosing to identify or not identify as Aboriginal, pride in belonging to an Aboriginal group was identified as an important factor (Australian Bureau of Statistics, 2012).
A person may be a Registered Indian without reporting Aboriginal identity for a number of reasons. This is the case, in particular, with non-Registered women who married a Registered Indian prior to 1985, because legislation at the time allowed women to obtain status by marriage. For a detailed analysis of the multiple features of concepts relating to Aboriginal affiliation and overlap, refer to Guimond and Robitaille (2009), and Guimond et al. (2007).
One pair simply comprises two individuals who share the same household, regardless of the link between them. The pair may be a couple, unrelated persons, parent and child, etc.
See note on Figure 1 for a definition of census division.
It should be noted that this linkage was not performed for the purposes of this project (see footnote 1). Rather, the initial objective was to study the effect of the response mode (Internet versus paper) on census results. In 2006, respondents could not use the Internet to respond to form 2D, so it was excluded from the linked file.
According to the Employment Equity Act, the visible minority population includes “… persons, other than Aboriginal peoples, who are non-Caucasian in race or non-white in colour” (Statistics Canada, 2007).
In 2006, a census metropolitan area was defined as a region of at least 100,000 inhabitants with an urban core of at least 50,000 people. There were 33 census metropolitan areas in Canada in 2006 (Statistics Canada, 2007).
The Prairies include the provinces of Manitoba, Saskatchewan and Alberta. The Atlantic region, also used in this article, includes the provinces of Newfoundland and Labrador, Prince Edward Island, Nova Scotia and New Brunswick.
As a result of the excluded populations, these factors are essentially births, deaths, institutional admissions and releases, migrations to and from Indian reserves and territories, ethnic transfers involving groups other than the three analysed and emigration. This last factor is generally considered negligible among Aboriginal people.
Idle No More is a protest and mobilization movement by Canada’s Aboriginal peoples which began in the autumn of 2012, in reaction to legislation by the Government of Canada that, according to demonstrators, did not respect treaties.