Migration and Ethnic Nationalism: Anglophone Exit and the 'Decolonization' of Québec

This article explains the effects of ethnic nationalism on Anglophone and Francophone migration. The rise of Québec ethnic nationalism in the 1960s dismantled the cultural division of labour, which created new opportunities for Francophones but threatened Anglophones' traditional dominance over the Québec economy. This had negative consequences for Anglophones but positive outcomes for Francophones, which in turn accounts for differences in migration patterns. Drawing from the internal colony model as well as migration and exit-voice theories, and using ecological census data, micro-census data and election panel data, I find that the key variables that increase the likelihood of Anglophone out-migration either do not explain Francophone out-migration or have opposite effects. This is because ethnonationalist policies decreased the economic return particularly for well-educated, higher earning, professional Anglophones in Québec, while increasing the economic position of Francophones and in particular well-educated professionals.


Introduction
Ethno-nationalist movements seek to define nationality as an inherent trait, creating -as Liah Greenfeld and Daniel Chirot (1994: 87) suggest -an impermeable 'us' vs. 'them' boundary. Therefore, it is not surprising that aggression, warfare, brutality and sustained violence often accompany nationalist movements. However, not all reactions to ethnic nationalism involve violence. Far less attention has been paid to out-migration as a response to ethnic nationalism. Those studies that do discuss the link between emigration and nationalism generally view it as an alternative to rebellion (see Wood 1994) within contexts where the regime is generally oppressive and out-migration is blocked (see Colomer 2000;Pfaff 2006 on blocked exit), and/or where there is sustained violence or brutal aggression, and/or where migration is forced (e.g. Boswell and Dixon 1990;Weiner 1996). Others have examined how immigration I begin with a brief overview of the internal colony model (and CDL) and the ways in which this perspective helps explain the link between migration and ethnic nationalism. I then discuss key aspects of migration theory, particularly the ways in which changes brought on by the Quiet Revolution are interpreted differently by Anglophones and Francophones. I draw from exitvoice theory to shed light on out-migration relative to alternative options like protest, and outline certain expectations that help disentangle the political and economic motivations for emigration. I use an ecological analysis of Québec regional migration to establish which regional characteristics differentially explain Anglophone and Francophone migration. I then use individual-level census data that includes individuals who left Québec to directly model the effects of key variables on Anglophone and Francophone migration. Finally, I analyse Canadian election panel data in order to shed light on possible motivations behind out-migration using attitudinal variables (like opinions on bilingualism, federal and provincial governments, etc) not available in census data. In the conclusion I discuss the impact of the findings, as well as the utility of the internal colony model in understanding out-migration as a consequence of decolonisation. I also address the theoretical implications of the Québec case, and discuss avenues for future research.

Theoretical background
Internal colonialism (Gellner 1964;Hechter 1974;1975: 9;Hechter and Levy 1994) refers to the ways in which uneven industrialisation and development emanating from the core (or several centres) towards the periphery creates resource and power inequalities between groups such that an elite class seeks to maintain their dominance over the periphery by limiting access to the subordinate group. Internal colonialism explains the rise of ethnic nationalism through the existence of a CDL -a system of stratification where cultural or ethnic differences are superimposed upon class differences. Ethnic identity in the periphery becomes increasingly salient when the group is bounded within certain occupational niches and when life chances are increasingly determined by ethnic group membership (Hechter and Levy 1994: 185-6). Subordination creates strong solidarity in the periphery, where the subordinate group forms a separate collective national identity in order to redress inequality by seeking independence from the core. Thus, the internal colony perspective is a 'classbased', reactive ethnicity model (see Nielsen 1980) such that ethnic nationalism is more likely where there is a sustained CDL.
For much of Québec's history, a CDL defined the relationship between the majority Francophone and minority Anglophone populations. In 1938, sociologist C. Everett Hughes (1938a: 348;1938b) noted that 'such people [Anglophones], with their families, constitute about one-twentieth of the population -an alien and socially isolated element, mixing little with the native [Francophones] population and not at all with the native laboring in 4 David Pettinicchio the inner part of the town'. It is therefore not surprising that Québec has often been compared to other internal colonies like Wales and Brittany. As Milner and Milner (1973: 57) state, 'Only in a colony are the majority disadvantaged and the minority privileged. Only when the language of economic control is foreign and when its organisational objectives are remote from the needs of the great majority of the population and lie in the hands of an external elite does such a situation arise.' However, by the 1960s the Quiet Revolution effectively 'decolonised' Québec, which meant dismantling the de facto CDL that had for decades been upheld by provincial governments and the Catholic Church (Fraser 2001; see also Tetley 1982). The internal colony approach has been met with criticism on a variety of levels and alternative explanations have been offered (see McRoberts 1979;Meadwell 1983;Olzak 1982; see also Bélanger and Pinard 1991;Darroch 1979;Lieberson 1970;Meadwell 1993;Olzak and Nagel 1986). Although these critiques are compelling, I suggest that the utility of the internal colony model lies not necessarily in its ability to explain the emergence of regional variation in Francophone ethnic nationalism (which is not the subject of this article), but rather for the insights it provides on the link between ethnic nationalism and politically motivated out-migration. Indeed, Québec nationalism was a means by which Francophones could improve their socio-economic position (Hechter 1992: 275), but in so doing created negative conditions for many Anglophones, which consequently led to increasing emigration in the 1970s. Thus, the internal colony perspective provides the background for understanding Anglophone out-migration. However, it does not provide the necessary mechanics outlining the specific factors that explain migration vis-à-vis its alternatives. Therefore, it is necessary to draw from migration and exit-voice theories to shed light on these processes.

Individuals, contexts and migration
Studies of migration have found that certain individual characteristics consistently increase the likelihood of migration. Educated, professional individuals with high incomes are more likely to migrate because they are aware of the conditions at the place of origin relative to conditions elsewhere and can afford the costs of migrating (Gimpel and Schuknecht 2001;Irwin et al. 2004;McInnis 1971;Ritchey 1976;Stone 1974;Trovato and Halli 1983). Migration theory also suggests that there are push and pull factors at the place of origin and destination, respectively, that explain patterns of migration. These may be cultural, economic or political in nature (see Irwin et al. 2004;Trovato and Halli 1983). Therefore, migration is best understood when contextual characteristics are linked to individual preferences and motives; as Lee (1966: 50) explains, 'the set of +'s and -'s at both origin and destination is differently defined for every migrant or prospective migrant'.
This article approaches Anglophone out-migration as a consequence of ethnic nationalism such that ethno-nationalist policies of the 1960s and 1970s Migration and ethnic nationalism 5 increased the costs, while reducing the benefits, for many Anglophones of staying in Québec. However, some scholars have called this relationship into question, arguing that Anglophone emigration may in fact be part of broader demographic and economic trends unrelated to political tensions (Amit-Talai 1993;Caldwell 1983;Caldwell and Fournier 1987;Locher 1988;Maheu 1983;Termote 1980 and2002). The most compelling argument against politically motivated out-migration is the broader movement of business westward beginning after World War Two and Toronto's growing link with the American economy, which consequently affected migration patterns (Razin and Langlois 1996;Stone 1974). But this does not explain why, for instance, the traditional disparity between Anglophone and Francophone out-migration increased from about ten times to as high as 17.5 times following the Quiet Revolution. An important study by Locher (1988), who sought to establish the non-political motivation behind Anglophone emigration, found that 29 per cent of Anglophones reinterviewed in his panel study still mentioned political motivations and dissatisfaction with the political situation in the province as reasons for leaving -which is greater than the 24 per cent who cited economic opportunities or transfers by their employer. It is important to consider the economic consequences of ethno-nationalist policies for both Anglophones and Francophones. Harry G. Johnson (1994: 239) described nationalistic economic policies as having three characteristics: control over manufacturing to demonstrate 'industrial competence'; ensuring that managerial and professional jobs are given to the middle and welleducated classes; and public ownership of economic enterprises. All three, to some extent, describe the policies of the Quiet Revolution. In the 1960s, the Liberals consolidated private Anglo-run utility companies into a national utility (Hydro-Québec), supported French-Canadian business and founded the Caisse de Dépôt, which manages the Québec pension plan (Fraser 2001). In the 1970s, changes in the banking and insurance sector, like the merging of certain Québec banks and new business alliances between Francophone-run financial institutions, increased the economic power of Francophone entrepreneurs (Bélanger and Fournier 1987). In addition, Francophone-run financial co-operatives diversified their economic interests and grew at the expense of traditional banking establishments.
For many Anglophones, these policies were seen as an attack on institutions to which they had traditionally controlled access (see Carens 1995;Lemco 1994). In addition, the economic return, especially for Anglophone men, to speaking only English in Québec went from positive to negative (Vaillancourt, Lemay and Vaillancourt 2007). Overall, the percentage of Anglophones in the well-paid labour force dropped by almost 32 per cent whereas it increased by almost the same amount for Francophones (see Levine 1990: 197). Anglophone business ownership dropped by 8.1 per cent and Francophone ownership increased by 7.7 per cent between 1961 and 1978 -a trend that has continued into the 2000s (Vaillancourt, Lemay and Vaillancourt 2007). This is especially true for the Québec financial sector, David Pettinicchio where 53.1 per cent of businesses were Anglophone-owned versus 25.8 per cent Francophone-owned in 1961. By 1978, the per cent of Anglo-and Franco-owned employers in the financial sector were virtually tied; by 2003, there were twice as many Franco-owned employers than Anglo-owned employers. In addition, between 1970 and 2000, the wage differential between Francophones and Anglophones in Québec fell by 25 per cent but only by 10 per cent in the rest of Canada (Albouy 2008).
If it were the case that Quebeckers were strictly responding to pull factors, such as better economic opportunities outside the province, then both welleducated, professional Anglophones and Francophones (especially Francophones who speak English) are more likely to migrate no matter what the political circumstance because they are the most knowledgeable of external opportunities and presumably are interested in improving their economic position. However, Francophone nationalism and the secession movement were in large part driven by the desire to improve the economic position of Francophones (not Anglophones). In other words, how can nationalistic policies be similarly interpreted by both groups when ethno-nationalistic policies altered Anglophone and Francophone incentive structures in opposing ways?

Political efficacy and out-migration
Out-migration is not the only option available to disgruntled citizens. Individuals also have the option to accept new conditions, or to protest (see Lichbach 1994). Exit-voice theory (see Hirschman 1970 and suggests that exit is usually less costly than voice. Consequently, most studies using an exit-voice model tend to find that invested individuals who can bear the cost of collective action are more likely to choose protest over emigration (see Lyons and Lowery 1989;Opp 1994;Pfaff 2006). However, both the costs of, and opportunities for, exit and voice are dependent on an individual's attachment to the place in which they live, as well as on obstacles imposed on exit and voice (see Colomer 2000;Pfaff 2006). If there are no suitable alternatives and/or exit is blocked, the costs of exit might be higher than the costs of voice but if exit is fairly inexpensive and available, then individuals may prefer emigration to protest. In the case of Anglophone Quebeckers, emigration was a fairly inexpensive option (because most who left settled within Canada, especially in neighbouring Ontario). At the same time, voice, although not repressed, was ineffective. Thus, high levels of outmigration might be partially accounted for by the inefficacy of Anglophone voice, which increases the attractiveness of leaving.
Although there are few systematic studies of the perceptions of Anglophone political efficacy during this time, Gary Caldwell's (1981) study found that young Anglophones expressed a sense of fatalism (see Odile-Magnan 2005: 49), suggesting that exit or acquiescence (but not protest) are the only two options available to Anglophones. Stevenson (1999), who draws directly from Migration and ethnic nationalism 7 Hirschman's theory, also attributes part of the increase in out-migration to the growing political ineffectiveness of Anglophones. Why were Anglophones politically ineffective, given that they had, until the 1960s, maintained privileged access to the economy and had political sway? When prior attempts at collective action are met with success, then consequently voice is perceived as efficacious (Opp 1998). However, when prior collective action is met with failure, voice is seen as ineffective, which in turn makes voicing discontent less attractive. This may explain why Anglophones were more willing to voice their discontent about language and education policies prior to 1976. Successive governments up to that period tended to back down on language policies when Anglophones protested. Both the Union Nationale (UN) and the Liberal Party promised to protect the right to English-language instruction (Levine 1990: 72) and generally took the side of Anglophones when the community expressed discontent. 2 The Anglophone business community also protested. Letters were allegedly leaked to the media in an attempt to highlight the negative economic impact of a Parti Québécois (PQ, the separatist party) victory. A notable example is the 'Brinks Show', where the Royal Trust arranged for a convoy of Brinks trucks to cross the Ontario border allegedly carrying securities out of Montréal. But, despite Anglophone protest, the PQ still won 23 per cent of the popular vote and five Montréal ridings in the 1970 election. Their win signalled the growing ineffectiveness of Anglophone voice. More importantly, the Liberal Party's enactment of Bill 22 in 1974, which made French Québec's official language for the first time in its history and also attempted to curb immigrant admission into the English-school system, was seen by Anglophones as a betrayal by 'their' party (see Coleman 1981). A petition signed by 600,000 Quebeckers was sent to Ottawa urging the Canadian government to disallow the bill. However, the federal government responded by stating that English minority-language rights in Québec were still better protected than the French minority-language rights in other provinces (Tetley 1982). As retribution, many Anglophones did not vote for the Liberal Party in the 1976 election (Levine 1990: 109). Thus, while Francophones benefited from new economic gains and their political voice became increasingly effective, Anglophones saw a decline in economic opportunities and their voice became ineffective. This left two viable alternatives: accepting the new political and economic conditions in the province, or exit.

Ecological analysis 3
The aim of the ecological analysis is to explain regional Anglophone and Francophone migration. This involved regressing the estimated net Anglophone and Francophone migration rate on several key division-level characteristics using ordinary least squares (OLS) regression. The sample 4 8 David Pettinicchio contained 100 census divisions. 5 Studies of Québec regional migration tend to suggest that Anglophones are more likely to leave regions (usually rural) where there is an established but small and declining Anglophone community (Amit-Talai 1993). Usually, economic motives are offered as reasons for their move to Montréal but some (like Gignac 2003) suggest that ethnic or linguistic tensions are also at play because Anglophone outmigration from rural regions to Montréal is considerably higher than it is for Francophones (Odile-Magnan 2005). Although the ecological analysis sheds a great deal of light on regional factors accounting for differential patterns of migration, it cannot account for whether out-migrants actually left the province.

Dependent variable: migration
The dependent variable in the ecological analysis was a measure of the estimated rate of Anglophone and Francophone netmigration in each census division between 1971 and 1981. I obtained division-level data by manually coding 1971 and 1981 census volumes. I then used the life-table survival rate method to estimate the predicted 1981 Anglophone and Francophone populations (see Shryock, Siegel, Larmon, et al. 1975). The estimated 1981 populations were obtained by applying a province-wide life table for 1975-6, the midpoint of the time interval, to the enumerated populations in 1971. 6 The estimated net migration was found by subtracting the predicted number of individuals in a division from the observed number of individuals in a division in 1981. This is then converted to a rate by dividing the absolute netmigration by the observed population in 1971. A positive value indicates net in-migration while a negative value indicates net out-migration. The Anglophone rate was transformed using a 'plus-one' log transformation in order to make the distribution more normal.

Independent variable: division resources
The key independent variable was a scale ranging from 1 to 5. The scale was intended to capture the extent of a division's physical and human resources and/or capital in order to explain regional variation in out-migration. Divisions were assigned '1' if the values for those divisions on income, per cent professionals (including managers, teachers and persons in the technology and medical sectors), value of dwellings, per cent with ties to the financial sector and per cent with college degrees were greater than province-wide averages. Thus, divisions that received a '5' were extremely resource-rich because they received a '1' on each of these characteristics.

Control variables
(i) Credit -this variable measured the extent to which Anglophones were involved in the growing credit union sector in the province, which was an alternative to the Anglophone-dominated traditional banking sector. This information is available in a 1973 document titled Répertoire des Caisses

Analysis using micro-level census data
The ecological analysis examined cross-regional variation in Anglophone and Francophone migration, but it did not speak directly to possible individuallevel motives for leaving Québec, and assumed that Anglophones and Francophones had access to division resources. In order to supplement the ecological analysis, I analysed the 1981 Canadian micro-census (i.e. individual-level) data, which was available through IPUMS International. I used logistic regression models with key independent variables predicting the odds of Anglophones and Francophones leaving Québec. 8 Dependent variable: migration A dichotomous variable disaggregated by mother tongue (left Québec: 1 = yes, 0 = no). I identified Anglophones and Francophones by respondents' selfproclaimed mother tongue. In order to obtain a measure of migration, I used a survey item that asked respondents in 1981 if they lived in another province 5 years previously ('province of residence 5 years ago').

Independent variables
(i) Education -a dummy variable for whether individuals received a university certificate or diploma or a bachelors degree/professional degree (1 = yes, 0 = no). (ii) Occupation -this is based on the International Standard Classification of Occupations. Occupations were recoded intro three types: professionals (including legislators, senior officials, managers and professionals), clerks and manual/service workers (service workers and shop and market sales, skilled agricultural and fishery, crafts and related trades workers, plant and machine operators/assembly, and elementary occupations). (iii) Income -recoded as a categorical variable (<$10,999, $11,000-$19,999, $20,000-$29,999, >$30,000). (iv) Financial Sector -a dummy variable based on industrial codes (1 = individual is in financial, real estate or insurance sector; 0 = not). . This variable came from the 'knowledge of official languages' question, which asked whether respondents could conduct a conversation in English ('Can you speak English well enough to conduct a conversation?'). Although this may not capture in great detail the quality of the person's English-language skills (there are no other measures available in this dataset), it does capture whether the respondent has a working knowledge of the language. I also included an interaction term for Francophones (speak English ¥ professional) in order to determine whether professional Francophones who speak English were also seeking better opportunities outside Québec, or whether they were remaining in Québec to take advantage of improved economic opportunities as a result of nationalistic policies.

Analysis with election panel data
Census data did not provide any attitudinal measures regarding Québec nationalism. In order to ascertain whether dissatisfaction explained Anglophone out-migration, I used data from Canadian National Elections and Québec Referendum Panel Study (Clarke, Jenson, LeDuc & Pammett 1982. Although the data were not directly linked to Canadian census data, they still provided insight into the individual-level motivations that complemented both the aggregate and micro-level analyses. The data were based on personal interviews conducted with 2,562 individuals following the 1974 federal election, and 1,295 individuals from the 1974 sample were once again interviewed following the 1979 federal election. Because the survey asked respondents in which province they lived in 1974 and in 1979, I was able to identify the 152 individuals who once lived in Québec, but in 1979 lived in other provinces (approximately 80 per cent of the individuals who left were Anglophones, over half of whom moved to neighbouring Ontario). I used difference of means tests to determine whether there were significant mean differences across samples on key demographic and attitudinal variables. Attitudinal variables (i) Closer to federal government -per cent of individuals who answered 'federal' to the question 'Would you say that you feel closer to the federal government in Ottawa or to your provincial government?'. (ii) Separation unlikely -per cent of individuals who believed that Québec separation was not very likely. (iii) Avoiding bilingualism -this variable was meant to capture the notion that Anglophones, although not opposed to the concept of bilingualism, would rather avoid it. Thus, I used the per cent of individuals who answered 'no compulsory second language' and 'French where there are French people' to the question 'What do you usually think of when bilingualism is mentioned (prescriptive)?'. (iv) Bilingualism negative -this variable was meant to capture outright negative feelings towards bilingualism (in response to the question 'What do you usually think of when bilingualism is mentioned?'). This represents the per cent of individuals who had completely negative attitudes towards bilingualism.

Results
Sixty-four of the 100 divisions experienced net Anglophone out-migration between 1971 and 1981 compared to forty-two divisions that experienced Francophone out-migration. Over one-third of all divisions that experienced net Anglo out-migration were on the island of Montréal and twelve of the top twenty divisions with the highest net Anglophone out-migration rate were in Montréal. This, in addition to the one-third of divisions on the island of Montréal that experienced net Francophone in-migration, casts doubt on the argument that internal migration is limited strictly to economic motives and suggests, as Gignac (2003) argues, that political motivations matter as well. In addition, the data may suggest that Francophone in-migration to Montréal occurred because Francophones were taking advantage of new economic opportunities (especially in the financial sector) created by the Quiet Revolution. I will examine migration patterns more systematically in the following sections. The ecological analysis (see Table 1) includes five models whereby the net Anglophone and Francophone migration rate at the division level is regressed on key economic, political and cultural variables. In Model 1 (Table 2), the Anglophone migration rate is regressed on economic variables only. Models 2 and 3 control for political variables as well as other variables that may explain variation in regional migration. Across all three models, the key independent 12 David Pettinicchio variable -division resources -is negatively related to Anglophone migration. This means that more resource-rich divisions have greater Anglophone outmigration (or less in-migration), net the effects of other political, economic and cultural variables. On the other hand, division resources are positively related to Francophone migration, which means that resource-rich divisions experience Francophone in-migration. The ecological analysis lends support to the argument that Anglophone migration in Québec is not simply a matter of individuals responding to better economic opportunities. The divergent effects of division resources on Anglophone and Francophone migration suggest that, because nationalistic policies negatively affected Anglophones' economic position in Québec, while positively affecting Francophones' position, Anglophones in resource-rich divisions were exiting. However, the ecological analysis only explains within-province variation in migration and cannot assume that out-migrants were necessarily leaving their region for a destination outside the province.

Migration and ethnic nationalism 13
The individual-level analysis (which specifically includes individuals who left the province altogether) provides further support for politically motivated out-migration. Table 2 presents odds ratios for two models: Models 1 and 2 predict the odds of Anglophones and Francophones leaving Québec, respectively. Values greater than 1.00 denote an increased likelihood and values less than 1.00 a decreased likelihood of exit. In general, the findings in Model 1 14 David Pettinicchio (Table 2) are congruent with migration theory. More highly educated, high earning, single Anglophones, with ties to the financial sector, have a greater likelihood of migrating. These findings are not surprising (with the exception of occupation, which is not significant). The more revealing aspect of this analysis is the comparison between Francophone and Anglophone migration. First, although income explains Francophone and Anglophone migration similarly, education works in the opposite direction for Francophones. That is, highly educated Francophones are less likely, not more likely, to emigrate. Education is a key variable in understanding Francophone and Anglophone relations because, as David Albouy (2008) claims, about 50 per cent of the reduction in the wage gap between Anglophones and Francophones (in Québec only) is attributable to increasing Francophone education levels. In addition, Francophone clerks and professionals are less likely to migrate than service/skilled manual workers, whereas occupation is not significant for Anglophones. Both of these findings lend support to Johnson's (1965)) argument that ethnonationalist policies tend to advantage the well-educated and professional class. Importantly, Anglophones attached to the financial sector are more likely to migrate, whereas this variable is not significantly related to Francophone migration. This supports existing economic data, which shows that since the 1960s the share of Anglophone-owned companies in the financial sector declined while the share of Francophone-owned companies increased dramatically. This suggests that francisation, especially of the workplace (and financial sector) -evidence of the dismantling of the CDL -created favourable conditions for well-educated professional Francophones, but did the opposite for Anglophones. This claim is also supported by the 'speak English' variable and the interaction term. If it is true that Quebeckers in general were migrating because of pull factors, then Francophones -especially Francophones who speak English (i.e. they can compete in the English-speaking labour market because they speak English) -should respond like their well-educated, professional, high-earning Anglophone counterparts. But this does not appear to be the case. In fact, Francophones who speak English are less, not more, likely to migrate than those who do not speak English. Even Francophone professionals who speak English are less likely to migrate. Because Francophones who speak English (including Francophone professionals) are more likely to remain in Québec, it implies that the ability to compete in both the French and English job markets does not explain their out-migration, which raises doubts about resource competition as proposed by Susan Olzak (1982). It also suggests that decisions to emigrate are not purely economic. The individuallevel analysis, together with the ecological analysis, suggests that Anglophones and Francophones are not merely responding to pull factors, but rather that Anglophone and Francophone migration is explained by the ways in which nationalistic policies within Québec were differently interpreted by Anglophones and Francophones (i.e. push factors for Anglophones, but not for Francophones).

Migration and ethnic nationalism 15
An important assumption in both migration and exit-voice theories is that individuals who exit are dissatisfied with the conditions in the place of origin. To my knowledge, no data exist that systematically seek to uncover dissatisfaction during this time period. However, it is possible to infer dissatisfaction and, to a lesser extent, loyalty using election panel data. As Table 3 illustrates, Anglophone out-migrants are more likely to be professionals or executives and are wealthier than both their non-emigrating Anglophone counterparts and Québec Francophones (which is congruent with both migration theory and the findings presented in the ecological and individual-level analysis). 9 Yet they also share important attitudinal differences. The most important and relevant difference concerns attitudes towards bilingualism. Comparing Anglophone migrants, Anglophones who stayed and Francophones, the study finds that a higher proportion of Anglophone out-migrants either preferred to avoid bilingualism or had completely negative attitudes towards it -even more so than Francophones. Again, this raises questions about bilingualism as a proxy for resource competition because Francophones overall do not appear to have negative feelings about speaking English.
There are other interesting findings. First, there is no considerable difference in the average number of respondents believing Québec would separate. This suggests that although ethnic nationalism played an important

16
David Pettinicchio part in Anglophone out-migration, it is not necessarily attitudes about Québec sovereignty that motivated out-migration. This is important because many studies of Québec ethnic nationalism use the sovereignty movement as a proxy for ethnic nationalism. Second, although there is no significant difference in feeling towards the federal government between Anglophones who stayed and those who left, Anglophones overall feel significantly closer to the federal government than Francophones. In terms of internal colonialism, this perhaps suggests a stronger Anglophone affinity with the core rather than with the periphery (i.e. the increasingly nationalistic provincial government). These findings suggest that at the heart of politically motivated Anglophone outmigration is francisation -especially francisation of the workplace and financial sector, which threatened the economic dominance of Anglophones in the province. This again provides support for the link between decolonisation and out-migration.

Discussion and conclusion
Ethnic nationalism in Québec successfully dismantled the CDL that had defined the social system since the eighteenth century. For educated, well-paid professional Anglophones tied to the financial sector, this meant a declining economic return to staying in Québec. Meanwhile, Francophones increased their economic position. As a result, many Anglophones left the provinceespecially because exit was fairly cheap and their voice had become ineffective. Theories like internal colonialism and resource competition have called into question the impermeable nature of ethnicity, suggesting that ethnicity is politically constructed (Nagel 1986). They also 'reject the assumption that primordial ethnicity will decline in modern states' (Olzak 1982: 256), highlighting the relevance of ethnic nationalism in industrialised countries. Québec is regarded as an example of peaceful ethnic nationalism (see Laitin 1999), and in this article, I show that responses to ethnic nationalism are not restricted to instances of sustained violence or warfare. In fact, it can be argued that one of the reasons for the peaceful nature of ethnic nationalism in Québec was the relative ease of exit: louder Anglophone voices left, in turn decreasing intergroup conflict.
A fruitful way of understanding Anglophone out-migration is as a consequence of decolonisation. There are striking parallels with other internal colonies, most notably Wales (see Hechter 1975: 164-207), where the nationalist movement of the 1960s referred to the CDL as 'English domination' (see Knowles 1999) -a similar frame to that of Québec's Quiet Revolution, whose chief slogan, 'maítres chez-nous', literally means 'masters in our own home'. In fact, Québec has much in common with other internal colonies like Brittany, Estonia, Scotland and Catalonia. However, these examples do not provide much insight into the ways in which ethnic nationalism and decolonisation prompt out-migration.

Migration and ethnic nationalism 17
This article sought to explain Anglophone out-migration as politically motivated; that is, as a response to push factors, not simply as a response to economic pull factors. The Quiet Revolution transformed Québec society. In fact, it had such a dramatic impact that 50 years later, according to a November 2010 poll, more highly educated Anglophones and Francophones (perhaps for different reasons) still see it as the most important event for Québec's identity (more so than the 1980 and 1995 referendums on Québec sovereignty) (Scott 2010). If the major reason behind ethno-nationalistic policies is to improve the economic position of an ethnic group, then these policies succeeded. Albouy's (2008) analysis of the Anglophone-Francophone wage gap suggests that nationalistic policies increased not only the economic position of Francophones but the demand for French-language skills, especially in the workplace, and decreased the demand for English-language skills. There was a negative demand shift for Anglophone labour in Québec, suggesting a loss of job opportunities in Québec and better opportunities in the rest of Canada. Francophones, particularly well-educated professionals in the financial sector, had no reason to leave Québec. A recent Montréal Gazette article commemorating the 35-year anniversary of the 'Anglo exodus' (Johnston 2009) provides anecdotal information about motivations behind Anglo emigration. For example, as one woman recalls, '. . . all of a sudden you would go into stores in downtown Montréal and people wouldn't speak English to you anymore. That was a real eye-opener.' Another interviewee said, 'We were anxious to move on with our lives. And it was really clear to us that our opportunities in Montréal were going to be pretty limited, based on the fact that we did not have the language that we would need, to have the kind of careers that we would hope for. ' To be clear, I am not suggesting that pull factors do not matter in explaining Anglophone out-migration. However, those who take the position that out-migration from Québec was purely economically driven (i.e. focus on pull factors) neglect the fact that Anglophones left because they were dissatisfied with political conditions in the province. A major reason why it has been difficult to disentangle economic and political explanations for out-migration is that there are no existing data that directly capture the motivation behind migration -whether at the ecological or micro levels. Although my approach is an improvement on existing work in this area, there are several limitations. First, an indirect method was used to estimate net Anglophone and Francophone migration at the ecological level because there are no age distributions by language group for each census division. However, for divisions where it was possible to construct an age distribution by language group, there did not appear to be great variation in age structure across these divisions, and their age structure appeared fairly similar to the provincial age structure. Second, the micro-level census data do not include individual characteristics from the exact time that individuals left the province. In migration studies, researchers are often able to explain migration only using end-of-

18
David Pettinicchio period values (see Greenwood 1975). Thus I test for simultaneity bias using the Hausman test and find that my models do not suffer from contemporaneous feedback effects. Another disadvantage of the micro-census data obtained through IPUMS is that the data are not nested and it is not possible to use multi-level modelling. Multi-level data are ideal for investigating how individual-level motivations for migrating operate, net the effects of regional characteristics. Third, there is a paucity of variables capturing ties/loyalty to, or dissatisfaction with, the place of residence. This is unfortunate because these variables are important parts of both migration and exit-voice theories. However, I do include some measures of social ties -like marriage and having children, and attitudinal measures (e.g. attitudes about bilingualism) -that are meant to capture dissatisfaction. Despite these limitations, the three different types of data, taken together, tell a consistent story about Anglophone and Francophone migration. Importantly, my analysis reveals differential responses to ethno-nationalist policies both within and between ethno-linguistic groups.
Internal colonialism is an analytically useful concept in understanding the history of Anglophone-Francophone relations in Québec. Migration and exitvoice theories shed light on politically motivated out-migration and specify some of the mechanics behind the effects of decolonisation on migration patterns. First, this article reinforces the idea that in comparing migration across ethnic or linguistic minorities, it is especially important to establish how different groups are affected by positive and negative evaluations of the territory in question. Second, it provides an empirical alternative to the numerous studies that have found that those with resources and vested interests (economic or otherwise) choose to stay and protest, rather than emigrate (Hirschman 1993 andPfaff 2006 with East Germany;Colomer 2000 with Cuba). Anglophone out-migration from Québec highlights important differences that explain why this is so: (a) voice was ineffective regardless of individuals' resource availability to invest in collective action (Okamoto and Wilkes 2008 also find that opportunities for voice matter more than costs of voice); (b) exit remained fairly cheap (especially compared to voice) and open; and (c) ethnic nationalism diminished Anglophone dominance in the economy, reducing the economic return, especially for high-resource Anglophones with 'louder voices' who may have otherwise stayed and protested.
In the 1970s, out-migration was the preferred option for many Anglophones for the reasons discussed in this article. Important work on the relationship between out-migration and protest (Dowding, John, Mergoupis, & Van Vugt 2000;Pfaff and Kim 2003;Tolnay and Beck 1992) suggests that when there is too much exit, erosion of networks decreases the efficacy of voice. This is especially applicable to Anglophone mobilisation and migration in the 1980s and 1990s. This period saw a resurgence of ethnic nationalism and a referendum on Québec sovereignty in 1995. In the 1990s, staunch nationalists and certain members of the PQ government increasingly Migration and ethnic nationalism 19 used the term 'pure laine' (translated to 'pure wool') to refer to old-stock French Canadians while blaming Anglophones, older European immigrant groups (e.g. Italians, Jews, Greeks, etc) and those with money as obstacles to sovereignty. But by this time, the CDL no longer existed (see Meadwell 1993). Anglophones no longer dominated the Québec economy; most Anglophones were required to learn French if they planned on having careers in Québec; and most new immigrants were, by law, required to attend French schools. Presumably, the Anglophones who remained accepted the new conditions in Québec. Yet the 1980s saw new attempts at mobilising discontent (e.g. the Anglo-rights group, Alliance Québec), despite the decline in the Anglophone population. Future work should seek to link the larger first wave of out-migration to the smaller wave of the 1990s, while considering the impact of out-migration on political efficacy.
6 Survival ratios for groups aged 15 and older using a 1975-6 life table (the middle of the time period of interest) were constructed and applied to the estimated English-speaking and Frenchspeaking populations. This produced the expected Anglophone/Francophone population for the census division by age group. Population estimates for groups aged 15-19 and 55-64 were then calculated. Age categories 15-19 to 55-64 were used because population values for these were available in both census years. 7 There is no systematic dataset that tracks business exit from the province. Neither the Registraires des Entreprises du Québec or Statistics Canada's Business Registry provide the necessary data. Because seven census divisions were missing manufacturing data in 1982, I replaced the missing data with available data from the next closest year (1977, 1978 and 1979). In addition, not all the Montréal area census subdivisions had this information. I used the number of remaining manufacturing establishments in Montréal that were not assigned a census subdivision and distributed them equally across these Montréal area divisions that had no data. 8 I mean-centred the independent variables to reduce multicollinearity (i.e. the mean VIF for both models was <2.00; no individual VIF was greater than 5.00; highest condition number associated with the Eigen value was approximately 10.72). The models did not suffer from specification error (using linktest in Stata 10) and fitted the data well (using the Hosmer and Lemeshow's goodnessof-fit test). 9 Although Anglophones who left Québec were more likely to have attended post-secondary institutions (and this is significantly higher than Francophones), there does not appear to be a statistically significant difference across groups in terms of receiving a post-secondary degree. Table A1.