Research on political attitudes and behaviour in Canada tells us that
Canadian political culture is not what it used to be. Canadians are
becoming less deferential than they once were. Depending on which research
you read Canadians are either becoming more like Americans (Nesbitt-Larking, 1998; Nevitte,
1996) or less like Americans (Adams,
1998; 2003; Lipset,
1990; Peacock, 1998). Voter turnout is
down, cynicism is up, confidence in leaders is down, distrust is up. The
fundamental tenets of Canadian political culture, to the extent that we
believe what they had to say about Canadian political attitudes, appear to
be shifting. If these attitudes are changing, we have reason to believe
that other tenets, such as regionalism, are also not static. This article
argues first that interprovincial differences in political attitudes are
either stable or declining, depending on the indicator. Second, it
demonstrates that on measures of trust and efficacy, regional rather than
provincial affiliations account for variations among Canadians.
Regionalism is one of the accepted facts of Canadian political life.
The distinct fragments in eastern Canada and present-day Quebec are
credited with tempering Canadian liberalism with a Tory or feudal touch
(Horowitz, 1966; McRae, 1964; Lipset, 1968). Our
understanding of the brokerage system of political life in Canada hinges
on the existence of regional demands. Literature on different political
approaches in various provinces has received considerable attention and
has linked social practices and demography to the development of party
politics.1
For the most part
these works were neither comparative nor quantitative, but rather efforts
to account for the tenor of political life in a particular region or
province.
2 Systematic attention to quantitative data on regional political
cultures owes much to Richard Simeon and David Elkins, who examined
results of the Canadian Election Studies in the 1960s and 1970s (Simeon and Elkins, 1974; Elkins and
Simeon, 1980). The authors concluded that there are variations in
efficacy and trust across Canada, producing four distinct citizen types:
critic, supporter, deferential and disaffected. Respondents in the eastern
provinces and French Canadians displayed much lower levels of political
efficacy than their counterparts in British Columbia and Ontario. This
pattern also holds true for political trust, although French Canadians
appear more trusting that their Atlantic counterparts. The Atlantic
provinces, which displayed low trust and efficacy, were deemed
disaffected. Voters in Ontario and British Columbia were characterized as,
respectively, critics and supporters, French Canadians as deferentials.
The authors noted that each of the four categories would be present in any
provincial population, but that the overall classification of the province
flows from the existence of cultural boundaries that are meaningful to
voters. In short, Canadian provinces possessed distinct political cultures
as a result of distinct provincial institutions. Variations in demographic
characteristics that might otherwise account for deviations in political
attitudes are thus not solely responsible for these differences.
Provincial differences remained even when community size, social class,
education, age and sex are controlled for.
Elkins and Simeon subsequently updated their analysis to incorporate
later data and to clarify their argument (Elkins and Simeon 1980). Small Worlds contains a broader range
of attitudes on which to examine provincial political cultures and shows
significant differences according to national identity, role of
government, social policy preferences and spending and partisan
identification. Throughout, the collected papers argue that the province
presents an enduring boundary, although often acting in concert with
socio-demographic factors such as class, employment and immigrant status.
On some measures such as national identity, constitutional renewal,
support for a multicultural Canada and patterns of public expenditure,
provinces present diverging political attitudes and behaviours. On most
areas of public policy, however, provinces appear to have converging
preferences.
These findings are consistent with research on provincial political
cultures in Canada (Nevitte, 1995; Wilson, 1974). Among such works there is a belief that
provincial sub-cultures reflect very real variations in political
practice. In his analysis of regional political cultures in Canada, for
example, John Wilson made reference to past practices such as the absence
of Hansard in British Columbia, the underdeveloped role for the opposition
in Ontario, the absence of nominating conventions for the Liberal party in
Newfoundland and differing approaches to patronage, some more enthusiastic
than others (Wilson, 1974). Such variations,
according to Wilson, were symptomatic of different stages of political
development, which in turn could account for variations in political
values. Researchers on political cleavages also notes that there are
variations in attitudes and behaviour between women and men, French and
English speakers, among different religious groups and rural and urban
voters (Burt, 1986; Everitt,
1998; Mendelsohn and Nadeau, 1997; Nevitte, 1996; O'Neill,
2001; Wearing and Wearing, 1991; Gidengil, 1989; 1990; Ornstein, 1986). Some of these characteristics cluster
in different regions, compounding the potential existence of regional
political preferences.
Regionalism and the existence of provincial political cultures has
been an enduring element in attitude research since Simeon and Elkins
first published their research. This analysis should be revised, however,
for two main reasons. First, the original analysis draws on data that are
almost forty years old. But, the questions on which the original work
rests its conclusions have been replicated in many of the subsequent
election studies. Thus, an investigation of more recent data would allow
us to determine whether there are still significant variations in
political attitudes. Updating the analysis does not allow much for its
advancement but tests its continued applicability. As a result this would
be more of a mechanical manipulation than a conceptual advancement of the
work. The second reason for revisiting this work is the fit between theory
and data. Simeon and Elkins argue that provincial political cultures exist
first, because provincial political institutions create meaningful
cultural boundaries around citizens, and second, because of the varied
relationship between the federal government and voters across the country.
Although the updated analysis in 1980 incorporated perceptions of
different levels of government, thus far analyses of provincial political
cultures have conflated the notions of provincial culture and regional
reaction. With the exception of 1984, the Canadian Election Study has
never asked respondents about their views of provincial politics, the
supposed motor behind pan-Canadian variations. Thus, when survey
respondents identify their perceptions of the federal government, they are
responding to indicators that tap into the regionalized aspect of Canadian
political culture rather than the existence of provincial sub-cultures.
Responses for these individuals are then aggregated at the provincial
level, forcing a provincial analysis on to regional data.
The following analysis revisits Simeon and Elkins' original data
and then addresses the conceptual relationship between provincial
sub-cultures and regional variant cultures. It examines the existence of
regions within Canada by analyzing the social, demographic and economic
characteristics of federal election districts. In his assessment of the
renaissance of political culture William Reisinger claimed that future
research must seek to explain how sub-cultures relate conceptually to the
overall societal culture (Reidinger, 1995).
Elkins and Simeon, in a methodological piece on political culture, also
suggest that the “culture-bearing unit” must be identified
(Elkins and Simeon, 1979). For both these works,
sub-cultures require bounded communities, as would be the case with
provincial political culture. In contrast, regional variations in
political attitudes that do not conform to provincial boundaries reflect
the different experiences that voters have with the federal government.
There is a conceptual difference between asking individuals how they feel
about politics during a federal election—an exercise that would
heighten awareness of regional divisions—and arguing for the
existence of provincial political cultures as created by provincial
institutions. Simeon and Elkins argued, for example, that voters in
Atlantic Canada possess low efficacy and low trust. It is worth
determining if variations in attitudes can better be explained by province
or by region. At present, asking about Ottawa and dividing by province
does not allow us to determine whether both, one or neither is relevant.
This article, then, determines whether regional clusters can substitute
for provinces as the constituent units of political culture, without loss
of explanatory power.
Data and Analysis
The analysis relies primarily on two datasets. The first dataset is
the 2000 Canadian Election Survey (CES), which contains a rolling
cross-section sample of respondents who completed two waves of interviews
(before and after the election) and a self-completion questionnaire. The
questions have a considerable amount of continuity from the first election
study in 1965. Where applicable, data have been compared with earlier
years that relied on similar coding. In most cases this facilitates a
comparison between 1974, when 5-point Likert scales were used, rather than
earlier years. The second dataset consists of Federal Election Profiles
(FEP) for all federal constituencies, data for which are drawn from the
2001 census. The profiles offer counts for a variety of ethnic,
linguistic, racial, economic and demographic measures. These data were
recoded to represent percentages or rates rather than raw counts for each
constituency. Further manipulation, explained later in this section,
facilitated the sorting of constituencies into coherent clusters for
analysis. Each constituency in the FEB dataset was then assigned to a
cluster. To determine the relationship between a regional cluster and
political attitudes, each respondent in the 2000 CES was assigned the
cluster that corresponded to the federal constituency in which he/she
resided. Differences in political attitudes could then be tracked
according to province and to region. The use of both datasets allows for
an analysis of the differential impact of province or region on the
political attitudes of respondents.
Contemporary Data
Simeon and Elkins argued that significant differences among provincial
respondents pointed to the existence of provincial political cultures.
Central to this conclusion was an analysis of efficacy, trust and
political involvement data from the 1965, 1968 and 1974 election studies.
Table 1 compares the responses of provincial
respondents for the efficacy measures in 1974 and 2000. Data from 1974 are
used, as this year employed a Likert scale similar to the one currently
used in the CES. Before this year, respondents were asked if they agreed
or disagreed with a particular statement, rather than if they agreed, or
agreed strongly, disagreed or disagreed strongly. Furthermore, Table 1 only contains responses for the efficacy
questions, as trust variables in the earlier studies have not been
replicated in recent versions of the CES. With this in mind, Table 1 reports the percentage of individuals who
indicated that they agreed or agreed strongly with three measures of
political efficacy. Architects of institutional reform argue that change
is needed because voters are becoming increasingly frustrated with
politicians, and possess declining confidence and satisfaction, efficacy
and trust. As Table 1 shows, however, the
widespread decrease in efficacy that politicians have feared has not
materialized. Over the last thirty years the proportion of people who feel
that they have no say in politics has decreased in Canada, as has the
proportion of individuals who feel that politics is too complicated. The
proportion of respondents who indicate that it is not worth voting remains
unchanged in Canada. We know, of course, that one of the best predictors
of efficacy, formal education, has increased in the last thirty years.
While not the focus of this paper, changes such as these could account for
improved efficacy. These are, of course, not the only measures of
political dissatisfaction. In addition, the perception that voting is not
a worthwhile activity, stable since 1974, cannot account for the decline
in turnout from 1974 to the present. What we do know, however, is that by
these limited measures the proportion of ‘low-efficacy’
respondents in Canada has decreased over time. This in itself is
interesting, but for our purposes the differences among provinces prove
more revealing.
Table 1: Efficacy Scores in Provinces over Time
In 1974, provincial respondents offered widely varying responses. As
Simeon and Elkins noted, far more Atlantic residents expressed low
efficacy, while voters in British Columbia expressed more positive
responses. In the following thirty years, however, there appears to have
been dramatic changes within provinces and a levelling off among
respondents. In 1974, Quebec francophones, Newfoundlanders and New
Brunswickers respectively had the largest proportion of low-efficacy
voters according to the three questions listed here. Today, almost thirty
years later, Saskatchewan voters record the largest proportion expressing
low efficacy on two of the three questions, while a large proportion of
Newfoundlanders remain convinced that politics is too complicated for
them. At the same time, while British Columbians once had the lowest
proportion of low-efficacy respondents, this mantle has now passed to
anglophone and francophone voters in Quebec, and residents of PEI. In
fact, of francophone voters in Quebec, only 14 per cent believe they have
no say in politics. By contrast, the proportion of Canadians who feel this
way is 37 per cent. Provinces with previously large proportions of
respondents indicating minimal efficacy have seen a drop in their numbers.
Efficacy appears to have improved in most of the Atlantic provinces. At
the same time, provinces on the opposite end of the scale in 1974 also
appear to have had a change in attitudinal patterns. Ontario, once home to
a relatively small number of respondents expressing a lack of efficacy,
has seen a small rise in the number of voters who feel there is little
point in voting. Although the ranges in proportions appear similar across
time, the patterns of change are interesting. Within the Atlantic
provinces, some have seen a marked drop in the proportion of respondents
with low efficacy, while others have seen a dramatic rise. If culture has
an enduring influence on politics, and can explain the presence of
critics, supporters, disaffecteds or deferentials in any province, and if
Atlantic provinces possessed similar ‘deferential’ cultures in
1974, how can we account for opposing trends within categories? An
analysis of regional political cultures addresses this point.
Province or Region?
Students of political culture in the United Kingdom have long
attempted to prove the existence of national political cultures. In part,
the argument for devolution was based on the existence of different
approaches to policy and government in Scotland and Wales. Quantitative
research on the subject, however, has had very limited success in proving
the existence of constituent cultures in the United Kingdom (Henderson, 2001; Miller, Timpson et
al., 1996). In most cases, proof of difference is found only upon
stretching political culture as a concept to include voting preferences
for nationalist parties or national identity (Dickson,
1996; Brown, McCrone et al., 1998). More
often, research on political culture in the United Kingdom indicates that
the boundaries around the historic nations are less relevant than the
social and demographic characteristics of various regions (Curtice, 1988; 1992; 1996). John Curtice argues that there is neither a
distinct Scottish political culture, nor a Welsh or English political
culture, but northern and southern variations of political attitudes
fostered mainly by the different economic experiences of voters. For
Curtice, region feeds nationalism, a fact not disputed in most literature
on nationalism in Scotland (Brown, McCrone et al.,
1998; Mitchell, 1996).
Simeon and Elkins acknowledged the impact of socio-economic
characteristics and indicated that many of these characteristics were
distributed unequally across the country. Pointing to the diversity of
political attitudes within such provinces as Alberta and Saskatchewan,
however, they warned that “it is dangerous to assume citizens of
geographically proximate areas share the same attitudes” (Simeon and Elkins, 1974: 401). For them, not all
Albertans think alike, nor do all Westerners. This view does not preclude
the existence of regions that occur within and across jurisdictional
boundaries, a topic of research much more developed in American political
science.
In his analysis of American federalism, Daniel J. Elazar identified
three main political cultures in America: an individualistic culture, a
traditional culture and a moralistic culture (Elazar,
1972). For those in the individualistic type, politics and
government operate as a marketplace, while for the moralistic type, the
responsibility of government is to build a commonwealth in which all
benefit. The traditionalistic view falls between these two poles and tends
to advocate elitist or deferential approaches to politics. The different
groups were created, Elazar argued, by patterns of early migration. The
arrival of ethnic and religious blocks of voters, coupled with the pattern
of Western migration, led to pockets of homogeneity in which pre-existing
views of politics could influence American political life. Once cemented
in institutions, these pre-existing approaches continually reinforce
political attitudes. In this, Elazar's analysis fits well with
fragment theories of Canadian political culture. The analysis is based on
an intuitive understanding of the way American politics has developed.
Subsequent research has sought to improve upon Elazar's model,
subjecting it to a more rigorous statistical analysis (Clynch, 1972; Schlitz and Rainey,
1978; Savage, 1981; Sharkansky, 1969). One approach in particular is
useful for our purposes.
In his 1993 article Joel Lieske sought to
determine whether there was a more systematic way to identify regional
political cultures in the United States. To accomplish this goal he
created a database that contained aggregated data for counties in the
United States. The racial, ethnic, religious and linguistic profiles of
the counties were recorded alongside information on urban population,
population mobility, presence of certain age cohorts, employment in
various economic sectors, university education and income. The author then
employed cluster analysis to identify the existence of coherent groups of
counties and identified seven distinct clusters of counties in the United
States. In an effort to determine whether his method provided a better
explanation of variations in political attitudes in the US, the author
assigned respondents in the national election study to each of the
clusters based on their county of residence. He then ran a series of
analyses of variance tables to determine whether the clusters demonstrated
statistically significant differences in political attitudes. One of the
advantages of Lieske's approach is that it does not identify the
clusters based on responses to political attitudes, but rather it relies
on what he refers to as the constituent units of culture, namely
ethnicity, race, language, religion and social structure. As much of the
literature on political culture argues, voter attitudes are proxy measures
of culture. Individuals themselves do not possess political cultures
(Simeon and Elkins, 1980), nor regional variants
nor sub-cultures, but have attitudes and behaviours that allow us to
examine what the dominant political culture might look like. Relying on
the sources of political culture rather than its products allows for an
investigation that is conceptually cleaner. This method has been employed
in a Canadian context once before. In his 1990
article Robert MacDermid sought to identify different regional clusters
within Ontario. His research points to seven distinct clusters within
Ontario, which vary according to two axes: urban-rural and economic
activity. This reflects the indicators selected by MacDermid, which
include measures of income and employment, and single indicators for
language, ethnicity, religion and education (MacDermid,
1990).
In an effort to determine whether there are regional variants of
Canadian political culture, this article draws on a similar methodology.
In particular, the analysis relies on a dataset of all federal
constituencies and relevant demographic information from Statistics
Canada. The data draw on the profiles of federal election districts that
contain information on ethnicity, race, income, education, mobility,
population and employment. Rather than mimic exactly Lieske's
categories, which are better suited to the American migration patterns and
the particular racial profile of the United States, the following analysis
employs a list of indicators that better reflect variation among Canadian
constituencies. The cluster analysis relies on four main variables, with a
number of constituent indicators.
Ethnicity/Race/Religion: %
Québécois/Acadian, % British, % Scandinavian, %
Protestant Europe, % Catholic Europe, % Asian, % Black/African, %
Slavic, % Aboriginal origins, % Jewish
Language: % English mother tongue, % French mother
tongue, % bilingual, % Native languages, % Asian, % Scandinavian, %
Western European language, % Slavic, % sub-continent, %
Gaelic
Immigration: % immigrants, % old immigrants, % new
immigrants, % 3rd generation Canadians or older
Social structure: % over 65, % internal migrants, %
non-movers, % common law, % university degree, % managerial, %
professional, % natural resources, % women in labour force
These four variables are intended to capture the ethnic and linguistic
migration patterns that were considered to have an impact on Canada's
initial fragment cultures, in addition to social-demographic information
that currently influences political attitude variations. Ethnicity has
been coded so that it identifies those reporting Jewish ethnicity or an
ethnic background of a Protestant or Catholic European country.3
A more explicit test of religion produced
similar results. Income has not been included in large part because it
does not provide an accurate measure of wealth or poverty. The variable
cost of living in Canada could mean that someone earning $35,000 could
have widely varying levels of relative wealth in, for example, Iqaluit,
Toronto and Halifax.
Federal election districts were then
classified on the basis of common cultural characteristics, using cluster
analysis. The number of clusters was set at nine, in the expectation that
districts would be grouped according to the following categories: eastern
Canada, Prairies, western provincial north, southwestern Ontario, northern
Ontario, homogeneous Quebec, metropolitan Toronto, metropolitan Vancouver,
metropolitan Montreal. Such an arrangement reflects the vision of Canada
as a country of five regions, while recognizing the multicultural nature
of its larger urban centres. Results of the cluster analysis support
grouping the constituencies into nine relatively homogeneous clusters.
4 The analysis was repeated by setting the number
of clusters both higher and lower than nine. The nature of the analysis
remained similar. When forced to select a smaller number of clusters, the
urban, suburban and metropolitan clusters merged. When forced to select a
smaller number of clusters, the far north cluster fractured to distinguish
Nunavut from the rest of the territorial north.
The
characteristics of the clusters suggest the following descriptive labels
based on their geographic location:
- Cosmopolitan Quebec
- Suburban Toronto and Vancouver
- Urban Canada
- Rural and mid-northern
- Manufacturing belt
- New France
- British North America
- Far north
- Metropolitan Toronto
Some of these clusters exist entirely within a province while others
cross provincial and territorial boundaries.5
A full list of the constituencies within each cluster may be
obtained from the author.
As expected, the analysis produced
separate clusters for metropolitan Toronto and homogeneous Quebec
(labelled here as New France). Atlantic Canada is not a coherent region,
although many of the Atlantic constituencies appear in a cluster
characterized by large proportions of respondents with British ancestry.
This cluster has been labelled British North America as it contains areas
to which United Empire Loyalists emigrated both in the Atlantic provinces
and in Ontario. Other constituencies in Atlantic Canada are incorporated
into the manufacturing belt cluster, which also contains constituencies in
Ontario such as Sudbury and the ‘nickel belt.’ Metropolitan
Vancouver did not receive its own cluster although much of metropolitan
Montreal is contained within the cosmopolitan Quebec cluster. Instead,
many of the Vancouver constituencies appear along with other urban centres
in the urban cluster. This cluster contains urban constituencies, mostly
from the richer ‘have’ provinces, and unlike the others, it is
not geographically concentrated in one region. Suburban Toronto and
Vancouver include constituencies from the 905 area and Vancouver area
constituencies such as Burnaby. There are two clusters for the north. The
rural and mid-north cluster contains respondents from across the western
portion of the provincial north, including Ontario, Saskatchewan,
Manitoba, Alberta, British Columbia and the Yukon. The far north includes
the NWT and Nunavut, in addition to two larger northern constituencies in
the western provinces.
The typical vision of regional variations in Canadian political
culture is that they operate on an east-west basis. The cluster analysis
suggests that we must acknowledge the north-south dimension of Canadian
politics, not only as it relates to the territorial north but to the
provincial north as well (Coates and Morrison,
1992). This analysis also highlights the heterogeneous nature of
western Canada, which is often lumped together as a distinct entity. At
its most nuanced, the west often is treated as three distinct groups:
British Columbia, Alberta and the prairie provinces of Manitoba and
Saskatchewan. Here, however, the west contains elements of an urban
cluster, which includes cities such as Edmonton and Winnipeg, and a rural
cluster, which includes the elements of provinces across the west. In
addition, these are distinguished from suburban Vancouver, which has more
in common with suburban Toronto than with more proximate constituencies in
rural British Columbia. The characteristics of each cluster reinforce this
point.
The far north cluster contains a large proportion of native residents
and the British North America cluster contains a larger proportion of
English, Scottish, Irish and Welsh residents. Metropolitan Toronto has a
high mean score for immigrants from ‘old Europe,’ while
suburban Toronto and Vancouver contain more Asian and sub-continent
residents. The rural and mid-northern constituencies have larger mean
scores for Slavic residents, a fact that reflects the ethnic
characteristics of early migrants to the prairies. The New France cluster
contains far fewer constituencies of British stock. The size of each
cluster ranges from four, for the far north cluster, to the sixty
constituencies that comprise the New France cluster. To be useful,
however, the division of constituencies into regions must be accompanied
by an analysis of their impact on political attitudes.
In light of the comparison of efficacy indicators earlier, Table 2 contains the proportion of respondents who
agreed or agreed strongly, sorted by regional cluster. The regional
clusters possess varying efficacy responses that reflect, in many cases,
the answers in relevant provinces. The New France cluster, for example,
has a small proportion of respondents who feel they have no say.
Interesting differences appear in clusters that are geographically close
to each other. Suburban Toronto and Vancouver have a larger proportion of
low-efficacy respondents than metropolitan Toronto. Rural and mid-northern
constituencies have a larger proportion of low-efficacy respondents than
urban constituencies in ‘have’ provinces. The constituent
attitudes of any political culture include more than measures of efficacy.
The following analysis employs seven main variables. Attitudes to basic
responses to efficacy questions are included alongside groups of questions
probing perceptions on voting and views of political parties. Further
explanations of these variables are contained within the appendix.
Table 2: Efficacy Scores in Clusters
The analysis also includes an additive index of political activity,
counting respondents who have performed any of five protest behaviours.
Individuals who would consider acting in a similar manner were not
included in the index counts. An additional index tracks confidence in
public institutions such as the government and police. Last, the analysis
contains two measures of political ideology, one old, one of more recent
origin. Placement on a left-right spectrum is included, as is a test for
post-materialism. In sum, these seven variables were used to compare the
differences among respondents according to province of residence of
regional cluster.
Table 3 summarizes the analysis of variance
for province and regional cluster. There is a significant difference among
provincial respondents according to measures of efficacy, which confirms
the earlier findings of Simeon and Elkins, and significant differences
according to left-right ideology. There is also a significant difference
among the respondents when clustered by region. These differences are
summarized in the ANOVA F scores, which examine the variation between
groups, and the variation within groups. A significant F score indicates
that there are clear divisions among the categories of the variables.
Here, efficacy, left-right placement and political activity produce
significant deviations in results among clustered respondents. Results of
the multinomial logistic model further illustrate this point.
Table 3: Comparison of Province and Regional Clusters
Tables 4 and 5
contain the predicted probabilities of membership in the province or
regional cluster based on the seven political variables. The results
demonstrate first that political attitudes are able to predict membership
in regional clusters as well as they are able to predict province of
residence. The pseudo-R2s for both models are similar. We are
hoping for a McFadden's score greater than .2, which does not occur
in either model.6
A McFadden's
R2 below .2 would suggest that less than 20 per cent of the
variation of the dependent variable can be attributed to the variation in
the independent variable.
Second, the tables show that some
variables produce statistically significant estimates of membership. The
betas in multinomial logistic regression report the maximum probability of
observing the values of the dependent variable that were observed, given
the values of the independent variables. These predicted probabilities are
recorded as the natural logs of the odds ratio. In other words, as
efficacy increases by one unit the probability of correctly predicting
residence in Quebec increases by a multiplicative factor of .62. Within
the regional model, as left-wing beliefs increase by one unit a respondent
is four times more likely to be classified as residing in cosmopolitan
Quebec. The tables also indicate the percentage of cases in any grouping
that were correctly predicted by the model. The proportion of cases
correctly predicted for both models appears similar, at 37 per cent for
the provincial model and 30 per cent for the regional model. The preceding
analysis suggests that further research should more fully explore the
existence and influence of regions in generating variations of political
attitudes and behaviour across Canada.
Table 4: Predicted Probabilities of Provincial
Membership
Table 5: Predicted Probabilities of Cluster Membership
Existing research on political culture highlights the importance of
regionalism in Canada, something that this article confirms. From earlier
works exploring the relevance of migrant groups in New France and British
North America, to more recent evidence on varying levels of efficacy
across the country, such research emphasizes both the existence of
east-west variations, and provincial political cultures. The first
significant finding, then, is that regionalism in Canada is not an
exclusively east-west phenomenon, nor does the role of provinces appear to
be of unparalleled influence, as previously thought. By noting that
regional clusters can equally account for variations in political
attitudes and behaviour across Canada, this article clearly notes that
regionalism in Canada does not conform to provincial boundaries. The
existence of multiple cultures within the west, and the identification of
two distinct northern clusters suggest that the west is much less
homogenous than once assumed, and that the long-ignored north is not only
relevant, but diverse. The analysis notes also the existence of not one
but three distinct urban clusters, something suggested in MacDermid's
1990 research on Ontario. This in itself
suggests that the rural-urban dichotomy should in future distinguish among
those living in large metropolitan centres, suburban areas, mid-sized
urban towns and more rural areas.
Perhaps most significant, the research questions the importance of
political institutions in the generation of assessments of their
performance. Previously, regional political cultures in Canada have been
seen as the products of institutions. Provincial legislatures and the
policy they generate could account for much of the variation in attitudes
to politics, the remainder of which was attributed to the different ways
in which the Canadian Parliament and its legislation affected citizens in
different portions of the country. If regional clusters perform as
effectively as provincial boundaries in accounting for variations in
political attitudes and behaviour, then there is reason to doubt, at the
very least, the impact of provincial institutions. We should be cautious,
though, in our assessments of provincial impact, if only because
indicators directly probing efficacy and trust in provincial politics have
long been excluded from the Canadian Election Study. Clearly, though,
provincial boundaries are not the only way to account for variations in
political culture across Canada. That these clusters exist within and
across provincial boundaries points to the greater need for research into
the interaction among socio-demographic variables, provincial boundaries
and attitudes to the federal government.
Conclusion
This article performed two tasks. First, it examined variations in
provincial attitudes to federal politics in light of recent data in the
CES, and demonstrated that the number of low-efficacy respondents is
falling and that inter-provincial differences appear stable over time.
Second, it demonstrated the existence of regional clusters that could
account for variations in attitudes to federal politics. Regionalism in
Canada, usually seen as an east-west phenomenon, also contains an
important north-south dimension. In addition, indicators of political
culture, including perceptions of parties, voting, efficacy, left-right
beliefs and post-materialism, can predict membership in regional clusters
as successfully as they can residence in provinces.
In its exploratory investigation of the current state of provincial
political cultures in Canada, this paper sought to clarify whether
provincial political institutions or regional clusters could better
account for variations in federal political attitudes. Additional research
would help to further clarify the relationship between province and
region. Such research could examine the predictors of provincial political
attitudes, re-examine the measures of political culture at the provincial
level and determine the role of regional clusters in driving attitudes to
provincial politics. The data explored here do not disprove the existence
of significant provincial variations, but suggest that greater attention
should be paid to the role of regions in the generation of political
attitudes. It would appear that what we have been measuring thus far in
political culture research is not the existence of provincial
sub-cultures, but that of regional variations. The structure of any
existing sub-cultures remains an under-explored area of research.
Acknowledgments
The author would like to thank Neil Nevitte and Steve White and the
Journal's anonymous reviewers for helpful comments and
suggestions.
