![Sample our Urban Studies journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110826/TOC-Subject-Sample-2021-Urban-Studies.jpg"})
Abstract
Recent research conducted in both the United States and Canada has found that residents of inner cities and suburbs are diverging in their voting behavior and political attitudes. The mechanisms producing such a divergence, however, have remained unclear. After identifying a set of distinct hypotheses for why one might expect residents of inner cities and suburbs to differ in their political views, this article draws on a survey undertaken by the author in one electoral district in the Toronto region to empirically test the relative contribution of each of the hypothesized mechanisms in explaining the geography of party preferences. This study suggests there is no single explanation for the city-suburban cleavage, and that the mechanisms producing it are complex. Spatial segregation (based on individual attributes such as race, ethnicity, and class) is clearly important; however neighborhood self-selection, local experience, and, to a lesser extent, mode of consumption all have significant independent effects. Particularly important is the self-selection of supporters of political parties on the left into the inner city, stemming either from a search for a “sense of community” or the desire to link their lifestyle choices to their political convictions, whereas supporters of parties farther to the right are more likely to choose postwar suburban neighborhoods out of a preference for private space. In contrast, there is little evidence that housing tenure or the sharing of political information between neighbors are factors independently producing city-suburban political differences within the study district.
Acknowledgments
This research was facilitated by a doctoral fellowship from the Social Sciences and Humanities Research Council of Canada (SSHRC). Thanks to the anonymous reviewers for their helpful comments and to Audrey Kobayashi for her editorial suggestions.
Notes
1. The Progressive Conservative (PC) party of Canada formally dropped the word “progressive” from its name after it merged with the far-right Canadian Alliance party in December of 2003, leaving the “new” Conservative party the main federal party of the right.
2. Geographic differences in the scope and usage of local media may also produce unevenness in local opinions and voting ( 11 Books and Prysby 1991 ). Some forms of media are national or international in scope (national television, Internet); others, including local newspapers, service a geographically-defined area, such as an urban region or a particular municipality (i.e., central city), or even a local neighborhood. Media that cover a wide geographic area may nonetheless contain information primarily relevant to people living within particular districts or spaces, or contain cues to a political understanding of events occurring within voters' neighborhoods. It is not easy to see how city-suburban differences might be structured through this mechanism, particularly in the absence of a specialized media targeting one or the other zones. This area remains ripe for future study.
3. Party mobilization strategies are out of the scope of this research because the survey was undertaken in only one electoral district (thereby preventing cross-constituency comparisons). This strategy was followed purposefully in order to control for the effects of uneven targeting of constituencies by the political parties.
4. In 1997, the Ontario provincial government passed legislation that automatically remaps the provincial electoral districts to the federal constituency boundaries after each new federal redistricting exercise.
5. Respondents were asked which political party they usually supported in federal and provincial elections. If no party was given, respondents were then asked which party they were closest to at the moment, and if again if no response was forthcoming, which party they would vote for if an election were to be held today. This method assigned a party preference to approximately 80 percent of respondents. The remaining respondents either did not vote or refused to answer the question.
6. This method was chosen because it provides the simplest and most accurate picture of the maximum potential contribution of the spatial distribution of each variable to the spatial distribution of votes. Note that this method is followed for each variable separately, and cannot control for the simultaneous partial contribution of other variables. It is thus possible that this method overestimates the importance of each variable/mechanism, but it will not underestimate. The method is as follows: First, for each variable, three-way contingency tables are constructed and compared against both the zonal distribution of the total population for each set of party preferences, and the (actual) average zonal variation in the vote, the latter calculated as the weighted (by survey population) zonal deviation from the district mean in the support for each set of provincial and federal parties. A weighted (by zonal population) average zonal deviation score is calculated for each set of party preferences, representing the “actual” distribution. New contingency tables (by zone and party preferences) are then built by estimating the value of each cell in the table if the variable in question were to be zonally distributed in line with the total (survey) population. This is the “expected” distribution. From this expected distribution, a new weighted-average zonal deviation score is calculated. The degree to which each variable contributes to the zonal variation in each set of party preferences is the proportionate difference between the expected and actual deviation scores. Take, for example, the mechanism of direct self-selection due to political convictions. Three-dimensional contingency tables were built showing, for each set of party preferences and each zone, the number of people in each cell who said their political convictions were and were not behind their move to their current residential location. This is the original or actual distribution. As this distribution is highly skewed toward both NDP supporters and the first zone (inner-city Toronto), the pattern does not match the contingency table for the total population. The degree to which direct self-selectors were more or less prevalent in each cell (a cell location quotient) was then calculated (as the ratio of the cell proportion to the total proportion, divided by the ratio of cell's total [survey] population to the district total population), and compared to the table for the total survey population. A redistribution factor was then calculated and applied to each cell that corrected for the over- and underrepresentation of self-selectors across the zones (but not across party preferences), producing a new expected distribution of self-selectors in each cell. From this new contingency table, a new weighted-average zonal variation score (deviation from the sample mean) in party preferences was calculated for each set of party preferences and compared to the original score. For federal and provincial NDP supporters, the original (actual) score is 0.61541 and the expected score (if self-selectors had been zonally distributed in accordance with the actual population) is 0.46362, a difference of 15.18 percent. Thus, it is estimated that the zonal distribution of direct self-selectors explains up to 15.18 percent of the zonal distribution of those survey respondents who supported the NDP at both the federal and provincial levels.
7. The 1996 census shows that tenants make up 49.4 percent of the constituency population; however only 19.2 percent of the survey respondents are tenants. Tenants are underrepresented partly due to the random method for selecting streets. A significant portion of the district's tenants are concentrated in a small number of high-density apartment districts. The largest is the Crescent Town complex. These streets were not selected for interview. Furthermore, the surveyor was not granted entry to all the apartment buildings that did come up for interview. Thus, although a small number of apartment buildings did get surveyed, the majority of tenants in the sample lived in duplexes, rooming houses, or basement apartments. The underrepresentation of tenants is likely responsible for the underrepresentation of low-income earners, unemployed, and recent immigrants in the sample. The underrepresentation of recent nontraditional immigrants can also be explained by the higher refusal rate of non-English speakers. The results reported here are thus less valid for such groups.
8. Prior to the 1990s, two separate electoral districts covered the territory of what is now the Beaches–East York constituency. Until the 1960s, the southern inner-city Toronto portion consistently elected candidates from Canada's main party of the right, the PC party. Between 1962 and 1988, the NDP was able to win the seat in all but one election (1979) at the federal level and held it since 1975 at the provincial level. On the other hand, the northern portion, the constituency of York East, fluctuated between the Liberals and PC candidates for most of its history, tending to be one of the “swing” ridings that helped elect the government of the day. The Liberals captured the amalgamated seat federally in 1993 and have held it since, and the NDP have held onto it at the level of Ontario provincial politics over this period.
9. The 1997 results are shown in Figure 2 because polling station addresses were not readily available for the 2000 federal election. The geography of the vote, however, changed little between elections ( 7 Blais et al. 2002 ).
10. At the federal level, the major difference between the election results and the survey proportions is lower support in the latter case for the Canadian Alliance (which changed its name from The Reform Party shortly before the election), with the slack apparently going to both the PC and Liberal parties. The 2002 survey followed the significant fall from grace of Stockwell Day, then leader of the Canadian Alliance, and the party clearly suffered in popularity as a result. Provincially, the survey registered fewer NDP supporters than would be expected given the 1999 provincial election results, with most of the difference going to the provincial Liberals. This may reflect sampling error (perhaps related to the underrepresentation of tenants in the sample; see Note 8) or a shift in support from the provincial NDP to the Liberals. The more closely matching federal results (and the subsequent Liberal landslide in the 2003 provincial election) suggest that the latter explanation warrants some consideration.
11. It is common in Canada for voters to support different political parties in federal and provincial elections. Even for parties with the same moniker (Liberal party, etc.), the provincial wings often differ substantially in policy terms from their federal cousins. Additionally, it has been suggested that some Canadian voters prefer to offset the power of the governing party at the federal level by electing opposing candidates at the provincial level, though this has been the subject of debate within the literature ( 9 Blake 1985 ).
12. Furthermore, those choosing their residential locations out of desire for a sense of community were more than seventeen times less likely to say they supported the PC party at both the federal and provincial levels, whereas no federal NDP supporters said that their fondness for the house was the motivating factor for their residential location. However, the addition of these variables to the model alters only slightly the impact of zone of residence and mode of transportation on the outcomes, suggesting that residential choice has an effect that is independent and potentially weaker than either of these variables. Meanwhile, federal Liberal party supporters were not differentiated in their residential choices.
13. For those instances in which respondents did not see their experience as linked to attitudes on the left or right but their answer clearly indicated movement in the direction of one or the other based on typical conceptions of left and right in the Canadian context, the results were then coded as pushing respondents to the left or right, respectively. For example, in response to the question “How has living here influenced your political views?” one respondent replied “I have made up my mind after seeing how people live around here. … I see too many teenagers and foreigners loitering around … they're moving in too many immigrants” (Zone 3 resident, federal Liberal/provincial PC supporter). This respondent was coded as being moved to the right, despite not volunteering that his or her position was affiliated with either. This method was necessary for approximately 20 percent of the answers to this question, as a limited number of respondents both understood the terminology of left–right and at the same time saw their experience as falling into one of these categories. This method of reinterpreting respondents' answers obviously introduces the biases of the coder into the results. Only responses that were at the appropriate local scale were included in this analysis.
14. Ideally, it would be possible to determine the partial contribution of each mechanism to spatial unevenness (i.e., to simultaneously control for each effect). Testing of potential models of partial determination, however, produced widely varying and thus unreliable results, so the more simple bivariate estimates are reported here. The author would be interested in hearing from spatial analysts applying models of partial determination to similar spatial contexts and problems.
15. This points to the potentially “tectonic” form of social relations and interaction between such groups and the surrounding communities, as has been claimed for middle-class gentrifiers elsewhere in Toronto and abroad ( 66 b72 Robson and Butler 2001; Slater 2004 ).
16. The residential choices of those on the right may therefore appear natural and uncontested to them, particularly in homogenous suburbs where there may be fewer experiences impinging on a doxic mode of understanding. “Doxa” is the term 12 b13 Bourdieu (1977, 1984) uses for a high degree of fit between an individual's objective positions and her or his subjective beliefs, such that the social world appears self-evident, natural, and nonarbitrary, and thus is taken for granted and unquestioned. Doxic experiences and beliefs are thus naturalized “commonsense” ones, in contrast with the orthodox and the heretical, defined as they are in relation to contested (dominant and alternative) claims for legitimacy. Thanks to Deb Cowen for further insights in this and related areas.