GENDER AND THE ELECTORAL OPPORTUNITY STRUCTURE IN THE CANADIAN PROVINCES

POLITICAL RESEARCH QUARTERLY, March, 1998

 An earlier version of this paper was presented at the Midwest Canadian Studies Association meetings in, Madison, Wisconsin, October 11-13, 1996. Support for the research was provided by the Canadian Faculty Research Program through the Canadian Embassy, Washington, D.C. Further assistance was provided by Deborah Wituski, Henry Rieke, George Krause, and Al Olivetti, all at the time at West Virginia University.

ABSTRACT

Gender and the Electoral Opportunity Structure in the Canadian Provinces

  Richard E. Matland and Donley T. Studlar

We use multivariate analyses to test hypotheses concerning the electoral opportunity structure for women across a twenty year period of Canadian provincial elections. We find that party, political context, and social variables affect the likelihood that a woman is elected to a provincial parliament. While similarities between U.S. state legislative elections and Canadian provincial elections are found, there are distinct differences across the two polities, especially concerning where women first made inroads in winning representation. While women first gained a beachhead in small amateur legislatures in rural states in the United States, in Canada they first won significant numbers of seats in metropolitan areas. We find there continues to be great differences across riding types with women doing much worse in rural ridings than either urban or metropolitan ridings. The implications of these differences for redistricting are considered. Canadian courts have generally been sympathetic to plans that insure representation of geographic "communities of interest", even when this has meant overrepresentation of rural areas and underrepresentation of urban areas. We argue that a consequence of this policy is that Canadian provinces risk underrepresenting women, a nonterritorial "community of interest".

Most analyses of representation by gender in democratic polities in recent years have been concerned with central-level legislatures. Although there is an abundant literature on political recruitment and representation by gender for state legislatures in the United States, there is relatively little on sub-central levels for other countries (see Darcy, Welch and Clark 1994; Welch and Studlar 1990; Raaum 1995). Analyses of representation by gender in the Canadian provinces usually have been province-specific and limited to a single point in time. Existing data analyses have tended to only consider one or two explanatory variables. This paper provides the first comprehensive, longitudinal, multivariate comparison of representation by gender in all ten provinces of Canada. We focus on the question of what conditions, political and social, facilitate the election of women candidates and how these conditions have changed over time.

THEORETICAL PROBLEM AND APPROACH

Political recruitment can be conceived of as having a supply and a demand side. Individual characteristics such as social background, resources, and ambition affect the supply side. A number of conditions influence the demand for candidates to run for office. The electoral opportunity structure refers to those long-term conditions external to individuals that affect the demand for candidates and affect candidates’ ability to achieve official positions (Schlesinger 1966; Carroll 1994; Norris and Lovenduski 1995; Welch and Studlar 1996). Among the factors that define the electoral opportunity structure are the party context, the political context, and the social context (Norris 1993). The party context refers to internal party conditions such as party rules about the representation of groups when fielding candidates or general party ideology that can affect which candidates are seen as attractive. The political context refers to structural factors such as the number of offices, the level of competition for office, turnover rates among existing office holders, and the electoral system that can affect the electoral opportunities for candidates. Finally, the social context refers generally to the political culture, social values and attitudes that affect the demands for various characteristics of candidates. Electoral opportunity structure does not refer to short-term campaign factors that can affect individual outcomes such as an effectively run campaign or positions on campaign issues.

While the electoral opportunity structure is fairly stable in an advanced industrial democracy such as Canada, it can change through social evolution, because institutional rules are changed, or in response to demands for increased access by an effectively organized insurgent group. For instance, in several countries the electoral opportunity structure has changed through internal and external pressure being placed on political parties to increase representation through the introduction of affirmative action and quota policies for women candidates (Norris 1994; Matland 1995; Davis 1997). This paper examines the electoral opportunity structure by gender for Canadian provincial legislatures over a 20-year period in which women moved from being token members to a more substantial, though still minority, position in the provincial assemblies.

Although comparative provincial politics, especially regarding issues of women’s representation, has received increased attention in recent years (Domalgaski 1986; Moncrief and Thompson 1991; Moncrief 1994a; Studlar and Matland 1996; Arscott and Trimble 1996), some fundamental questions remain inadequately explored. Studies of women’s recruitment to legislative office in Canada, on both the provincial and federal levels, have emphasized party, party competitiveness, turnover, and political culture as explanatory factors (Brodie 1977; 1985; Vickers and Brodie 1981; Domalgaski 1986; Moncrief and Thompson 1991; Pelletier and Tremblay 1992; Studlar and Matland 1994; 1996; Brodie and Chandler 1991; Bashevkin 1991; 1993; Young 1991; Erickson 1991; 1993; Arscott and Trimble 1996). This paper considers an expanded list of variables that may affect women’s chances to become legislators. While mainly testing hypotheses from previous literature on gender representation in the Canadian provinces, the paper will also incorporate research findings from state legislatures in the United States and from the Canadian House of Commons. An important goal of this research is to look for dynamic elements, to see if conditions affecting women’s representation in the Canadian provinces change over time. Longitudinal studies in the United States have found that some early findings failed to hold up as women increased their share of seats in the state legislatures (Nechemias 1987; Rule 1990; Darcy, Welch and Clark 1994). We will look for similar results in the Canadian provinces. In comparison to previous empirically based research (Studlar and Matland 1996) this research focuses on what influences electability rather than candidacy.

VARIATIONS IN THE ELECTORAL OPPORTUNITY STRUCTURE

As noted, the electoral opportunity structure is defined by the party context, the political context, and the social context. While no party in Canada has enforced strict quotas for female candidates, all parties have made attempts at recruiting more female candidates (Erickson 1993). Of the parties, clearly, the New Democratic Party has been the most active in developing inclusionary policies. The NDP has promoted many policy proposals with feminist elements, has had many women in leadership positions, and has been the most aggressive in trying to insure that women are seriously considered as possible NDP candidates in every riding. These policies are consistent with the NDP’s leftist ideology and are an important part of NDP’s profile. Not surprisingly, the NDP is the party with the highest proportion of women among its nominated and elected candidates (Moncrief and Thompson 1991; Studlar and Matland 1996). Where the NDP does well, women also do well.

The political context is made up of several elements that are part of the electoral opportunity structure. One element that has been investigated closely is whether a party’s competitiveness in a district affects whether women are nominated. The suspicion was that women were more likely to be nominated as sacrificial lambs, i.e., that women could be expected to win nominations largely in districts where the party had little chance of doing well (Brodie 1985; Bashevkin 1993). A careful empirical analysis found that while in the 1970s women were more likely to be selected for seats where the party could not expect to do well, this effect disappeared by the 1980s (Studlar and Matland 1996).

More generally, an important factor in the political context is how much competition there is for an elected position and against whom one must compete. One measure of competitiveness is whether an incumbent candidate is in the race. As legislators have been historically male, incumbents running and winning are expected to hurt women’s access (Darcy, Welch, and Clark 1994). The desirability of the job of elected representative also affects how much competition there will be. Early research on female representation in U.S. states found a strong negative relationship between prestige, as measured through size of constituency, salaries, and amenities, and women’s representation (Werner 1968, Diamond 1977; Hill 1981; Welch 1978). Women were best represented in "amateur legislatures," those where being a representative was not a full time job and where prestige was quite low. When the pay and prestige are low less competition exists for legislative positions. The presumption behind this line of reasoning is that as positions become more attractive more men, and especially men with significant resources, will seek these positions and it will be more difficult for women to win.

A third element of the political context is the electoral system. While all Canadian provincial elections presently use a single member district simple plurality system, there existed until the mid 1990s ridings that had two members. When a party has to nominate two candidates, they often show a concern for balancing their tickets in the hopes of attracting more voters (Matland and Brown 1992). One might expect this would result in more women being nominated and elected in dual member ridings than in single member ridings.

At the broadest level, the social context is a crucial part of the electoral opportunity structure. Different ridings are expected to have different social contexts. Preliminary tests at the provincial level find that, when tested as a dichotomous variable, urban ridings elect women at higher rates than rural ones (Brodie 1977; Moncrief and Thompson 1991). It is not urbanization itself that should lead to better representation; rather urbanization serves as a proxy for other variables that effect the social context. First, urban areas have higher levels of labor force participation for women. Participation in the labor force has a strong positive effect on political activism. Second, average levels of education are higher in urban areas. This should have a positive effect, because education teaches tolerance and there should be a greater willingness to accept women as candidate. Education is also likely to affect the supply side as more women have the necessary educational and professional background that parties typically seek in candidates. Third women should do better in urban ridings because these ridings will have higher concentrations of highly educated and professional women who can form the core of women’s interest groups. Both politically oriented interest groups and professionally oriented groups serve as recruiting grounds for women candidates and serve an important function in providing backing for female candidates.

Regional and provincial political cultures are an additional part of the social context expected to affect representation levels. Although the relative importance of different levels of political culture is a much-discussed topic in Canadian politics (Dyck 1996; Landes 1995), previous studies of representation by gender have found provincial and regional differences in women’s access to elected office. The Western provinces of British Columbia and Alberta have tended to elect more women, while fewer women have been elected in the Atlantic provinces, especially Newfoundland and Nova Scotia (Brodie 1977; Vickers and Brodie 1981; Domalgaski 1986; Arscott 1996; Crossley 1996; Studlar and Matland 1996). By the late 1980s and into the 1990s these differences were large enough that Atlantic Canada had significantly lower levels of representation than the rest of Canada while the West was not far enough above other regions to be significantly different in a statistical sense (Studlar and Matland 1996).

Several scholars have argued that the political culture of the Atlantic provinces makes it harder for women to gain access there than in other Canadian provinces (Brodie 1977, Arscott 1996). Adamson and Stewart (1991) describe the political culture of the Atlantic provinces as traditionalist. In analyzing the low levels of women’s representation in Newfoundland and Nova Scotia, Arscott (1996) emphasizes the traditionalist culture of the provinces and the parties as an important part of the explanation. She argues the elites within the parties have been relatively unwilling to increase access for women. The traditionalist bent in internal party politics is seen in the Atlantic provinces tradition of sons following their fathers into party positions (Brodie 1977). Atlantic party organizations are more centralized, traditional, and patronage-based than those in the other provinces (Dyck 1996; Noel 1976). These characteristics can all be expected to work against incorporating a recently-aspiring group such as women.

In summary, a number of contextual factors at the party, political, and social levels define the electoral opportunity structure for women, and they can be expected to vary across ridings. Several of these have been tested individually, but they have not been subjected to simultaneous testing. Neither has there been systematic testing across time. It is to those tasks we now turn.

DATA AND METHODS

Data on candidates and elected MLAs were gathered, primarily using the Canadian Parliamentary Guide, supplemented by material from Canadian Censuses of 1981, 1986, and 1991, for all parliamentary elections in the ten provinces between 1975 and 1994 (see Appendix A for a complete list of the elections included). More than 11,000 candidates received at least 5 percent of the vote in their riding, with women constituting 14.7 percent of these. Of 3,755 victorious MLAs in this time period, women constituted 384, or 10.2 percent. The Progressive Conservatives won 37.1 percent of the seats, the Liberals 31.6 percent, the New Democrats 18.5 percent, the Parti Québécois 7.5 percent, Social Credit 4.4 percent, and other parties 1.0 percent.

(Table 1 about here.)

Table 1 shows the basic contours of representation in the ten Canadian provincial legislatures at the beginning (1975), middle (1984), and end (1994) of the period our data cover. As can be seen from Table 1, the Western provinces have tended to lead and the Atlantic provinces to lag in women’s representation. All provinces have had increases in the percentage of female MLAs over these two decades, and, except for Newfoundland and Nova Scotia, these increases have been substantial. The regression equation in Table 1 indicates that the percentage of women MLAs in a province in 1994 is a function of how many there were in 1975. If the regression coefficient for percent women in 1975 had been exactly 1.00, it would mean the absolute difference between provinces that were on the high and low end of the scale had remained constant over the 20-year period. The fact that it is greater than 1.00 means the gap between those who were high and low in 1975 has increased over this 20-year period. In Canada, as in the United States (Darcy, Welch, and Clark 1994), women have made the greatest gains recently in the same places they made gains initially.

We test for the effect of the party, political, and social context variables, described in the previous section, on women’s electability in a set of probit analyses. The unit of measure is individual provincial ridings and the dependent variable is whether the winning candidate in the riding election is a woman (1=female, 0=male). To test for party differences in advancing women, we create an NDP party dummy that equals one if the NDP wins the seat.

To test the political variable, level of competition, we include a dummy variable that indicates whether an incumbent is running in the riding. We further want to test if women in Canada first make inroads in the less prestigious legislatures where competition was less fierce, as they did in the United States. While legislative prestige may be easy to describe, it is more difficult to operationalize (Moncrief 1994a). We have taken two variables relevant to the concept, standardized them, and added them to form an additive scale of legislative prestige. The first variable used is legislative pay, adjusted for changes in the consumer price index. The second is the average number of constituents per MLA in a province. Legislative pay measures directly the economic attractiveness of being an MLA; average number of constituents is a more indirect measure of the prestige associated with being an MLA. We assume the more people one represents, the more prestigious the position. Both variables have been used in U.S. state studies (Rule 1981; Nechemias 1987).

To test whether multimember districts lead to better representation of women, we include a dummy variable measuring whether a riding has two seats. The hypothesis is where dual mandates exist parties are more likely to balance their tickets with respect to gender.

Among the social variables we expect to affect women’s electability are education levels and labor force participation. Unfortunately data on education levels and labor force participation at the provincial riding level are not available. While we do not have individual riding data, we do know that women’s labor force participation rates and education levels are higher in urban areas, and we can test if urban districts are more favorable to women than rural areas. In addition, we wish to extend the existing work on urbanization by examining whether a distinction exists between metropolitan districts, i.e., ridings in very large metropolitan areas, and ridings in districts that clearly are cities, but are more moderate in size. Metropolitan ridings are assumed to be more advantageous for women than urban ridings because they are more likely to have interest groups supporting women, a larger pool of well qualified possible female candidates, and less traditional views on proper roles for women. We have created two dummy variables, one for metropolitan ridings in urban areas of greater than 500,000, and a second for urban ridings located in urban areas of between 50,000 and 500,000.

To test the "regional political culture" hypothesis, we create an Atlantic dummy variable, which equals one for ridings in the four Atlantic provinces, the region suspected of being most inimical to women’s political advancement. This multivariate analysis will indicate whether the earlier findings were due to a uniquely regional culture or whether other factors lie behind the laggard status of the Atlantic provinces. The chart below reviews the variables to be tested and their expected effects on women’s representation in the provincial legislatures.

Expected Effects on Women’s Representation in Legislative Assemblies

Variable	Expected Effect on Women’s Rep
NDP Victory	+
Incumbent Candidate	-
Prestige Scale	-
Multimember District	+
Urban	+
Metropolitan	+
Atlantic Province	-

MULTIVARIATE ANALYSIS OF WOMEN’S REPRESENTATION

We believe the electoral opportunity structure has changed over the twenty years we study. If this is true then pooling all twenty years into a single regression is inappropriate. On the other hand, because elections are generally held every four or five years and not annually, comparing successive cross sectional regressions for individual years would be flawed because each year would contain a very different set of provinces. We do not believe a single dramatic defining event (or even multiple events) can be pointed to as creating a marked change in women’s chances of election in this time period. Such events, if they existed, could be used to identify where we should split the sample. Rather, we suspect a gradual evolutionary movement has occurred. Given this state of affairs, an ideal would be to split the data into equal size blocks with each province represented once in each block. This is impossible, however, because election frequency varies across provinces. What we have done is split the twenty-year time period into five separate blocks. Although the blocks are not designed to reflect specific political eras, we did weigh the need to keep the time periods short, so that the years included could all be considered part of the same era, against the desire to include all provinces in each block. Ultimately we created five blocks with at least eight provinces represented in each block and 8-13 elections in each block.

Table 2 presents the probit results for each time period. The party variable produced results about as expected. Victories by the NDP have a positive effect on women’s representation in all periods, but the effect only reaches strong standards of statistical significance in the last time period (1990-1994) when the NDP hit record highs in several provinces and rose to power in Ontario, British Columbia, and Saskatchewan with a high proportion of women MLAS. We cautiously suggest that women’s fortunes are positively correlated with the fortunes of the NDP. The obvious reason is that the NDP has nominated more women than the other parties; when the NDP does well, women do well.

[Table 2 about here.]

The incumbency variable has a statistically significant effect in three of the five time periods. This confirms a fairly universal assumption that incumbency hurts women’s access. The failure to have a statistically significant effect in the mid 1970s and mid 1980s is likely to stem from different causes. In the mid 1970s women had not yet begun to increase as a percentage of candidates; the few women who ran were just as likely to be incumbents as challengers (Studlar and Matland 1996). From the late 1970s onward the number of women candidates has grown, but it has grown unevenly. The 1983-1986 period is a period of consolidation where the percent of women candidates did not increase significantly over the immediately preceding time period. Women were more likely to be challengers than incumbents, but not by much. An additional reason this variable is not significant in the mid-1980s is that it only measures whether an incumbent is running in the race. A considerable number of incumbents do go down to electoral defeat (Moncrief 1994b), and this mitigates the gender bias incumbency creates.

The prestige of the office was expected to effect women’s access negatively. Unlike the early studies in the United States, however, the prestige of a legislative position does not appear to have affected the likelihood women are elected to provincial parliaments. The prestige factor score never nears statistical significance and even fails to have a consistent effect, alternating between positive and negative effects.

Multimember districts have a very weak effect in the first three time periods but an extremely powerful effect in the last two. This delayed effect may have occurred because until the mid 1980s gender was not a significant concern for parties when composing their tickets. By the late 1980s, however, the issue of women’s representation had become important enough, women’s groups within local parties had become strong enough, and parties were sufficiently concerned about the issue that they began to balance their tickets in terms of gender (Crossley 1996). Consequently, a substantial increase in the number of women MLAs in two-member districts occurred. In other words, the electoral opportunity structure changed over time in response to societal pressures. Ironically, just as it emerged that women are helped by multimember districts, these ridings have now been eliminated in all provincial legislatures. The effects of this development are brought home starkly by the first provincial election held on Prince Edward Island after redistricting from dual member districts to all single member districts. After the 1996 PEI elections the percentage of women in the legislature plunged from 25 percent to 14.3 percent.

Surprisingly an effect failed to appear for the variable testing for a regional effect in the Atlantic provinces. This variable shows a small negative effect that never reaches statistical significance in any of the time periods. While it is clearly true that the Atlantic provinces have lower rates of representation for women, the explanation for these rates should not be attributed to a uniquely Atlantic political culture. Rather, the lower levels of representation in Atlantic Canada are likely explained through the lack of a viable NDP, the more rural nature of the provinces, and a slightly higher tendency for incumbents to seek and win reelection.

The final variable tested is urbanization. The results show that women have always done better in metropolitan ridings. In the mid-1970s women did substantially better in metropolitan ridings than they did in their rural or urban counterparts. From the mid-1970s until the late 1980s we see the gap between urban ridings, i.e., smaller and mid-sized cities, and the large metropolitan centers slowly diminishing. By the late 1980s and to the present women did as well in urban ridings as they did in metropolitan ridings. Women still did substantially worse in rural ridings. We return to some interesting implications of these findings later in the paper.

In summary, we find that riding characteristics strongly affect electability of women, with metropolitan districts having been historically the most advantageous. Running in a district where an incumbent was on the ballot decreased the likelihood that a woman would be elected in most time periods. In the later time periods being in a dual mandate riding significantly increased the probability of a woman getting elected. The NDP shows a modest effect, but women are helped by NDP wins. Legislative prestige and running in an Atlantic province have no significant independent effect on women’s representation.

An Alternative Analysis. While Table 2 provides information on which variables are statistically significant, probit coefficients are quite difficult to interpret. To get a better picture of the effects of the variables we repeated the probits, dropping variables that were never significant and including those that were significant at least once. We then used the revised probits to develop Table 3, which presents estimates of the likelihood that an elected MLA is a woman under various conditions across time. There are a dozen separate conditions described with respect to whether the competition includes an incumbent candidate, whether the NDP wins the race, and the characteristics of the riding shown in Table 3.

[TABLE 3 ABOUT HERE]

One immediately obvious finding is that the probability a woman will be elected has increased steadily over time under all conditions. In the mid-1970s the probability that an MLA would be a woman was less than 10 percent under all conditions. By the 1990s under all conditions, except one, the probability of an MLA being a woman was greater than 10 percent. The gains are especially dramatic in urban districts (conditions 2, 5, 8, 11) where the probability of an MLA being a woman is 14-34 percent higher in the 1990s than it was in the mid-1970s. Rural ridings have consistently lagged behind the metropolitan ridings and, in more recent years, the urban ridings also. We see that except for rural ridings where the NDP wins and no incumbent is running (i.e., condition 10) the gains across the twenty year period in rural ridings are modest, only in the 6-12 percent range (conditions 1, 4 and 7).

The most favorable conditions for women candidates are metropolitan and urban ridings where the NDP wins and there is no incumbent running. Under these conditions (conditions 11 and 12) in the last time period the probability of a woman MLA is approximately 38 percent. The probability of the MLA being a woman drops to about 26 percent when the winning party is not the NDP (conditions 8 and 9). It drops an additional 7 percent to about 19 percent when an incumbent is running in the riding (conditions 2 and 3). Finally it drops to only an 8 percent likelihood when the riding is a rural riding, when the NDP does not win the seat and when an incumbent is running in the race (condition 1).

The effect of incumbency can be seen by comparing conditions on the top half of the chart with their mirror condition on the bottom half (compare 1&7, 2& 8, 3& 9, 4& 10, 5 & 11, and 6 &12). In 1975-1978 and 1983-1986 incumbency has very little effect, but for the other three time periods women were much more likely to be elected when no incumbent was running. In time periods where an incumbent candidate running did have an effect, the effects went from a modest 4-5 percent to a more substantial 15 percent greater probability that a woman was elected if no incumbent was running.

Women also are helped by NDP victories; an MLA is approximately 10-12 percent more likely to be a woman when the NDP wins in urban or metropolitan districts in the 1990s (compare conditions 2 & 5, 3 & 6, 8 & 11, and 9 & 12) , and 6-8 percent more likely to be a woman when there is an NDP victory in rural districts (compare conditions 1& 4, 7 & 10).

REPRESENTATION, RIDING CHARACTERISTICS, AND REAPPORTIONMENT

We now wish to take a more intensive look at the issue of riding characteristics. The probits show that over this twenty year period significant differences existed in where women gained access. Different representation rates across riding types are also significant because they are relevant to the on going debate in Canada on provincial reapportionments. We wish to look at both issues more closely.

[FIGURE 1 ABOUT HERE]

Figure 1 provides a graphic presentation of how women’s representation has changed in the three riding types over the 20-year period. Figure 1 uses rolling three year averages, i.e., the 1977 result shows what percent of all seats contested in 1975, 1976, and 1977 were won by women. In the 1970s the metropolitan ridings clearly elected more women than either rural or urban ridings. The rolling average of women’s representation in urban and rural ridings remained below 5 percent until 1981, while in metropolitan districts representation was near 10 percent. In the 1980s representation in rural ridings rose modestly; the three-year rolling average went from 3.7 percent in 1980 to 9.6 percent in 1989. In only one year in the 1980s did women win more than 10 percent of the rural ridings that were at stake in an election. In urban ridings, however, much more dramatic movement occurred, with women’s percentage of winning MLAS rising from 4.6 percent to 18.5 percent. Metropolitan ridings exhibit a noticeable rise from 10.7 percent to 20.8 percent. In the 1990s results have been on a generally upward trend but also have been volatile. Some inroads have been made in rural ridings, where the three-year average has finally started to stay consistently over 10 percent. For both urban and metropolitan ridings, women have won at least 25 percent of the seats in three of the four election years. Nevertheless, for both riding types there have been bad election years that have caused the rolling averages to plummet. Perhaps the most striking feature of Figure 1 is that women’s representation in rural ridings continues to lag significantly behind urban and metropolitan ridings.

The probits and Figure 1 help explain the evolution of women’s representation over time. Initial gains were made in metropolitan districts and only later did women start to gain in other types of constituencies. The difference between the urban dummy and the metropolitan dummy is statistically significant in the first period and shows that it is not just cities, but more specifically in metropolitan areas that women first made inroads. This is unexpected since early U.S. literature emphasized that women had the greatest possibility of gaining access in rural districts where politics is less intense (Werner 1968; Diamond 1977). Our Canadian findings, on the other hand, suggest that women first made gains in exactly those districts where politics was most professionalized, most competitive, and where substantial electoral organization was needed. These ridings were, however, also those where women had the greatest resources available in terms of women’s organizations and significant numbers of professional women who might be seen as attractive candidates by parties. Suggesting that metropolitan areas are likely to be less traditional than either rural areas or small urban centers also seems reasonable. Over time, in the U.S. as well women have started doing better in these types of districts than rural districts (Darcy, Welch and Clark 1994; Rule 1990; Volgy, Schwarz and Gottlieb 1990).

The substantial differences in women’s representation across types of ridings takes an unexpected twist when we look at another controversial issue in provincial politics, redistricting. Canadian provinces do not require districts be drawn so equal numbers of voters are behind each representative (Dyck 1996). The controversial debate over how district lines are to be drawn occurs at two levels. At the level of ideas, competition occurs between the principle of "one person/one vote" and the principle of insuring representation for "communities of interest" (Small 1991; Courtney et al. 1992). The "communities of interest" argument states that a demand for perfect equivalency in constituents per riding will force map drawers to create, in rural areas, very large districts, often with very disparate constituencies. A possible consequence of forcing several different communities together is that some viewpoints or interests will fail to gain representation and not be heard. Therefore district lines should be drawn to insure all communities are heard, even if this leads to fewer constituents in some provincial ridings.

At the level of power politics, political parties see an opportunity to maintain their political power through gerrymandering such that those areas of the province where they are strongest have the greatest representation. Being allowed to draw constituencies of uneven sizes makes this even more tempting. The issue of what principles should be used come to the fore either when a substantial lag occurs since the last redistribution (Prince Edward Island is the runaway winner, going almost 100 years before redistricting) or when new redistribution proposals do not adjust for population shifts to insure equal number of citizens behind each MLA. Often population has moved toward urban centers and out of rural districts. When a party in power depends upon MLAs elected from sparsely populated rural ridings to provide its legislative majority, the party is loathe to redraw lines to provide ridings of equal size. Several times a party which finished second in total votes in a province has had the most seats in the Legislative Assembly and formed the government (Dyck 1996). This occurred most recently in British Columbia in the election of 1996.

We consider the apportionment question, and the consequences for female representation, by looking at the province of Alberta, which is fairly typical with respect to redistribution. Alberta has two major metropolitan areas, Calgary and Edmonton. These two metropolitan areas have approximately 1,394,000 people, or 52 percent of the total population of Alberta. Only two cities meet our criteria for being urban (i.e., above 50,000 and less than 500,000), Lethbridge and Red Deer. Combined they have 125,000 people, or approximately 5 percent of the population of the province. The remaining 43 percent of the population lives in rural ridings, but they elected 49 percent of the representatives in the Legislative Assembly in 1993. Instead of the 36 representatives they would have if the seats were apportioned solely based on population, they have 41 representatives. Calgary and Edmonton, on the other hand, with 52 percent of the province’s population, have only 46 percent of the representatives (38 seats rather than 43). This pattern is repeated throughout the country, with urban or metropolitan areas receiving less than their proportional share of seats, while rural districts receive more than their proportional share of seats. The significance for women’s representation is, as our probit analyses show, that women are more likely to win races in metropolitan and urban ridings than in rural ones. In Alberta women won 26.3 percent of the seats in Calgary and Edmonton in the 1993 election while winning only 14.6 percent of the seats in the rural ridings.

Baker v. Carr (1962) and Reynolds v. Sims (1964) largely decided the issue of legislative redistricting in the United States in favor of "one person, one vote." U.S. court rulings since then have required districts to be virtually identical in population. In Canada the operating principles for legislative redistributions remain more contentious. The Carter case was brought to the Supreme Court of Canada in 1991 by citizens protesting the drawing of district lines in the Representation Act of 1989 (an Act of the Saskatchewan Legislative Assembly) to consciously over represent rural (and Conservative) ridings. While the Saskatchewan Court of Appeals ruled in favor of the citizen group, the Supreme Court of Canada reversed the Saskatchewan Court of Appeals decision and upheld the initial redistribution plan, arguing that what was important was "effective, not necessarily equal, representation" (Dyck 1996:449). The court, in making this ruling, helps maintain a system that makes it more difficult for women to gain effective rather than merely token representation in the provincial legislatures.

CONCLUSION

Our analysis shows the electoral opportunity structure within ridings does make a difference in women’s chances of being elected. Our analysis also shows the effect of various elements of the electoral opportunity structure change over time. MLAS are most likely to be female in metropolitan (and urban) ridings, in districts where no incumbent candidate is running, and when the NDP wins.

Among the dynamic elements seen in our analysis is the change in how women have done in urban ridings and in multimember districts. Initial breakthroughs occurred in metropolitan ridings. Over this twenty year period there is a general upward trend, but also different rates of increase as urban districts "catch up" to metropolitan districts and both are lagged considerably by the rural districts. An optimistic perspective would predict that over time rural ridings will catch up with the urban and metropolitan ridings, just as the urban ridings caught up to the metropolitan ridings. A more sanguine perspective would note significant differences exist between these types of ridings in the size of the candidate pool, education levels, labor force participation rates, and women’s interest organizations. Those differences may well continue to work to the advantage of women in urban and metropolitan districts and to the disadvantage of women in rural constituencies.

The effect of multimember districts also provides a dramatic case of change. Over the twenty years there appears to have been a significant change in which factors parties looked at when balancing their tickets. Dual ridings were initially created in Prince Edward Island at the end of the last century expressly to insure ticket balance with respect to religion (Dyck 1996: 96). Looking at the time period our data covers, in the late 1970s and early 1980s parties nominating candidates in dual ridings did not appear to be concerned about gender when trying to balance their tickets. By the late 1980s, however, it appears that gender had become a very important part of balancing considerations. Women did much better in dual member ridings than they did in single member ridings. This finding is theoretically significant as it shows that increased number of seats can be advantageous to women within a majoritarian system and not just in proportional representation systems.

While the general trend has been in the direction of increasing representation, other tendencies hinder this movement. The demise of multimember districts is one negative factor. Another is the systematic underrepresentation of urban and metropolitan districts. Redistributions are usually seen as partisan issues, however, women as a group also have an interest in electoral districts that more closely follow the "one person/one vote" principle. Our results show that women face higher barriers in rural districts than elsewhere. A more egalitarian seat distribution would lead to more women being elected. In this case, the Canadian legal penchant for preserving geographical "communities of interest" in legislative districting leads to neglect of an emerging, non-geographical "community of interest," namely women.

In the interests of cumulative social science, we can compare our results with previous findings from U.S. states. Several variables performed similarly. History seems to matter in both countries. Canadian provinces and U.S. states where women have done well in the past tend to be where they continue to do well. Incumbency hinders women’s advancement at the sub-national level in both countries, too. Higher incumbent return rates in the United States, however, mean that this variable is likely to be more important in the U.S. (Moncrief 1994a, 1994b). Multimember districts also act similarly, having very little effect in early periods when few women were legislators but subsequently having a strong positive effect (Matland and Brown 1992).

Still, clear differences exist. Women in Canada have the advantage of a leftist party, the NDP, advocating feminist concerns, including greater representation by gender, on both the central and provincial levels. The NDP has even attempted to use quotas to nominate women on the federal level, a difficult situation in a single-member district system. This pressure affects the competing parties as well (Matland and Studlar 1996). In contrast, in the United States parties on either the federal or state level lack sufficient organizational control over nominations to implement such a plan.

The most striking contrast occurs, however, when looking at where women first made inroads into sub-national legislatures. In the United States, women were first elected in amateur legislatures and in largely rural states. This led to the suggestion that women were most likely to first gain access in low prestige legislatures and areas where politics was less competitive. In contrast, our results show that legislative prestige has never affected women's representation in provincial legislatures. Furthermore, it is in metropolitan areas, the very antithesis of where U.S. women first got their start, that Canadian women first began gaining representation. This lends credence to those who argue that the major impetus for women initially gaining better access to the legislatures in New England states was not the low prestige of the bodies but the moralistic political culture of those states (Nechemias 1987).

Cultural arguments, however, should be approached with caution. Our multivariate analysis shows that previous claims of a distinctive Atlantic political culture antipathetic to women’s representation were exaggerated. Political culture may be significant, but it needs to be carefully specified, and testing of other explanations should be exhaustive before political culture is invoked.

This study demonstrates the value of studying lower level political systems outside the United States to understand the conditions advancing and inhibiting greater women’s legislative representation. The electoral opportunity structure for women is somewhat different, even in two federal polities with similar social structures. Some factors, such as a tradition of electing women, incumbency, multimember districts, and, more recently, urbanization, seem to be constant across the two countries. Others, such as legislative prestige, urbanization (in the early stages), and the presence of a leftist party actively promoting women’s representation, suggest differences across the two countries. Women’s representation in the Canadian provinces is assisted by the presence of the NDP and high levels of legislative turnover but hampered by the end of multimember districting and the refusal of provinces to redistribute seats according to "one person, one vote." As demonstrated above, continued adherence to the preferred Canadian theory of territorial representation for "communities of interest," will retard women’s possibilities for gaining legislative seats. In short, the electoral opportunity structure has political consequences.

REFERENCES

Adamson, Agar and Ian Stewart. 1991. "Party Politics in Atlantic Canada: Still the Mysterious East?" In Hugh G. Thorburn (ed) Party Politics in Transition. 6th ed. Scarborough: Prentice-Hall Canada.

 Arscott, Jane. 1996. "Between the Rock and a Hard Place: Women Legislators in Newfoundland and Nova Scotia." In Jane Arscott and Linda Trimble (eds.) In the Presence of Women. Toronto: Harcourt, Brace and Company.

Arscott, Jane and Linda Trimble (eds.) 1996. In the Presence of Women. Toronto: Harcourt, Brace and Company.

Bashevkin, Sylvia. 1991. "Women's Participation in Political Parties." In Kathy Megyery (ed) Women in Canadian Politics Toronto: Dundurn Press.

 Bashevkin, Sylvia. 1993. Toeing the Lines: Women and Party Politics in English Canada. 2nd ed. Toronto: University of Toronto Press.

Brodie, M. Janine. 1977. "The Recruitment of Canadian Provincial Women Legislators, 1950-1975." Atlantis 2: 6-17.

Brodie, Janine. 1985. Women and Politics in Canada. Toronto: McGraw-Hill Ryerson.

 Brodie, Janine and Celia Chandler. 1991. "Women and the Electoral Process in Canada." In Kathy Megyery (ed) Women in Canadian Politics. Toronto: Dundurn Press.

Carroll, Susan J. 1994. Women as Candidates in American Politics. 2nd ed. Bloomington: Indiana University Press.

Crossley, John. 1996. "Picture This: Women Politicians Hold Key Posts in Prince Edward Island." In Jane Arscott and Linda Trimble (eds.) In the Presence of Women. Toronto: Harcourt, Brace and Company.

Courtney, John C., Peter MacKinnon, and David E. Smith (eds.). 1992. Drawing Boundaries; Legislatures, Courts, and Electoral Values. Saskatoon, Saskatchewan: Fifth House Publishers.

 Darcy, R., Susan Welch and Janet Clark. 1994. Women, Elections and Representation. 2nd ed. Lincoln: University of Nebraska Press.

Diamond, Irene. 1977. Sex Roles in the Statehouse. New Haven: Yale University Press.

 Domalgaski, Marilyn. 1986. "The Role of Canadian Women in Federal and Provincial Politics." In Robert J. Fleming (ed) Canadian Legislatures: The 1986 Comparative Study. Toronto: Office of the Legislative Assembly.

Dyck, Rand. 1996. Provincial Politics in Canada. 3rd ed. Scarborough: Prentice-Hall Canada.

Erickson, Lynda. 1991. "Women and Candidacies for the House of Commons." In Kathy Megyery (ed) Women in Canadian Politics. Toronto: Dundurn Press.

Erickson, Lynda. 1993. "Making Her Way In: Women, Parties, and Candidacies in Canada." In Joni Lovenduski and Pippa Norris (eds.) Gender and Party Politics. London: Sage.

Landes, Ronald G. 1995. The Canadian Polity: A Comparative Introduction. 4^th ed. Scarborough: Prentice Hall Canada.

Lovenduski, Joni and Pippa Norris (eds.) 1993. Gender and Party Politics. London: Sage.

 Matland, Richard E. 1995. "How the Electoral System Structure Has Helped Women Close the Representation Gap." In Lauri Karvonen and Per Selle (eds.) Closing the Gap, Women in Nordic Politics. London: Dartmouth.

Matland, Richard E. and Deborah Dwight Brown. 1992. "District Magnitude and Women's Representation in State Legislatures." Legislative Studies Quarterly 17: 469-492.

Matland, Richard E. and Donley T. Studlar. 1996. "The Contagion of Women Candidates in Single-Member District and Proportional Representation Electoral Systems: Canada and Norway." Journal of Politics 58: 707-733.

Moncrief, Gary F. 1994a. "Professionalism and Careerism in Canadian Provincial Assemblies: Comparison to U.S. State Legislatures." Legislative Studies Quarterly 19: 33-48.

Moncrief, Gary F. 1994b. "Turnover in Canadian Provincial Legislatures, 1960-1994." Paper presented at meeting of Canadian Political Science Association, Calgary.

Moncrief, Gary F. and Joel A. Thompson. 1991. "Urban and Rural Ridings and Women in Provincial Politics: A Research Note on Female MLAs." Canadian Journal of Political Science 24: 831- 837.

Nechemias, Carol. 1987. "Changes in the Election of Women to U.S. State Legislative Seats." Legislative Studies Quarterly 12: 423-442.

Noel, S.J.R. 1976. "Leadership and Clientelism." In David Bellamy, Jon H. Pammett and Donald C. Rowat (eds.) The Provincial Political Systems: Comparative Essays. Toronto: Methuen.

Norris, Pippa. 1993. "Conclusions: Comparing Legislative Recruitment." In Joni Lovenduski and Pippa Norris (eds.) Gender and Party Politics. London: Sage.

Norris, Pippa. 1994. "Labour Party Quotas for Women." In David Broughton, David Farrell, David Denver and Colin Rallings (eds.) British Parties and Elections Yearbook 1994. London: Frank Cass.

Norris, Pippa and Joni Lovenduski. 1995. Political Recruitment. New York: Cambridge University Press.

Pelletier, Rejan and Manon Tremblay. 1992. "Les Femmes Sont-elles Candidates Dans De Conscriptions Perdues D/Vance? De L'examen D'une Croyance." Canadian Journal of Political Science 25:249-267.

Randall, Vicki. 1987. Women and Politics. 2^nd ed. Chicago: University of Chicago Press.

Raum, Nina. 1995. "Women in Local Democracy" in Lauri Karvonen & Per Selle (eds.), Closing the Gap: Women in Nordic Politics. London: Dartmouth.

Rule, Wilma. 1981. "Why Women Don't Run: The Critical Contextual Factors in Women's Legislative Recruitment." Western Political Quarterly 34: 60-77.

Rule, Wilma. 1990. "Why More Women Are State Legislators: A Research Note." Western Political Quarterly 43: 437-448.

Schlesinger, Joseph R. 1966. Political Ambition. Chicago: Rand McNally.

Small, David (ed). 1991. Drawing the Map: Equality and Efficacy of the Vote in Canadian Electoral Boundary Reform. Toronto: Dundurn Press.

Studlar, Donley T. and Richard E. Matland. 1994. "The Growth of Women's Representation in the Canadian House of Commons and the Election of 1984: A Reappraisal." Canadian Journal of Political Science 27: 54-79.

Studlar, Donley T. and Richard E. Matland. 1996. "The Dynamics of Women's Representation in the Canadian Provinces: 1975-1994." Canadian Journal of Political Science 29: 269-293.

Vickers, Jill McCalla and M. Janine Brodie. 1981. "Canada." In Joni Lovenduski and Jill Hills (eds.) The Politics of the Second Electorate. London: Routledge and Kegan Paul.

Vickers, Jill, Pauline Rankin and Christine Appelle. 1993. Politics As If Women Mattered. Toronto: University of Toronto Press.

Volgy, Thomas, John E. Schwarz and Hildy Gottlieb. 1990. "Female Representation and the Quest for Resources: Feminist Activism and Electoral Success." Social Science Quarterly 71: 156- 168.

Welch, Susan and Donley T. Studlar, 1990. "Multi-Member Districts and the Representation of Women: Evidence from Britain and the United States." Journal of Politics 52: 391-412.

Welch, Susan and Donley T. Studlar. 1996. "The Opportunity Structure for Women's Candidacies and Electability in Britain and the United States." Political Research Quarterly 49: 861-874.

Werner, Emmy E. 1968. "Women in State Legislatures." Western Political Quarterly 21: 40-50.

Young, Lisa. 1991. ""Legislative Turnover and the Election of Women to the Canadian House of Commons." In Kathy Megyery (ed) Women in Canadian Politics. Toronto: Dundurn Press.

TABLE 1

INCREASE IN PROPORTION OF WOMEN IN PROVINCIAL LEGISLATURES

1975- 1994

Province	1975 %	1984 %	1994 %	% Increase 1975 -1994
Alberta	2.7	7.6	19.3	+ 16.6
British Columbia	10.9	10.5	25.3	+ 14.4
Manitoba	0.0	12.3	19.3	+ 19.3
New Brunswick	3.4	6.9	17.2	+ 13.8
Newfoundland	2.0	5.8	5.8	+ 3.8
Nova Scotia	2.2	5.8	9.6	+ 7.4
Ontario	5.6	4.8	21.5	+ 15.9
Quebec	0.9	6.6	18.4	+ 17.3
Prince Edward Island	6.3	6.3	25.0	+ 18.7
Saskatchewan	3.3	7.8	18.2	+ 14.9
Mean	3.7	7.4	18.0	+ 14.3
S.D.	3.2	2.3	6.1	+ 2.9

REGRESSION EQUATION:

Percent Female1994 =	13.66	+ 1.15* Percent Female1975
	(2.64)	(.55)

* t statistic = 2.08, significant at .07 level, two-tailed test.

Adjusted R²= .27
S.E. Estimate= 5.23
N=10

TABLE 2: RESULTS OF PROBIT ANALYSES, Five Discrete Time Periods, 1975-1994

Dependent Variable = Sex of Winning Candidate (0=male, 1 = female)

YEARS	1975-1978	1979-1982	1983-1986	1987-1989	1990-1994
N	862	861	713	589	730
Constant	-2.15*** (.23)	-1.37*** (.16)	-1.58*** (.17)	-1.11*** (.18)	-1.13*** (.14)
NDP Win	.02 (.20)	.31 (.20)	.23 (.16)	.14 (.24)	.31** (.13)
Incumbent Candidate	.06 (.21)	-.47*** (.15)	-.07 (.15)	-.30** (.15)	-.23** (.12)
Prestige Scale	.06 (.05)	-.05 (.04)	-.01 (.05)	.01 (.05)	.00 (.04)
Multi-Member District	.19 (.32)	-16 (.29)	.09 (.21)	.89*** (.30)	.72*** (.28)
Atlantic Province	.26 (.28)	-10 (.21)	-.15 (.25)	-.18 (.22)	-.07 (.19)
Urban District	.28 (.22)	.25 (.20)	.43** (.18)	.46** (.19)	.50*** (.15)
Metropolitan District	.67*** (.20)	.41** (.17)	.60*** (.16)	.49*** (.17)	.50*** (.14)

* = significant at the .05 level, 1-tailed test
** = significant at the .025 level, 1-tailed test
*** = significant at the .005 level, 1-tailed test

TABLE 3: PROBABILITY THAT AN MLA WILL BE A WOMAN, for different conditions with respect to time, opposing candidates, party, and riding type.

Con- Dition	Competition Party District 1975- 1979- 1983- 1987- 1990- Includes: Type 1978 1982 1986 1989 1994
1	Incumbent	Other Party	Rural	2.2%	3.3%	4.5%	6.8%	8.2%
2	Incumbent	Other Party	Urban	4.4%	5.2%	10.6%	16.1%	18.9%
3	Incumbent	Other Party	Metro	9.0%	7.2%	14.2%	18.1%	19.5%
4	Incumbent	NDP	Rural	2.0%	6.4%	7.5%	9.0%	14.5%
5	Incumbent	NDP	Urban	4.0%	9.5%	15.9%	19.8%	29.1%
6	Incumbent	NDP	Metro	8.4%	12.7%	20.6%	21.8%	29.8%
7	No Incumbent	Other Party	Rural	1.8%	8.7%	5.4%	11.9%	12.5%
8	No Incumbent	Other Party	Urban	3.7%	12.5%	12.1%	24.5%	26.1%
9	No Incumbent	Other Party	Metro	7.8%	16.3%	16.3%	27.1%	26.4%
10	No Incumbent	NDP	Rural	1.6%	14.9%	8.7%	14.9%	20.6%
11	No Incumbent	NDP	Urban	3.4%	20.3%	17.9%	29.1%	37.8%
12	No Incumbent	NDP	Metro	7.2%	25.5%	23.3%	31.9%	38.2%