Convergence and Divergence: Change in Support for Federal Parties Across the Western Provinces

It is common in Canadian politics to talk about Western Canada as a Conservative stronghold.  This association is most commonly based off of perceptions of Alberta, though often it travels to other provinces in the West, especially to Manitoba and Saskatchewan.  Observers from outside the West, and sometimes from the West, can find themselves transferring the assumptions that they have about Alberta to the other Western provinces.  To what extent is this justifiable?  In this post I look at the degree to which the three other Western provinces, Manitoba, Saskatchewan, and British Columbia differ from Alberta in their voting behaviour in federal elections.  Controlling for the different demographics of ridings in each of the Western provinces, I am able to assess the degree to which simply being in a particular province affects parties’ vote shares in a given riding.  Looking at trends between 1988 and 2021, I find three separate patterns in the differences between Alberta and each of three other Western provinces.  In recent elections support for parties in Saskatchewan has completely converged with Alberta.  By contrast, voting patterns in British Columbia differ significantly from Alberta, showing no sign of convergence.  Manitoba ends up in between Saskatchewan and British Columbia, there is some evidence of convergence between Manitoba and Alberta but it is far from complete and the trend towards convergence appears to end around the 2011 election.

To look at the convergence and divergence between support for parties in the four Western provinces, I look at the correlation between a federal riding being in a province and support for different parties.  I do this using seemingly unrelated regression models.  This allows me to control for a number of demographic factors that may affect a parties’ vote share independently from the province that a riding is in.  Using census data, I control for the proportion of people in the riding that speak French as their mother tongue, the proportion of Indigenous and ethnoracialzied minorities (recorded in the census as visible minorities), the proportion of individuals with college and university degrees, and the median income of the riding.  I also control for whether the riding is in a major urban centre- Winnipeg, Saskatoon, Regina, Edmonton, Calgary, Vancouver, Victoria, or a suburb* of one of these major centres.  Because of the impact that incumbency can have on vote share, I also control for whether the party has an incumbent running in the riding.  All of this allows me to compare similar ridings across provinces.  In effect, I am asking whether there is a difference in the way a riding in Alberta votes and a similar riding in one of the other three Western provinces votes. 

My analysis goes back as far as the 1988 election.  This requires grappling with the split on the right between the PCs and Reform/Canadian Alliance between the 1988** and 2004 elections.  To compare years in which the right was split in Canada with years in which the Conservative party was united, I add together the vote shares of the Progressive Conservatives and Reform/Canadian Alliance for the elections between 1988 and 2004.  While there were significant differences between the two parties, to look at just one would understate the strength of the conservative vote between 1988 and 2004.  Obviously the exclusion of the Reform/Canadian Alliance parties would miss a large share of conservative voters.  While the PCs were the weaker of the two parties in Western Canada, they still managed over 10% of the vote in every Western province in 1993, almost 15% of the vote in Alberta in 1997, and over 15% in Manitoba in 1997.  To focus only on Reform, would thus also understate the share of the conservative vote in the Western provinces.

Saskatchewan

Perhaps the most interesting province to look at is Saskatchewan.  The figure below shows a clear trend in convergence between support for each of three major parties in Saskatchewan and Alberta.  In the late 1980s and early 1990s the differences between parties’ support in Saskatchewan and Alberta were comparable to the differences between the parties’ support in British Columbia and Alberta.  The two conservative parties did about 24 percentage points worse in a riding in Saskatchewan than in a similar riding in Alberta in both the 1988 and 1993 elections.  This difference got sharper in 1997, when the conservative parties together were winning almost 30 percentage points fewer votes in Saskatchewan ridings than in similar Alberta ridings.  By contrast, the NDP did about 18 percentage points better in Saskatchewan ridings than in similar Alberta ridings in 1988, 16 percentage points better in 1993, and almost 25 percentage points better in 1997.  The Liberals also did better in Saskatchewan than in Alberta, though they benefited less from relative conservative weakness in Saskatchewan than the NDP did.

Dotted lines show the error bars for a 95% confidence level.

The period between 2000 and 2015 is a period of moderate convergence between Saskatchewan and Alberta.  The conservative parties were still doing better in Alberta than they were in Saskatchewan during this period (though in 2015 the difference between conservative support in Alberta and Saskatchewan is not quite statistically significant), but the gap between the two provinces was not as large as previous elections.  Conservative support was between 15 percentage points and 6 percentage points lower in Saskatchewan ridings than it was in similar Alberta ridings.  After the 2000 election, the advantage the NDP had in Saskatchewan as compared to Alberta also weakened.  Where prior to 2000 the NDP’s advantage to Saskatchewan was larger than the Liberals’, after the 2000 election the parties were pretty comparable.  The exception to this was the 2011 election in which the NDP was historically strong across the country and the Liberals were historically weak.  Over this period the Liberals and NDP did between 11 percentage points and 3 percentage points better in ridings in Saskatchewan than in similar Alberta ridings depending on the election and the party (excluding the Liberals’ particularly poor showing in the 2011 election).

The last two elections have seen complete convergence between party vote shares in Alberta and Saskatchewan.  In 2019, there was no statistically significant difference between Conservative vote share in similar Alberta and Saskatchewan ridings.  Then, in 2021, the Conservatives did better in Saskatchewan than in Alberta, with ridings in Saskatchewan giving the Conservatives almost 6 percentage points more support than similar ridings in Alberta.  The flip side of this is the elimination of any clear difference between Saskatchewan and Alberta with respect to either NDP or Liberal support. 

Manitoba

If Saskatchewan is a case where the last two elections have seen complete convergence with Alberta, Manitoba is a case of partial convergence.  As with Saskatchewan, the figure below shows the largest divergence between Manitoba and Alberta is in the 1988, 1993, and 1997 elections.  In 1988, conservative parties did 22 percentage points worse in Manitoba ridings than they did in similar Alberta ridings.  In 1993 and 1997 the discrepancy increased to 29 percentage points and 28 percentage points respectively.  In 1988, this discrepancy in conservative vote share between the two provinces worked in the Liberals’ favour, with the party doing 20 points better in Manitoba ridings.  In 1993, both the NDP and Liberals did better in Manitoba than Alberta, with the Liberals doing 15 percentage points better and the NDP doing 12 percentage points better.  The difference between the Liberals’ strength in Manitoba and Alberta declined a bit in the 1997 election while the NDP’s increased.  In 1997 the Liberals did 7 percentage points better in Manitoba ridings than they did in similar Alberta ridings, while the NDP did almost 20 percentage points better in Manitoba ridings than in similar Alberta ridings.

Dotted lines show the error bars for a 95% confidence level.

Like in Saskatchewan, the difference between Manitoba and Alberta declined in the mid-2000s and again in the late 2000s and early 2010s.  Between 2000 and 2006 conservative parties did between 16 and 18 percentage points worse in Manitoba ridings than in similar Alberta ridings depending on the election.  Meanwhile the Liberals and NDP did between 4 and 10 percentage points better in Manitoba ridings than in similar Alberta ridings depending on the party and the election. 

There is a further convergence between Manitoba and Alberta in the 2008 and 2011 elections, with the difference between Conservative support in Manitoba ridings and similar Alberta ridings declining first to 10 percentage points, then to 4 percentage points.  By 2011 the difference between support in Manitoba ridings and similar Alberta ridings was no longer statistically significant for both the Liberals and the NDP.

Manitoba differs from Saskatchewan between 2015 and 2021.  Where Saskatchewan converged with Alberta, Manitoba diverged.  The Conservatives did 13 percentage points worse in Manitoba ridings than in similar Albertan ridings in 2015 and 2019, and then 7 percentage points worse in 2021.  In 2015, the gap between Liberal support in Manitoba and support in Alberta was as high as it ever was during the 1988-2021 period, though it fell to about 9 percentage points in 2019 and 14 percentage points in 2021.  The NDP actually did worse in Manitoba ridings than they did in similar Alberta ridings in 2015, though the difference between the NDP’s support in Manitoba ridings and similar Alberta ridings was not statistically significant in either the 2019 or 2021 elections.

British Columbia

Of the three Western provinces, the dynamics in British Columbia are the most distinct.  Where there is evidence of at least some convergence between Alberta and Manitoba, the figure below shows no evidence of convergence between British Columbia and Alberta.  Granted, there are elections where conservative parties’ support in British Columbia is closer to their support in Alberta and elections where it is farther apart, but there is no clear pattern over time.  After 1997, conservative parties never did more than 20 percentage points worse in Manitoba or Saskatchewan ridings than in similar Alberta ridings.  By contrast, in 2015 the Conservatives did 22 percentage points worse in British Columbia ridings than they did in similar Alberta ridings, and in 2019 they did almost 26 points worse.  Indeed, in only one election, 2000, is the gap between conservative parties’ support in British Columbia ridings and their support in Alberta ridings less than 10 percentage points. 

Dotted lines show the error bars for a 95% confidence level.

Both the NDP and Liberals have consistently done better in ridings in British Columbia than they have in similar Alberta ridings.  The difference between NDP support in British Columbia and Alberta is greatest in 1988, when the NDP did almost 19 percentage points in British Columbia ridings than in similar Alberta ridings.  After 1988 the difference between NDP support in British Columbia ridings and similar Alberta ridings varies between almost 12 percentage points in 1997 and 2 percentage points in 2000.  It is statistically significant, though, in every election except for 2000.  The difference between Liberal support in British Columbia ridings and similar Alberta ridings peaks in 2015 at just under 12 percentage points.  Only in 1993 and 2000 does the difference between Liberal support in British Columbia ridings and similar Alberta ridings fail to reach statistical significance.  Like with the conservative parties there is not a clear pattern over time for either party.  The difference between NDP support in British Columbia and similar Alberta ridings never again reaches the levels it was at in 1988, but beyond that there is little evidence of convergence between NDP support in British Columbia and Alberta. 

Conclusion

One can draw a couple of conclusions from this.  The first is that, while there is convergence between some provinces in Western Canada, federal party support remains at least somewhat different across the Western provinces.  Only in Saskatchewan does party support match that of ridings with similar demographics as Alberta.  There was some convergence between Manitoba and Alberta in the late 2000s and early 2010s, but Manitoba seems to have diverged again in the last three elections (the possible exception in Manitoba is the NDP, whose support in Manitoba appears to have converged with its support in Alberta).  British Columbia has become the unique Western province, and now has the largest divergence between support for its parties and similar ridings in Alberta.  It is worth noting, however, that there is no evidence of divergence in British Columbia.  One might have expected to see such divergence as the province has gone from one of two (alongside Alberta) that produced the Reform party to a province that is in increasing conflict with Alberta over pipelines and resource development.

The other pattern that comes through is that the differences between the Western provinces cannot be explained away by the demographics of the provinces.  Even when one controls for the fact that many British Columbia ridings are more diverse than ridings in Alberta and that many Alberta ridings have higher median incomes than those in British Columbia, stark differences between the provinces remain.  The same can be said of Manitoba, but to a lesser degree, and of Saskatchewan until the most recent elections. 

There is some truth to the idea that Western Canada is a particularly conservative part of the country.  However, the stereotype that Western Canada provides a solid base of support to the Conservative party applies mostly to just Alberta and Saskatchewan.  British Columbia is certainly a unique political environment, and to a lesser degree Manitoba is as well. 

*In British Columbia I include the urban centres in the Lower Mainland such as Abbotsford, Burnaby, New Westminster, Richmond, and Surrey as suburbs of Vancouver. 

**While the Reform party did not win seats in the 1988 election, it contested the election winning 15% of the vote in Alberta and almost 5% in BC.

The More Things Change: Volatility and Stability in Recent Alberta Elections

At first glance the last three Alberta elections suggest a high level of volatility in the Alberta electorate.  The 2015 election produced the province’s first ever New Democratic (NDP) majority government.  This was followed by a large United Conservative Party (UCP) majority in 2019 and then a relatively close (at least by Alberta standards) election in 2023 won by the UCP.  However, these results mask a fair amount of stability in support for different ideological blocks in the province.  The combined vote share of right-wing parties has been relatively stable over the last few Alberta elections, as has the combined support for progressive parties.  This suggests that the NDP has established itself as the main challenger to the conservative parties largely by uniting progressive voters as opposed to winning over more conservative voters.  This dynamic is generally consistent across Alberta’s regions, though in recent elections the progressive block seems to have slightly expanded their share of the vote in Calgary.

To capture the degree to which there has been movement across ideological blocks in recent Alberta elections I measured combined vote shares for parties on the right and parties on the progressive or left side of Alberta politics.  I have labeled the progressive parties as left parties, though parties like the Alberta Party and Liberals have at times taken more centrist positions (I use the term left because it sounds better than “not conservative”).  In the right block I include the Progressive Conservatives (PC) and Wildrose Alliance (in 2004 called the Alberta Alliance) in elections prior to 2019 and the UCP in 2019 and 2023.  In the left block I include the NDP, Alberta Party, Liberal Party, and Green party (in the 2004 and 2008 the Alberta Greens and in 2012 the Evergreen party).  I exclude small parties from the analysis of blocks, so the percentage of votes across the different blocks will add up to slightly less than 100% in each election year.  I look at elections that took place between 2004 and 2023.  For regional analysis I break Alberta into Calgary, Edmonton, and Rural Alberta.*

Alberta as a Whole

The graph below shows that, despite the NDP’s win in 2015, voters on the right have outnumbered voters on the left in each election going back to 2004.  Much of the volatility in vote share between the blocks occurs between 2004 and 2012, not between 2015 and 2023.  Ralph Klein’s waning popularity in 2004 allowed the left block to get over 40% coming out of an era in which the Progressive Conservatives were dominant.  The replacement of Klein with Ed Stelmach saw an increase in right block support from 56% of the vote to 60%, pushing the left block down to 39%.  The combination of strategic voting against the Wildrose Alliance (polls showed the party competitive for government in 2012) and Alison Redford’s shift of the PCs towards the centre further led the right block’s support to balloon to 78% on 2012.

By contrast between 2015 and 2023 the right block’s vote share has remained between 52% (in 2015) and 55% (in 2019), with the 2023 election seeing it win 53% of the vote.  The left block saw a bit more volatility (voters can leave both blocks and vote for independents or small parties).  Nonetheless, the left block’s worst result between 2015 and 2023 (43% in 2019) was within 5 percentage points of its best result over these three elections (48% in 2015).  The gains made by the left block between 2019 and 2023 were relatively small, with the group of parties moving from 43% to 46% of the vote. 

The big shifts in the outcomes of Alberta elections have largely been driven by the vote splits between parties within the different blocks, shown below.  The most notable cases of fragmentation have been on the left in 2004 and 2008 when vote splits between the Liberals and NDP led to large PC majorities, and on the right in 2015 when vote splits between the PCs and Wildrose Alliance led to an NDP election win.  In less stark terms, the NDP’s decline and recovery between 2019 and 2023 appears to be paralleled by growth and decline of support for the Alberta Party as opposed to losses to the UCP.

Calgary

Calgary is perhaps the most interesting of the three regions in Alberta.  In the last two elections the city has been a key battleground between the NDP and the UCP.  The city has been becoming slightly more progressive over time, as shown below.  If one excludes the 2012 election, in which Alison Redford moved the PCs to the centre, there has been a slight but steady decline in the proportion of Calgarians voting for right parties and corresponding increase in the proportion of Calgarians voting for progressive parties.  In 2004 right wing parties won 57% of the vote on Calgary as compared to 43% for progressive parties.  Since 2015, the proportion of Calgarians voting for a progressive party has steadily increased, going from 44% in 2015 to 45% in 2019 and 50% in 2023.

It is notable that the NDP managed to hold on to most of its support in Calgary in 2019, despite losing a substantial number of seats in the city as a result of the unification of the PCs and Wildrose Alliance.  The graph below shows that the NDP won 34% of the vote in Calgary in 2015 and 33% in 2019.  They did this despite the gains made by the Alberta Party in Calgary between 2015 and 2019.  The NDP then made substantial gains in Calgary between 2019 and 2023, with their vote share rising to 49%.  If there is a region in Alberta where the progressive parties are making gains, it is Calgary.  To the extent that Calgary is likely to remain a battleground, this bodes well for the chances of the NDP, so long as they can keep progressive voters in Calgary united behind them.

Edmonton

The graph below shows that the dynamics in Edmonton largely reflect those in the rest of the province, with the difference between Edmonton and the rest of Alberta being that the left block is stronger in Edmonton.  The greater volatility of the 2004-2012 period across Alberta comes through in the results in Edmonton.  The right block’s vote share in Edmonton increased when both Ed Stelmach and Alison Redford led the PCs.  Stelmach grew the right block’s vote share from 37% to 46%.  The mix of Redford’s centrism and strategic voting against the Wildrose Alliance raised the right block’s vote share in Edmonton all the way to 60%.  Since 2015 Edmontonians have largely backed progressive parties.  In 2015, support for the left block in Edmonton rose to 69%.  In 2019 it fell slightly to 63% and remained at a similar 64% in 2023.

The NDP’s success in Edmonton has thus largely been about consolidating the progressive vote behind itself.  The graph below shows, that like in the province as a whole, vote splits between the Liberals and NDP prevented either party from dominating in the city prior to 2015 despite the fact that a solid majority of Edmontonians were voting for progressive parties in 2004 and 2008.  Like in the rest of Alberta, the NDP’s decline in support in 2019 in Edmonton reflected growth in support for the Alberta Party to a much greater degree than it reflected growth in support for the UCP (and the NDP’s recovery in Edmonton in 2023 reflected a decline in support for the Alberta Party). 

Rural Alberta

Rural Alberta (inclusive of some of Alberta’s smaller cities such as Lethbridge) has largely been the safest part of the province for conservative parties in the province.  This is reflected in the graph below. Over the 2004-2023 period, support for the right block in rural Alberta never fell below 60% while support for the left block remained below 40%.  Though support for progressive parties has not declined in rural Alberta, there is little evidence of a sustainable increase in support either.  In 2004 progressive parties won 31% of the rural Albertan vote, as compared to 31% in 2019 and 32% in 2023.  There was some volatility in between these elections, with support for progressive parties dropping when they faced PC campaigns led by Stelmach and then Redford.  It is also notable that the share of support for progressive parties increased to 37% when the NDP made their breakthrough in 2015.

The graph below shows that, for the most part, the NDP’s 2015 success in rural Alberta was driven by the vote split between the PCs and the Wildrose Alliance party.  With the two right-wing parties splitting the rural Alberta vote almost evenly between each other, the NDP had a vote share in rural Alberta comparable to both.  In the two elections since the PCs and Wildrose Alliance merged to form the UCP, there has been little evidence of the gap between the NDP and the UCP narrowing.

Conclusion

For all of the volatility in the outcomes of Alberta elections, there is an underlying stability to the support for the different ideological blocks in the province.  This is reflected to the greatest degree in the last three elections.  This has implications for the future of the province’s politics.  So long as the two blocks remain consolidated behind a particular party, for the right the UCP and for the left the NDP, it is hard to see a path to government for the NDP.  While the growth in support for progressive parties in Calgary is encouraging, it would need to take place on a much larger scale than it has over the past few elections for it to shift the balance of power in the province.  The results of the last three elections suggest that the most likely path to power for the NDP will need to involve a fracturing of the alliance between the more moderate right leaning and more extreme right-wing voters backing the UCP.  With the progressive vote in Alberta hovering around 45%, the NDP may not need the kind of equal split of right-wing voters that occurred between the PCs and Wildrose Alliance in 2015, but they do need a right-wing competitor to take a substantial share of the ideologically conservative vote from the UCP.  The challenge for the NDP in the meantime will be keeping the progressive vote united behind them.  The consistently weaker level of support for progressive parties in Alberta means that the NDP cannot lose support to a progressive challenger if they are to compete for government.

*I did analysis on North, South, East, and West Rural Alberta, but the analysis is similar enough that it makes sense to combine it into one analysis of rural Alberta for brevity’s sake.

Senate Reform’s False Promises Part 3: The Chrétien Years

In my past couple of posts, I have been grappling with how an elected Senate could have changed regional representation and policymaking in Canada.  A key plank of the Reform party’s platform when it first won seats in parliament in the early 1990s was to create a “Triple-E” Senate.  This Senate would have been elected, had equal representation from each of the provinces, and would be effective in that it would be able to block legislation passed through the House of Commons.  In my past two posts I looked at how an elected Senate might have affected Justin Trudeau’s as well as Stephen Harper’s and Paul Martin’s governments.  In this post I look at the governments led by Jean Chrétien during the 1993 to 2004 period.  This is perhaps the most interesting period in which to look at an elected Senate.  The Liberals would not have been able to win the same majorities in the Senate that they won in the House of Commons over this period of time.  However, Reform and its successor, the Canadian Alliance, would not have been the main beneficiary of Liberal minorities in the Senate.  Rather, the Progressive Conservatives (PCs) and NDP would have held the balance of power in the Senate and would have gained some influence over policymaking.  At the same time, the Liberals would have been able to play the PCs and NDP off against each other, leaving the Liberals in a fairly strong position.

For these posts I have been assuming an elected Senate would be modelled on the proposal in the failed Charlottetown Accord.  This proposal would have given each province six Senators to be elected through a single transferable vote electoral system (STV).  This approach is similar to the Australian Senate, though in Australia each state gets 12 Senators instead of six.  In this model I assume that each of the territories will have one Senator each.  It is worth noting that the two earliest elections I look at in this post occur before the establishment of Nunavut, and so there are only two territorial Senators for the 1993 and 1997 elections, but 3 for the 2000 election.  This leaves the Senate with 62 members in 1993 and 1997 and then 63 in 2000.  In either case a party needs 32 Senators to have control of the Senator without having to rely on finding a way to break ties in its favour to pass legislation.

To estimate the partisan distribution of seats in the Senate, I use vote shares from the 1993, 1997, and 2000 elections.  While voters will not behave the same way under an STV system as they will under a first past the post system, in the absence of actual from an STV election this is the best evidence available.  In the territories I assume that the same party that won each territory’s seat in the House of Common would have won its Senate seat.  I suspect that while some voters would have behaved differently under an STV system, that the number of individuals who would have voted differently would not have been large enough to dramatically change the results. 

1993: the Emergence of Reform

The 1993 election saw Reform emerge as a party arguing in part that the representation by population model for allocating seats in the House of Commons led Western interests to be under-represented.  The Reform Party’s breakthrough in the 1993 election came largely in the Western provinces of Alberta, British Columbia, and Saskatchewan, so would an elected Senate have increased Reform’s representation or the West’s power?  Figure 1 suggests that it would not have.  I estimate that Reform would have only won 11 Senate seats in 1993 compared to 30 for the Liberals and 13 for the PCs.  At 30 seats, the Liberals would have been just short of the 32 needed for control of the Senate and would have been able to get control with the support of any of the four opposition parties.

Figure 1: The PCs are in blue, the Liberals in red, the NDP in orange, the Bloc Québécois in light blue, and Reform in Green.

It is unlikely in this scenario that the Liberals would have turned to Reform to help them get legislation through the Senate.  Rather they would have had two palatable options in the PCs and the NDP to get the 32 Senate votes they needed to get legislation passed.  While it may be tempting given the current context to assume a Liberal-NDP alliance under these circumstances, it is important to remember than in the early 1990s the Liberals were pursuing a policy of cuts to social programs driven by a desire to eliminate the deficit and cut debt.  The NDP would likely have been uncomfortable working with a Liberal party pursuing such policies on issues related to social programs and the budget.  The PCs, however, may have been more sympathetic to the Liberals and offered them a partner.  Depending on the legislation, the Liberals in this Senate could always threaten to work with a different party if they did not get what they want in a deal.  If the PCs threatened not to work with the Liberals, the Liberals could always threaten to work with the NDP to get a deal to the left of what the PCs would have not been comfortable with.  The Liberals could have done the same with the NDP, threatening to work with the PCs and even Reform to get a deal much to the right of what the NDP was comfortable with. 

There is some irony in the situation an elected Senate in 1993 would have created in that by trying to empower Western Canada, Reform was arguing for a change that would have instead strengthened a party that would have received most of its Senate seats from Atlantic Canada.  Figure 2 shows that the PCs would have won two seats in each of the Atlantic provinces, making just over 60% of their Senate caucus Atlantic Canadian.  By contrast, the PCs would have only 3 Senators from the West.  Of the three parties most likely to work together, the NDP would have the largest proportion of its caucus coming from Western Canada, but it would have been the weakest of the three parties likely to work together.  The policies that the NDP would have sought would have also been very different than those sought by the plurality of Western Canadians that had voted for Reform.  Oddly enough, of the three parties most likely to work together, the Liberals would have had the most seats in Western Canada, but two thirds of the Liberal Senate caucus would have still come from provinces east of Manitoba. 

Figure 2: The PCs are in blue, the Liberals in red, the NDP in orange, the Bloc Québécois in light blue, and Reform in Green.

There are two reasons for what would have been Reform’s weakness in the Senate after the 1993 election.  In 1993 Reform was not yet dominant in Western Canada.  The large House of Commons seat shares that it won in Alberta and BC were driven in part by Reform’s plurality of support in both provinces and the way that the first past the post electoral system rewards support.  Using a proportional system to award seats in provinces significantly reduces Reform’s advantage in Western Canada.  Meanwhile Reform is so weak east of Saskatchewan (Reform only won 22% of the vote in Manitoba) that even allocating seats based on a proportional system does not help the party much.  Indeed, I estimate that Reform would have only won four seats east of Saskatchewan (one each in Manitoba, Ontario, New Brunswick, and Nova Scotia).  Given that there are seven provinces east of Saskatchewan, this would be a severe problem in a system that gives each province equal representation.

1997: the 1990s Liberals at their Weakest

The 1997 election saw the Liberals’ worst result of the three elections they fought with Jean Chrétien as leader.  In the House of Commons the party came close to losing its majority, winning only 155 seats (151 were needed for a majority).  Reform made gains in Western Canada while the PCs and NDP both made gains, largely in Atlantic Canada and, for the PCs, in Quebec.  Nonetheless, figure 3 shows that my estimates would have placed the Liberals in largely the same place in the Senate.  At 23 Senate seats, the Liberals would have been much further from a majority in the Senate than they would have been after 1993.  However, their potential partners would have remained the same (with the exception that they would no longer be able to gain control of the Senate by working solely with the Bloc Québécois- a prospect that would have been very unlikely in either 1993 or 1997 given the context of the Bloc’s support for the yes side of the 1995 Quebec secession referendum).  The PCs would have lost their Senate seat in BC, dropping from 13 to 12 Senate seats, but still would have had enough seats that a Liberal-PC alliance would have had control of the Senate.  The NDP, by contrast, would have made significant gains, largely in the Atlantic.  The party would have gone from 5 Senate seats to 11.  Thus, a Liberal-NDP alliance would have also had the ability to control the Senate as it would in 1993, with the Liberal losses between 1993 and 1997 being compensated for by NDP gains.  While the Liberals would have made up a smaller share of a 1997 Liberal-PC alliance or a Liberal-NDP alliance than they would have in 1993, they would have still been in a very powerful position in such an alliance.  They would have retained the threat that they could go and work with the other of the two parties if either the PCs or the NDP made too many demands in any alliance.

Figure 3: The PCs are in blue, the Liberals in red, the NDP in orange, the Bloc Québécois in light blue, and Reform in Green.

The regional breakdown of my 1997 estimates, shown in figure 4, suggests that the West would have been less well represented in Liberal-PC or Liberal-NDP alliance in 1997 than it would have been in 1993.  Reform would have made gains in Western Canada, picking up seats in BC, Alberta, and Manitoba.  The PCs, however, would have lost one of their three Western Senate seats, making them more reliant on their support in Atlantic Canada.  While the NDP would held what it had in the West in 1993, the entirety of its gains would have come east of Manitoba.  The party would have gone from having a Senate caucus exclusively from the West and the North to having just over half of its seats in Atlantic Canada and Ontario.  The Liberals would have lost two of their Western Canadian seats, one in Manitoba and one in Alberta.  The result would have been that, despite the growth of Reform, Western Canada would have had less influence over the likely alliances that would have had control over the Senate following the 1997 election.

Figure 4: The PCs are in blue, the Liberals in red, the NDP in orange, the Bloc Québécois in light blue, and Reform in Green.

2000: Chrétien’s recovery

The 2000 election saw the Liberals recover a significant amount of the support they had lost between 1993 and 1997.  Figure 5 shows that I estimate that this would have led to Liberal gains in the Senate, with the party ending up with 26 Senate seats.  By 2000 the Reform Party had rebranded itself as the Canadian Alliance in an effort to win seats outside of Western Canada.  My estimates suggest that this would have led to some, though limited gains in the Senate, as the party would have gone from 13 seats 1997 to 15 in 2000.  The power dynamics in the Senate would have largely been the same as they were after the 1993 and 1997 elections.  The PCs would have had the same number of seats as they had in 1997 (compensating for the loss of their Quebec seat by picking up a seat in PEI).  The NDP would have lost 4 of the 11 seats they would have won in 1997, but at 7 seats they would have just barely enough seats to give a Liberal-NDP alliance enough seats to control the Senate.   The Liberals would have thus been in the same dominant position in the Senate after the 2000 election as they would have been after 1993 and 1997.  They could get the policies they wanted by threatening to work with the NDP if the PCs were not willing to make enough compromises, or threatening the NDP that they would work with PCs if the NDP was not willing to make enough compromises.   The additional seats that the Canadian Alliance would have gained would not have prevented them from being marginalized in the Senate.

Figure 5: The PCs are in blue, the Liberals in red, the NDP in orange, the Bloc Québécois in light blue, and the Canadian Alliance are in teal.

Figure 6 shows my estimates for the regional breakdown of an elected Senate after the 2000 election.  One of the Canadian Alliance’s new seats comes in Saskatchewan and the other comes in Ontario.  The Canadian Alliance’s gain in Saskatchewan comes at the expense of the Liberals, while its gain in Ontario comes at the expense of the NDP.  This has the effect of making either Liberal-PC or a Liberal-NDP alliance slightly more reliant on Atlantic Canada and Quebec than it would have been in 1997 or 1993.  There are some regional shifts amongst the three parties most likely to work together.  The NDP also loses two of its Atlantic seats (one in PEI to the PCs and one in New Brunswick to the Liberals) and its Northern seat to the Liberals. The PCs meanwhile lose their Quebec seat to the Liberals. 

Figure 6: The PCs are in blue, the Liberals in red, the NDP in orange, the Bloc Québécois in light blue, and the Canadian Alliance are in teal.

Conclusion

Overall, my estimates of the impact of an elected Senate for the 1993-2000 elections follows similar themes to the earlier posts.  Governments with majorities, in this case the Liberals, would have had to negotiate more with opposition parties than they would have without an elected Senate.  Those negotiations would not have been insurmountable, however, and it is likely that the governing party would have been able to get what it wanted without having to make a large number of compromises.  This is particularly the case in the 1993-2000 period as the Liberals would have had their choice of compromising with a party slightly to their right, the PCs, or compromising with a party slightly to their left, the NDP.  It is hard to see an elected Senate paralyzing government in the way that it has in the United States, at least over the three periods that I looked at.  This is not necessarily because Canadian parties are less polarized or more willing to work with each other, but because Canada’s multi-party system often gives the governing party multiple options for parties that they can work with.

While this analysis suggests that an elected Senate may not be as paralyzing as it is in the United States, it also suggests that the gains made by Western Canadians would not be as extensive as many advocates of Senate reform often suggest.  To be sure, the alliances that would control an elected Senate usually would have included some Western Canadian Senators, but when the government is not Conservative (and thus includes substantial representation from the West in the House of Commons) those Senators would be the minority within their party’s Senate caucuses and would represent parties that tend not to prioritize Western interests as expressed by those that tend to favour an elected Senate.  In both the case of the Conservative party when facing Liberal governments led by Justin Trudeau and the Reform/Canadian Alliance party when facing Liberal governments led by Jean Chrétien, right leaning Senators from the West would have found themselves isolated in the Senate with limited ability to block legislation opposed by large shares of Western Canadians.  The reason for this is something that is fairly straight forward though often over-looked by Western advocates of Senate reform.  An elected Senate with equal representation for each province would increase the representation of the least populous provinces.  The least populous provinces in Canada, however, are in the Atlantic not the West.  In most cases, the struggles that the most popular party in Western Canada has in Atlantic Canada cancel out any advantage that they would get by winning the West.  As a result, an elected Senate would not do much to increase the power of Western provinces over legislation.

Addendum: As in the last two posts, using a first past the post system to allocate Senate seats, where the most popular party in a province would get all of its Senate’s seats, usually strengthens the party in government.  In 1993 and 2000 the Liberals would have won majorities in the Senate if Senate seats were allocated using a winner-take-all method.  In both 1993 and 2000 the Liberals would have won 7 provinces. They would have won Manitoba, Ontario, and the four Atlantic provinces in both elections.  In 1993 they would have added Saskatchewan to that total, and in 2000 they would have won Quebec.  By contrast in 1993 Reform would have won just Alberta and BC, while in 2000 Reform would have won Alberta, BC, and Saskatchewan (in 1993 the Bloc would have won Quebec). 

A winner-take all system would have produced interesting results in 1997.  The Liberals would have won just four provinces under such a system: Manitoba, Ontario, PEI, and Newfoundland.  Reform would have won three: BC, Alberta, and Saskatchewan.  The PCs would have won two provinces: New Brunswick and Nova Scotia, and Bloc would have won Quebec.  The Liberals would have split the two territories with the NDP, preventing a Reform-PC alliance from controlling the Senate.  A Liberal-NDP alliance, however, would also have been insufficient to control the Senate.  Control of the Senate would hinge on the ability of the Liberals to make compromises with the Bloc, or on their ability to get the PCs to work with them instead of Reform. 

Senate Reform’s False Promises Part 2: The Harper Years

One of the Reform party’s ideas for institutional change that survived the transition to the modern Conservative party was Senate reform.  After taking over as Prime Minister, Stephen Harper made attempts to appoint elected Senators and to get new appointees to commit to term limits.  He also explored the idea of moving to an elected Senate.  Harper’s attempts at Senate reform ultimately failed to produce an elected Senate, in part because the Supreme Court ruled he needed the consent of the provinces to make the Senate elected and in part because there was no provincial consensus around Senate reform.

What would an elected Senate have looked like during the Harper years?  In this series of posts, I estimate results for an elected Senate and discuss the implications for the governments’ ability to pass legislation.  In the previous post I discussed what an elected Senate would have looked like during the Justin Trudeau led governments from 2015 to today.  In this post I examine both the Harper and Paul Martin led governments.  As with my post on the Trudeau governments, I find that an elected Senate under during the Harper governments would not have changed very much.  Somewhat ironically, given the Conservatives’ efforts to pursue Senate reform, my estimates suggest that the largest impact that an elected Senate could have had would have been a slight constraint on the majority government the Conservatives won in 2011.

For this post I assume a Senate following the model developed in the Charlottetown Accord and similar to the one used in Australia (though Australia’s Senate has 12 Senators per state and, following Charlottetown, I assign 6 Senators to each province).  Under this model each province gets an equal number of Senators and Senators are elected using a single transferable vote (STV) electoral system.  For lack of actual results from an STV election, I use the vote shares of the elections for the House of Commons for each province to estimate the number of seats that each party would have won in each province.  Under the Charlottetown model each territory would have received one Senator, so I assume each territory would be represented in the Senate by the same party that won its House of Commons seat.  At 63 Senators in the Senate, a party would need 32 to have a majority and control of the Senate.  I also assume that the Senate would not have the power to determine the Prime Minister or vote no confidence in the Prime Minister, and that these powers would remain exclusive to the House of Commons.  I discuss and justify this approach in more detail in my previous post.

2011: the Harper Majority

My estimates for 2011 produce the most interesting results of the 2004-2011 period.  Figure 1 shows that I estimate the Conservatives would have won 30 seats in 2011, just short of the 32 they would need for control.  This Conservatives would end up slightly short of the majority needed to control the Senate and would not have an obvious partner to help them get control of the Senate.  The Bloc Québécois is sometimes suggested as a potential partner for the Conservatives because of the Conservatives’ willingness to respect provincial autonomy (though the Bloc is well to the left of the Conservatives on many issues), but the Bloc would have been one seat short of what would have been needed to give the Conservatives control of the Senate.  The NDP’s 20 seats combined with the Liberals’ 12 would have given them some power to block Conservative legislation.  The Harper Conservatives would have found themselves dealing with an elected Senate after 2011 that had somewhat similar dynamics to the minority parliaments they dealt with after 2006 and 2008.  The Conservatives would have been the largest party by a significant margin, but without a clear partner in a minority situation, they would have had to use a mix of negotiation and brinkmanship to pass legislation.

Figure 1: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

One should not overstate, however, the extent to which an elected Senate would have been a constraint on a Conservative majority.  The Liberals and NDP would have to be in complete lockstep to use the Senate to block legislation.  The defection of just one or two Liberal or NDP Senators to the Conservative side of any vote or the simple absence of three or for Liberal or NDP Senators (depending on the decision made by the Bloc Senator) would have been enough to allow the Conservatives control of the Senate.  Depending on the powers granted to the Senate by reform, the House of Commons may have been able to overrule the Senate or just simply bully the Senate into submission by using the House of Commons’ traditional role as the House with the legitimacy to overrule the other.  I suspect that this Senate would have made life slightly more difficult for the Conservatives during the Harper majority years, but I doubt it would have prevented the Conservatives from pushing though most of their agenda.

Figure 2 shows that the STV electoral system is largely responsible for the Conservative’s inability to gain control of the Senate in 2011.  Even though STV is the proportional electoral system friendliest to the largest party, it is still difficult under this system for any party to win more than half of the seats in any province.  Indeed, the Conservatives only win more than three seats in the two provinces in which they have dominated twenty-first century politics, Alberta and Saskatchewan.  The Conservatives, meanwhile are prevented from getting a majority by their inability to win more than two seats in Newfoundland, Nova Scotia, Quebec, and the Territories. 

Figure 2: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

2008 and 2006: The Harper Minorities

My estimates (shown in figures 3 and 4) suggest that the 2008 and 2006 Senates would have left the Harper Conservatives in more less the same place in the Senate as they were in the House of Commons after those two elections.  Like in the House of Commons they would have had a plurality of seats, but at 26 seats in 2008 and 25 in 2006, they would have been well short of the majority needed to control the Senate.  Indeed, the Liberals fall a couple of seats short of winning a plurality of seats in the Senate in 2006.  At three seats, the Bloc would have been unable to help the Conservatives get to a majority, leaving the party having to find a way to work with the Liberals or NDP (or to at least get one of those parties to absent themselves from votes) in order to get legislation passed.  In these cases, it is hard to see the Senate doing anything more than reinforcing the dynamics the characterized the House of Commons during this period, and thus leaving more or less the same decisions made by the House of Commons in place.

Figure 3: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

Figure 4: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

Figures 5 and 6 show that the broad story behind the 2006 and 2008 is more or less the same as the story for 2011, except that in most provinces (including Alberta and Saskatchewan) the Conservatives are slightly weaker than they were in 2011.  In these two elections only one province, Alberta, would have delivered more than half of its Senate seats to the Conservatives.  Meanwhile, only twice do Atlantic provinces deliver at least half of their Senate seats to the Conservatives (New Brunswick in 2008 and Newfoundland in 2006). 

Figure 5: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

Figure 6: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

2004: The Martin Minority

If things would not have been different for Harper during his minority governments, what about for Paul Martin and the minority government that he won in the 2004 election?  Martin may have had an easier time in the Senate than in the House of Commons.  Like Harper, Martin would have had a plurality but not a majority of seats in the Senate.  The 2004 election would have left the Liberals with 26 seats, well short of the 32 needed for control (shown in figure 7).  However, an NDP with 11 seats would have easily been able to get the Liberals to 32 seats.  By contrast the 2004 election left the Liberals and NDP with a combined 154 seats in the House of Commons, one short of the 155 needed for control.  Martin would end up needing the defection of Conservative MP Belinda Stronach and the support of independent MPs such as Chuck Cadman along with the support of the NDP to survive confidence motions in the House of Commons.  It is unlikely that the Senate that would have been produced by the 2004 election would have been more difficult to manage for Martin than the House of Commons that resulted from that election.

Figure 7: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

The provincial breakdown of Senate seats shown in figure 8 suggests Senate dynamics in 2004 that would have been similar to those I discussed in my previous post on the kinds of Senates that Trudeau would have faced.  The Conservatives would have had an advantage in the three Prairie provinces, but only would have won more than half of a province’s seats in Alberta.  By contrast the Liberals would have held an advantage in three of the four Atlantic provinces as well as in Ontario.  They also would have had an advantage over the Conservatives in Quebec, though the Bloc would have taken the plurality of the seats in that province.  As in 2015-2021 the problem the Conservatives would have faced in an elected Senate in 2004 would have been that any advantage that they would have had in the West would have been cancelled out by disadvantages that they faced in the four Atlantic provinces. 

Figure 8: The Conservatives are in blue, the Liberals in red, the NDP in orange, and the Bloc in light blue.

Ultimately, it is hard to see an elected Senate making a difference to what Harper or Martin would have been able to do while Prime Minister.  Of the two, it would probably have been Harper that faced the most difficulty in the Senate because it would not have given him the same majority in 2011 that he was able to win in the House of Commons.  That being said, Harper would have been very close to a majority in the Senate in 2011, and it is hard to see the Senate that the 2011 election would have produced preventing the Harper government from getting most of its legislation passed.  Ironically, it is the party that wants to abolish the Senate, the NDP, that would have benefited most from an elected Senate.  The STV system that was proposed under Charlottetown and is used in Australia would have ensured that the NDP would have received at least one seat from most provinces in most elections.  This would have left them with a larger proportion of Senate seats than House of Commons seats after most elections.  The NDP would have had the balance of power in the Senate after the 2004 election, though the fragmented House of Commons would have prevented them from doing much with it between 2004 and 2006.  The greater NDP representation in 2006, 2008, and 2011 would not have likely mattered as Conservative governments in the House of Commons in those elections would have kept the NDP from doing anything with their greater representation in the Senate.  The regional story is more or less the same as in the previous post.  The greater representation of the West in the Senate as compared to the House of Commons would have been counterbalanced by the greater representation of the four Atlantic provinces.  The result is a Senate that would have largely confirmed what was happening in the House of Commons .

Addendum: Assigning Senate seats using a plurality rule would not have increased the likelihood that the Senate would end up blocking legislation coming from the House of Commons.  The Conservatives would have had an easier time in 2011 than with a Senate elected under STV, as they would have won eight provinces compared to the NDP’s one (Quebec) and the Liberals’ one (Newfoundland).  The Conservatives would have also had a majority in in the Senate in 2008 with six provinces (BC, Alberta, Saskatchewan. Manitoba, Ontario, and New Brunswick) compared to three provinces for the Liberals (Nova Scotia, PEI, and Newfoundland) and one for the Bloc (Quebec).  A Conservative minority in the House of Commons, however, would have prevented them from using their majority in the Senate to increase what they could do in government.  In 2004 and 2006 a plurality electoral system would have led to bare Liberal majorities in the Senate with the party winning five of ten provinces and being able to use strength in the Territories to end up with a majority.  In both 2004 and 2006 the Liberals would have won Ontario, New Brunswick, Nova Scotia, PEI, and Newfoundland while the Conservatives would have won BC, Alberta, Saskatchewan, and Manitoba (the Bloc would have won Quebec both times).  A Liberal Senate majority after 2006, could have blocked legislation that had the support of the Conservatives and Bloc.  However, the Conservatives were more reliant on the Liberals to get legislation passed the 2006-2008 House of Commons than the Bloc.  Between 2004 and 2006 a Liberal Senate majority would not likely be able to do anything more than confirm what legislation that Martin Liberals were able to get through the House of Commons.

Senate Reform’s False Promises, Part 1: The Trudeau Years

From time to time in Canada debates arise over what to do with an unpopular appointed Senate and how to better ensure that Canada’s regional diversity is represented in the federal government.  Reforming the Senate so that it is elected and includes equal representation from each of the provinces is often offered as a way to give the Senate democratic legitimacy and improve the representation of the less populous provinces.  Senate reform has been a particularly popular issues with those that feel Western Canada has difficulty getting its interests heard in the House of Commons, where the larger populations in Ontario and Quebec give both of those provinces more influence over who controls the legislature.  Critics of an elected Senate often counter that such a Senate would create legislative gridlock.  If different parties win control of the House of Commons and the Senate, each’s ability to block legislation would make it difficult to get anything passed, an experience that is common in the United States.

In this series of posts, I estimate results for an elected Senate for elections between 1993 and 2021.  I suggest that both the potential for the Senate to provide a regional counterbalance protecting Western Canadian interests and for an elected Senate to create legislative gridlock are overstated.  An elected Senate would not have been so different from the House of Commons that the government would have had difficulty passing legislation, nor would it have given parties popular in Western Canada the ability to block legislation passed through the House of Commons.  In this particular post, I look at the 2015-2021 elections that resulted in Liberal governments under Justin Trudeau.  In future posts I will look at the 2004-2011 elections that produced governments led by Paul Martin and then Stephen Harper, as well as the 1993-2000 elections that produced governments led by Jean Chretien. 

Approach

In this work I assume an elected Senate similar to the one proposed in the failed 1992 Charlottetown Accord.  This proposal would have given each province 6 Senators, as well as 1 Senator to each territory.  This means that there would be 63 Senators total (62 before the creation of Nunavut in 1999), and that a party would need 32 Senate seats to control the Senate.   These Senators would have been elected using a single transferable vote proportional system (STV)- ensuring that parties would have won a share of Senate seats in each province that was roughly proportionate to the share of the vote in the province.  Because each territory only elects one Senator and STV does not work in single member districts, I assume the same party that one each territory’s House of Commons seat would have also one their Senate seat.* 

The Charlottetown model is similar to the structure used for the Australian Senate, though the fact that Australia assigns 12 Senate seats to each of its states instead of 6 reduces the proportion of votes needed in any state to win a Senate seat.  This means that it is easier for small parties to win Senate seats in Australia than it would be for small parties to win Senate seats under the Charlottetown model.  I also assume that the House of Commons would still determine which party forms government, and that the result of a Senate election would not change the party in government.

There are certainly other set-ups that could be used for a Canadian Senate.  Giving each province an equal number of Senators means that PEI ends up with the same number of Senators as Ontario despite the PEI’s population of under 200 000 people.  However, if the idea of the Senate is to provide a counterbalance to the power in the House of Commons of the provinces with the largest populations, it is hard to justify reducing the number of Senators in one province because of its small population.  The choice of an STV electoral system can also be debated.  It is awkward, though, to elect multiple members from a province using a multi-member plurality system (this system tends to over-reward the party with a plurality of the vote) and STV was the system proposed in Charlottetown and is the system used in Australia.

In estimating the number of seats that a party would win I assume that voters would have cast votes for the same parties in the Senate as they would have backed in the House of Commons.  One has to be careful about such an assumption as voters will behave differently under different electoral systems and voters may decide to vote for different parties for the House of Commons and Senate for a number of reasons.  However, without actual data from Senate elections, it is impossible to account for how voters might vote differently under a different system or how they might decide to split their vote between parties if given a chance to vote for both members of the House of Commons and Senate.  Real world election results would likely be slightly different than my estimates in this post, but I do not expect that they would be so different as to affect the broad conclusions that I draw from these estimates. 

The 2021 and 2019 Elections

My estimates for 2021 and 2019 Senate elections suggest a dynamic in the Senate that would be very similar to the House of Commons.  Figure 1 shows that in 2021 the Liberals would have a slight plurality in the Senate with 26 seats to the Conservatives’ 24.  Figure 2 shows that in 2019 this would be reversed with 25 Conservative Senate seats giving them a slight plurality over the Conservative’s 24.  In both cases, a Liberal alliance with the NDP (which would get 10 seats in 2021 and 9 in 2019) would be sufficient to give the Liberals control of the Senate.  This would put the Liberals in the same position that they were in from 2019 forward in the House of Commons, relying on the NDP to get legislation passed.  As in the House of Commons, the Liberals could always threaten to reach out to the Conservatives if the Liberals felt that the NDP was pulling it too much to the left.  The last few years have demonstrated that the Liberals are capable of building a quite comfortable working relationship with the NDP to get legislation passed.  Unlike in the House of Commons, the Bloc Québécois would not have enough Senate seats to help give the Liberals control of the Senate, but it is not clear that the Liberal party has a preference for working with the Bloc over the NDP in the House of Commons.  As a result, it is not likely that the Bloc’s weakness would matter to the Liberal’s ability to get their legislation through the Senate. 

Figure 1: The blue bar represents the Conservatives, the red the Liberals, the orange the NDP, the light blue the Bloc, and the green the Greens.

Figure 2: The blue bar represents the Conservatives, the red the Liberals, the orange the NDP, the light blue the Bloc, and the green the Greens.

A look at my estimates for the distribution of Senate seats by province (in the figures below) provides insight as to why an elected Senate does little to change Conservative fortunes.  The Conservatives are simply weaker than the Liberals in most provinces.  In only three provinces, Alberta, Manitoba, and Saskatchewan, do the Conservatives win more Senate seats than the Liberals.  By contrast the Liberals win more Senate seats than the Conservatives in Newfoundland, PEI, Nova Scotia, Quebec, Ontario, Northern Canada, and in 2021 in New Brunswick.  The two parties end up tied in British Columbia and in New Brunswick in 2019.  The problem that the Conservatives would have in a Senate election is similar to the problem that they have in elections to the House of Commons.  The party only really has a strong advantage in two provinces, Alberta and Saskatchewan.  In House of Commons elections that advantage is counterbalanced by the Liberals’ greater strength in the two most populous provinces, Ontario and Quebec.  In Senate elections Ontario and Quebec could no longer counterbalance the Conservative’s strength in Alberta and Saskatchewan.  Instead, the Liberals’ slight advantage over the Conservatives in much of Atlantic Canada would counterbalance Conservative strength in Alberta and Saskatchewan. 

Figure 3: The blue bar represents the Conservatives, the red the Liberals, the orange the NDP, the light blue the Bloc, and the green the Greens.

Figure 4: The blue bar represents the Conservatives, the red the Liberals, the orange the NDP, the light blue the Bloc, and the green the Greens.

The 2015 Election

I estimate the 2015 election would produce a Senate with a slightly different power dynamic than the 2015 House of Commons (shown in figure 5, below).  Where the Liberals won a majority in the House of Commons, I estimate they would have been two seats short of a majority in the Senate, at 30 seats.  This would have forced the Liberals to find compromises with either the Conservatives or the NDP to pass legislation.  However, with the Liberals just two seats short of a majority in the Senate, the Liberals would not need much support from those two parties to pass legislation, and they could likely play the two parties off against each other (presumably working with the NDP more often than the Conservatives).  While this Senate may have forced the Liberals to make slightly more compromises than they had to in reality, it is hard to imagine this Senate result keeping the Liberals from pursuing most of the agenda that put proposed in the 2015 election.

Figure 5: The blue bar represents the Conservatives, the red the Liberals, the orange the NDP, and the light blue the Bloc.

Figure 6 shows that the breakdown of seats by province shows that the Liberals’ Senate advantage in 2015 would have been driven by much the same dynamic as drove the Liberals’ advantage over the Conservatives in the 2019 and 2021 elections.  The Liberals’ advantage in Atlantic Canada essentially counterbalances the Conservatives’ advantage in Western Canada.  In this case though, the Liberals’ advantage in Atlantic Canada is stronger than the Conservatives’ advantage in the West.  The Liberals win four Senate seats in three of the four Atlantic provinces while the Conservatives only would have won four Senate seats in Alberta.  The big difference between this and the House of Commons result is that proportional electoral system allows the NDP to hold the Liberals to a minority.  That the NDP is able to win a seat in each Atlantic province and tie the Liberals’ seat shares in Quebec and British Columbia keeps the Liberals from getting majority control of the Senate.

Figure 6: The blue bar represents the Conservatives, the red the Liberals, the orange the NDP, and the light blue the Bloc.

An elected Senate would not have changed the Conservatives’ inability to influence legislation in the post-2015 election legislatures.  For all of the party’s claims to represent Western Canada as a region, they only would have had a decisive advantage over the Liberals and NDP put together in Alberta and Saskatchewan.  British Columbia and Manitoba are much more divided (in Manitoba because of the Liberals’ and NDP’s strength in Winnipeg) between the three parties.  Meanwhile, any advantage that the Conservatives had in the West gets matched by the Liberals’ advantage in the Atlantic provinces.  To the extent that the Conservatives have branded themselves as the defenders of the interests of Western Canada (and that is likely truer of Alberta and Saskatchewan than the rest of the West) they would have had little ability to use an elected Senate to force the Liberals to be more responsive to the concerns they were claiming to represent.  Rather, the Liberals would have had to make slightly more concessions to the NDP between 2015 and 2019 than they did in reality.  I suspect that this is hardly the outcome many advocates of Senate reform in Western Canada have in mind.

Addendum: It is unlikely that using a winner-take all system instead of STV would have helped the Conservatives.  In 2019 and 2021 the Liberals had a plurality of the vote in six provinces (Newfoundland, PEI, Nova Scotia, New Brunswick, Quebec, and Ontario) to the Conservatives’ four (Manitoba, Saskatchewan, Alberta, and BC).  In 2015 the Liberals had a plurality of the vote in eight provinces (Newfoundland, PEI, Nova Scotia, New Brunswick, Quebec, Ontario, Manitoba, and BC) to the Conservatives’ two (Saskatchewan and Alberta).  A winner-take all system would have thus left the Liberals with even more power in the Senate than a proportional system.

* It would be possible to determine territorial Senate seats using alternative vote, and that may produce slightly different outcomes in some elections.  At the same time alternative vote tends to produce roughly similar outcomes to single member plurality and it is difficult to simulate an alternative vote election without knowing voters’ second preferences.

How Confident Should the NDP Be? A Look at the Potential Costs of a Confidence and Supply Agreement to the NDP

In late March the federal Liberals and New Democratic Party (NDP) did something rarely done in Canadian politics.  In most cases in Canada when a party has a minority government it governs alone, seeking the support it needs from opposition parties to pass legislation on a bill-by-bill basis.  In March the Liberals moved away from this practice and entered in a confidence and supply agreement with the NDP.  The agreement commits the NDP to supporting the Liberals on confidence bills and motions (including the budget) in exchange for commitments by the Liberals to introduce programs important to the NDP such as dental care, pharmacare, and childcare (the CBC has a more detailed discussion of the agreement).  To be sure, the agreement is not a coalition, while the NDP will support the Liberals on key pieces of legislation, it is not receiving cabinet posts- a necessary condition for a coalition agreement.  Nonetheless, the Liberals and NDP are engaged in a level of cooperation that goes beyond what is usual in Canadian politics.

The confidence and supply agreement has led to speculation over the electoral consequences of the deal for the NDP.  Many commentators have pointed to the danger for the NDP that the Liberals will take credit for the more popular initiatives that come as a result of their agreement and, in doing so, take votes from the NDP.  While it is far from certain that a confidence and supply agreement or coalition would hurt its junior partner, it is worth looking at the proportion of votes that the Liberals would have to take from the NDP in order to move from a minority to a majority government.  I focus on the proportion of the vote that the Liberals would need to move from a minority to a majority government because the shift from a minority to a majority would end the Liberals’ need to rely on the NDP for support to pass legislation.  The influence the NDP has over government thus depends on whether the Liberals have a minority or majority government.  I show that if the Liberals manage to hold on to all of their voters, they need to take approximately 20% of the NDP’s vote in order to win a majority.  However, if the Liberals lose votes to the Conservatives, they need to take closer to 30% or 40% (depending on many votes they lose).

There is good reason to believe that the being the junior partner to a Liberal-NDP agreement is dangerous for the NDP.  The NDP has historically had to fight to keep voters who like NDP policies but do not think the NDP can win from voting Liberal.  The agreement with the Liberals may reinforce this idea, suggesting that the Liberals are the more competitive of the progressive parties.  Further, the agreement will see the Liberals implementing (at least in part) some of the progressive policies that the NDP campaigns on.  If the Liberals can successfully claim credit for these progressive policies they may be able to use them to win over voters from the NDP.  Finally, NDP voters who do not want a Liberal government may choose not to support an NDP that is responsible for helping to secure the Liberals’ position in government.  While these voters are unlikely to shift to the Conservatives, they could move to the Green party or to stay home in order to punish the NDP for the agreement with the Liberals.

A couple of agreements in other jurisdictions loom large for the NDP and the degree to which their electoral fortunes will be affected by the confidence and supply agreement.  After a minority parliament was elected in the United Kingdom in 2010 the Liberal Democrats entered into a coalition to give the David Cameron Conservatives a stable majority.  In the subsequent 2015 election the Liberal Democrat vote declined by 15 percentage points and the party lost 49 seats.  The Liberal Democrat case may be different than the NDP’s, however.  The Liberal Democrats are a centrist party, and as such, have a mix of voters that would prefer cooperation with the Conservatives and voters that would prefer cooperation with the Labour Party.  Cooperation with the Conservatives was thus likely to alienate some of the more left-leaning Liberal Democrat supporters.  By contrast, the NDP is a party that sits to the left of the Liberals.  There are thus likely to be few New Democrats who would prefer cooperation with a potential governing party other than the Liberals (essentially the Conservatives).  The NDP must only justify to its voters that cooperation with the Liberals is preferable to cooperation with no one, where the Liberal Democrats had to justify giving a Conservative government a working majority when a substantial share of its voters would have preferred a Labour government to a Conservative one.

The more informative example for the NDP is likely the 2017 and 2020 provincial elections in British Columbia.  In 2017 the Greens entered into a confidence and supply agreement with the NDP to allow the NDP to form a government after the 2017 BC election produced a minority situation.  In the subsequent 2020 election the BC NDP went on to win a large majority, claiming nearly 48% of the vote and 57 of 87 seats.  It is worth noting that, despite the BC NDP’s gains, the BC Green vote fell by just under 2 percentage points.  The BC Greens won only 2 seats as compared to the 3 they won in 2017, but the seat that switched from the Greens to the NDP was Oak Bay Gordon Head.  This was former leader Andrew Weaver’s seat, and Weaver had left the Greens before the election to sit as an independent. That it was no longer Andrew Weaver running for the Greens in Oak Bay Gordon Head may have contributed to the Green’s loss of the seat as much as the confidence and supply agreement.

It is finally worth noting Roy and Alcantara’s work on the impact a smaller parties support for a governing party in a minority situation has on both parties’ vote shares.  Using an experiment where different respondents are given different scenarios where smaller parties either choose to support or oppose a large party in a minority situation (and scenarios where the large party has a majority), Roy and Alcantara find that rather than punishing a small party for supporting a larger governing party, respondents rewarded it.*  This suggests that it is entirely possible that the NDP might not lose support at all from their cooperation with the Liberals. 

Having considered all of this, it is worth asking just how much support the Liberals would have to win from the NDP in order win a majority and deny the NDP future influence over policy.  To do this I look at the way shifts in vote share from the NDP to the Liberals and from the Liberals to the Conservatives affect the number of seats that each party wins.  I base my analysis on the results of the 2021 election.  I do this to isolate the impact of shifts that might relate to the confidence and supply agreement from other things that might affect the success of the Conservatives, Liberals, and NDP.  Between 2021 and the next election there will almost certainly be other factors that affect the votes of all three parties.  If the Liberals were to make large gains (or suffer large loses) as a result of other factors such as favourable economic conditions, a major scandal, or change in attitudes towards the party leader, it would be those factors and not the NDP’s decision to enter into a confidence and supply agreement with the Liberals that would be responsible for the NDP’s loss of influence over policy.**

In my analysis I consider three different scenarios.  One in which the Liberal party keeps all of their votes and adds varying proportions of NDP support, one in which the defection of centre-right Liberals to the Conservatives costs the Liberals 5% of their vote (and allows the Conservatives to pick up 5% of the Liberal vote), and one in which centre-right defections cost the Liberals 10% of their vote to the Conservatives.  For each scenario I estimate seat shares for a condition in the which the NDP loses 10% of their votes to the Liberals, one in which they lose 20%, one in which they lose 30%, and one in which they lose 40%.  Importantly, I look at percentages of each party’s vote instead of percentage point change in the vote.  The NDP losing 10% of their vote means losing 10% of the 17.8% of the vote they won in 2021 (1.8 percentage points) and going from a national vote share of 17.8% to 16%, not going from a national vote share of 17.8% to 7.8%. 

For simplicity’s sake, I assume that the proportion of losses will be the same across all ridings.  For example, for estimates where the NDP loses 10% of their vote share I assume that the party would lose 5 percentage points in all ridings where they won 50% of the vote and 4 percentage points in all ridings where they won 40% of the vote.  While this is unlikely to be the case in reality, it is impossible to predict which ridings would see the largest number of NDP to Liberal switching as a result of the confidence and supply agreement.  I also assume that the Bloc Québécois and Green party’s vote shares are unaffected (though the Bloc Québécois’ estimated number of seats is affected by changes in the strength of the Liberals***).

In the case where the Liberals hold on to all of their votes (shown in the figure below), I estimate that the Liberals need to win about 20% of the NDP vote to win the 170 seats they would need for a majority.  Winning just 10% of the NDP vote in this scenario wins the Liberals just 161 seats and leaves them short of a majority.  In this scenario that NDP would hold on to 22 seats should they only lose 10% of their vote to the Liberals and would still be able to extract concessions from the Liberals in exchange for propping up their government.  At a gain of 20% of the NDP vote, the Liberals would win a bare majority with 171 seats.  The NDP would suffer significant, but not catastrophic loses, winning 14 seats.  It is only when the Liberals win 30% of the NDP vote that they get into clear majority territory (184 seats) and where the NDP starts to suffer what would be considered catastrophic losses (being cut down to 4 seats).

The red line shows seat estimates for the Liberals, the blue line estimates for the Conservatives, the light blue line estimates for the Bloc Québécois, and the orange line estimates for the NDP.

If the Liberals lose votes to the Conservatives, it becomes much more difficult for them to win a majority.  In the figure below I estimate the number of seats that each party would win at different levels of NDP losses to the Liberals for a scenario in which the Liberals lose 5% of their votes to the Conservatives.  In this case, the Liberals need somewhere between 20% and 30% of the NDP vote in order to win a majority.  With losses of 5% to the Conservatives, I estimate that winning over 20% of the NDP vote only gets the Liberals to 160 seats.  The party would end up with 176 seats if they managed to win 30% of the NDP’s vote.  This would give them a majority, but not a very large one.  NDP loses in this scenario are similar to the estimates for the Liberals holding on to their vote.  I estimate the party will end up with 16 seats if they lose 20% of their vote to the Liberals and 4 if they lose 30% of their vote to the Liberals.

The red line shows seat estimates for the Liberals, the blue line estimates for the Conservatives, the light blue line estimates for the Bloc Québécois, and the orange line estimates for the NDP.

What happens if there are large numbers of centre-right Liberal voters who are pushed to the Conservatives by an agreement with the NDP?  In that case things become quite difficult for the Liberal party.  The figure below shows seat estimates for a scenario in which the Liberals lose 10% of their vote to the Conservatives.  In this scenario winning 20% of the NDP vote is not sufficient to keep the Liberals from falling to second place in seat share.  With 10% of the NDP vote the Liberals end up with 135 seats to the Conservatives 147 and with 20% of the NDP vote the Liberals end up with 143 seats to the Conservatives 145.  If the Liberals lose 10% of their vote to the Conservatives, they need to make up for the lost votes by gaining more than 20% of the NDP’s vote just to remain the largest party in the House of Commons.  Under these conditions it would take the Liberals winning 40% of the NDP vote just to get to a bare majority of 170 seats.  NDP seat losses in this scenario follow a very similar trajectory to the previous two scenarios.

The red line shows seat estimates for the Liberals, the blue line estimates for the Conservatives, the light blue line estimates for the Bloc Québécois, and the orange line estimates for the NDP.

These results suggest that the NDP can afford to lose some of their vote (though not 20% or more) to the Liberals without jeopardizing the influence over policy that they get under a Liberal minority government.  This is especially the case if cooperation with the NDP leads as few as 5% of Liberals to shift their support to the Conservatives (and the Liberals may have larger problems if they lose as many as 10% of their voters).  The NDP certainly must be careful, and they will need to work hard to make sure that they get at least some of the credit for the more popular initiatives that come out of the confidence and supply agreement.  The NDP can ill afford a collapse of the kind that the Liberal Democrats saw after their 2010 coalition with the Conservatives.  The BC Green party in 2017 and 2020 may, however, be more comparable case to the federal NDP, and the party could suffer the losses that the BC Greens did and still have a substantial influence over policy (provided the Liberals do not find a way to gain a large share of the vote from their centre-right competitor as the BC NDP did in 2020).  Indeed, Roy and Alcantara’s work suggests that there may even be an opportunity for the NDP to make gains from their cooperation with the Liberals.  The confidence and supply agreement that the NDP has entered into certainly carries some risk, but it should not be assumed that the agreement will lead to the demise of the NDP or even the end of the minority parliament situations that give the party a large amount of influence over policy.

*Roy, Jason and Christopher Alcantara. (2020). Winning and Keeping Power in Canadian Politics. Toronto: University of Toronto Press. (see chapter 6).

** It is also worth noting that there will be redistricting between the 2021 and the next election (assuming the Liberal/NDP deal holds until its conclusion in 2025).  It was impossible to account for this redistricting, however, as it has not been completed yet.

*** The Green seat share is not affected, so I exclude the Greens from the graphs to make them easier to read.

The Far-Right Pull: The Far-Right’s Influence on the Conservative Party is a Problem for Canadian Democracy

This past weekend was an ugly one in Ottawa.  The largely far-right trucker “convoy/protest” saw what Canadian Anti-Hate Network Chair Bernie Farber “worst display of Nazi and racist propaganda [he] had ever seen in [Canada].” The trucker convoy is unfortunately part of a worrying increase in far-right protests and riots across Europe and North America that include the January 6^th 2021 Capitol Hill riots in the United States.  As troubling as these events are, what may be more concerning is the response from the Conservative Party of Canada.  While the events in Ottawa of this past weekend are troubling in and of themselves, the reaction from the Conservative party is what creates the larger problem for Canadian democracy.  Much like the far-right in much of Europe and the United States, the far-right in Canada has limited ability to influence public policy on its own.  Rather, it threatens to shape policy by pulling the centre-right towards its views.  The more the Conservative Party is willing to pander to far-right activists, the greater the threat the far-right is.

With some exceptions, far-right parties and movements have demonstrated limited ability in European and North American democracies to win power or shape policy on their own.  While there are exceptions, such as Vicktor Orbán’s Fidesz Party in Hungary, most far-right parties that have managed to gain power have done so as junior partners in coalition governments with centre-right parties (the 2017-2019 coalition between the centre-right Austrian People’s Party and far-right Freedom Party stands out as an example).  In other cases, such as in Denmark, the Netherlands, and the United Kingdom, far-right parties have managed to get centre-right parties (and to a lesser degree centre-left parties) to co-opt some of their policies by threatening to take votes from the centre-right.  The best example of this is the UK Independence Party’s role in then Conservative PM David Cameron’s decision to hold a referendum on Britain’s membership in the European Union.  Finally, as illustrated by Donald Trump’s takeover of the American Republican party, far-right movements can gain power by co-opting existing centre-right parties, using their resources and brand to help win power. In all of these cases it is the willingness of a centre-right party to accept or court the support of far-right parties or voters that gives the far-right their influence.

The chances of the far-right winning power in Canada on its own is likely lower than in most other countries.  Canada’s single member plurality electoral system makes it difficult for new parties without geographically concentrated bases of support (and far-right voters do not appear geographically concentrated enough to overcome the barrier that a single member plurality electoral system presents) to win seats in parliament.  Indeed, despite winning almost 5% of the vote, the far-right People’s Party was shut out of parliament.  Any influence that the far-right is going to have on policy and broader politics in Canada at the federal level is thus likely to depend on its influence over the Conservative party.

This makes the response of the Conservatives to the trucker convoy disappointing.  Leader Erin O’Toole has been equivocal, meeting with some truckers and tweeting out his support for the convoy while also condemning extremism.  In a tweet supporting the convoy Andrew Scheer called Justin Trudeau “the biggest threat to freedom in Canada,” apparently ignoring the threats of violence against journalists and politicians that have been made by some on convoy present to freedom and democracy in Canada.  Other Conservative MPs such as Michael Cooper and Garnett Genuis have been supportive of the convoy also ignoring the threats of violence that have come from its members and its association with hateful groups.  At the very least there is a willful ignorance in the party of the extremism that has been associated with the convoy.  More likely, the party is trying to be just far-right enough to win over those taking part in the convoy while hoping it can disassociate itself enough from its more extremist elements that other voters will not notice how odious the group that it has embraced is. 

There are two possible outcomes from the Conservatives’ strategy, neither of which are good for Canadian democracy.  The more dangerous outcome is that the strategy is successful.  It is possible that the Conservatives are able to build a moderate right/far-right electoral coalition that is capable of winning power.  This is the kind of electoral coalition the American Republican Party built between 2008 and 2016, and which led Donald Trump to the Presidency in the United States.  As has been demonstrated in the United States, there is no guarantee that the moderate right can keep the far-right in check in such coalitions.  Though the far-right tends to be in the minority, it is often well enough organized and committed enough to wield a great deal of influence over parties that court it.  A Conservative party that wins power while courting far-right votes is likely to be a more extremist right government than past Canadian Conservative and Progressive Conservative governments, with likely troubling consequences for immigration and multiculturalism policy. 

While it is not impossible that a Conservative party that courts the far-right is successful, there are reasons to believe that such a party would struggle to win government.  Canada’s single member plurality electoral system makes it imperative that parties win swing ridings, and Canadian swing ridings tend to be culturally diverse.  Indeed, in the past two elections the Conservative party has failed to win more seats than the Liberals, despite winning more votes, because of their struggles in culturally diverse ridings in suburban Toronto and suburban Vancouver.  The Conservatives’ loss of Richmond Centre and Steveston Richmond East in the 2021 election coupled with the nearly 10-point Liberal margins over the Conservatives in most Brampton and Mississauga ridings suggest that, if anything, the Conservatives have been doing worse, not better, in each recent elections in the culturally diverse ridings they need to win if they want to form government.  An embrace of the far-right and the racism that was on display in Ottawa this past weekend is unlikely to help the Conservatives win these ridings.

An alternative outcome to Conservative success by embracing the far-right, a Conservative party that is relegated to opposition by an embrace of the far-right, is not good for Canadian democracy either.  Healthy democracies need multiple parties that can win government, if for no other reason than they keep the government in power in check.  Governments that are in power for a long time can see scandals pile up and can start to lose the trust of the electorate.  One might look to the Sponsorship scandal under Liberal governments of the late 1990s and early 2000s or the Senate expenses scandal under the Harper Conservatives in the lead up to the 2015 election as evidence that governments eventually need to be replaced.  Voters need to have reasonable alternatives to support for instances when government corruption or poor performance leads them to the conclusion that a new party in government is needed.  A Conservative party that disqualifies itself from government by embracing the far-right leaves Canadians with little ability to vote out a Liberal government they are unhappy with.  This makes a Liberal government, even if it is a minority government, less accountable and ultimately weakens the quality of Canadian democracy.*

The Conservative response to the convoy in Ottawa this past weekend is thus deeply troubling for Canadian democracy.  When centre-right parties draw a clear line between themselves and far-right movements there is at least some hope that the far-right movement can be isolated.  While an isolated far-right movement can still be a problem, especially when it threatens violence or engages in violence, its ability to influence policy making is quite limited.  Far-right movements become a problem for democracies when centre-right politicians and parties decide to either embrace them or refuse to condemn them.  In such cases the far-right movement either co-opts the centre-right party and uses it as a path to shape policy or the far-right movement makes the centre-right movement so unpalatable to most voters that it cannot win power.  Neither is good for a democracy.  If Canadian democracy is to remain healthy, the Conservative party needs to get better at identifying extremism and racism on the right, and it needs to be more willing to condemn both.

* This presumes that the NDP are strong enough to win government.  It is hard, however, to see how the NDP could win government without the complete collapse of the Liberal party as a lot of the moderate voters that the NDP would need to build a winning electoral coalition are Liberal voters.  The NDP is far more likely to be able to develop as a viable alternative government when the Conservatives are in power and the Liberals are a weak opposition party (as was the case between 2006 and 2011) than they are with a Liberal party that is strong enough to be in government.  There would thus have to be a very dramatic realignment of Canadian party politics for the NDP to be a viable alternative government in opposition to the Liberals.

The Campaign in Context: How 2021 Polls Compare to Parties’ Past Results

The first half of the 2021 federal election campaign gave pundits plenty to talk about.  In two weeks, the Liberals went from a solid lead over the Conservatives and a 50/50 shot at a majority to being consistently behind in the polls and facing speculation that the Conservatives may be able to form government.  Whenever a shift in the polls leads to a significant change in the likelihood that different parties win government there is a temptation to try to build a narrative around that change.  The decline in Liberal support is attributed to poor campaigning on the part of the Liberals while Conservative gains are seen as effective campaigning by that party.  There is a certain degree of truth to this.  Campaigns matter, and as a result, so do the strategic decisions that parties make during them.  At the same time campaigns happen in a broader context.  Short-term analysis of polling shifts should not lose sight of the broader context in which the campaign is occurring.  As much as a shift in polls can be the result of a party finding ways to bring new voters into their electoral coalition, it can also be a result of parties’ recovering the voters that usually support them.  Seen in the context of the past two decades of federal elections, none of the parties’ current polling numbers seem surprising.  This is particularly the case with respect to the Conservatives who are still polling slightly below most of their vote shares in elections since 2000.

There are a wide range of factors that shape vote choice.  Some of these factors are short term, fluctuating over the course of an election.  Some voters may consider different issues, such as which party is offering the best package of policies to deal with or recover from the pandemic or to respond to climate change.* Voters’ evaluations of leaders (as well as the party leaders themselves) will also change over time, leading some to change the party they support.  Other factors that shape voters’ decisions are harder to change.  Since Berlseon, Lazarfield, and McPhee’s 1950s study of voter behaviour in Presidential elections it has been accepted that voters’ social connections and the identities that shape them matter to voters’ decision making.**  Voters’ values or ideological leanings will also matter to voters’ decision making, and are unlikely to fluctuate over the course of a single campaign.  More stable than views on issues, but less stable than identities and values, voters’ partisan attachments (or partisan identity) will also shape how they vote.

Because voters’ decisions are in part shaped by factors that can change quickly over an election campaign, one should expect parties’ support to fluctuate to some degree from election to election and over the course of a campaign. At the same time, the factors that influence vote choice that are stable over time should ensure that parties’ support does not fluctuate so wildly that election results appear unrelated.  Understanding how parties’ polling numbers compare to past results can thus be useful to understanding fluctuations over the course of a campaign.  If a party goes into a campaign polling below its vote shares in previous elections, one might expect that it has a lot of room for easy growth.  In such circumstances there will be voters who might have social identities or values that predispose them to support the party but may have moved away for short-term reasons such as issues or attitudes towards the party leader.  If the party manages to shift perceptions of those short-term issues it should be able to win back those voters.  A party that is polling above its past results cannot do the same.  Indeed, it may be vulnerable to losing recently won-over voters who tend to vote for other parties.

To get a sense of how this is playing out in the 2021 election, I compared parties’ polling numbers to elections results over the past two decades, going back to the 2000 election.  For polling data, I used the projections for September 3^rd from 338 Canada, a polling aggregator, as well as the error margins provided by the site.  In the graphs below these are distinguished from the past election results by the slightly transparent dots and the dotted lines showing the error margins for each party.  For the 2000 election I combined the vote shares of the Progressive Conservatives and the Canadian Alliance to get a vote share for conservative parties that is comparable to Conservative vote share in subsequent elections.

Comparing current polling to past elections is most straight forward for the Conservatives.  Even with the Conservatives at 34%, the graph below shows that they are a bit below most of their election results over the past two elections (though they are within the margin of error of such results).  Only in two elections have the Conservatives done worse than they are currently polling.  In one case, 2004, the party’s poor result was likely caused in part by the just formed Conservative party’s inability to hold on to more moderate Progressive Conservative voters worried that the new party was too heavily influenced by the more right-wing Canadian Alliance.  The other election where the Conservatives did worse than they are currently polling was 2015, when almost 10 years of Harper government made the party unpopular (parties that have been in government for a long time often become unpopular).

The Conservatives are in blue, the Liberals in red, and the NDP in orange.

A comparison of the Conservatives’ current numbers to past results, suggests that the gains the Conservatives made over the first two weeks were largely a result of bringing voters who had voted Conservative in previous elections back to the party.  The 28% the Conservatives were polling at when the election was called was probably too low for the party.  The party started the election with the opportunity to make a lot of easy gains by winning over voters who tend to support the Conservatives but who had shifted away from the party in the lead up to the election.  The party may still have some room to grow, they are still very slightly below where they finished the 2019 election and somewhat below their vote shares in the 2006-2011 period when Stephen Harper was forming governments.  That being said, at this point the party is likely going from winning back disaffected Conservative voters to trying to win swing voters who were persuaded to vote Conservatives by Harper but who may also be reasonably willing to vote Liberal if the Liberals’ can find the right message, leader, or policy commitment.

It is harder to know what to make of the comparison between the Liberals’ and NDP’s polling and their past election results.  This comparison is complicated by the Liberals’ decline between 2000 and 2011 and their striking revival in 2015.  The Liberal’s current polling numbers match the party’s vote share at the beginning of the party’s decline in 2006.  Further, one can draw a near straight line from the Liberals’ peak in 2015, through their 2019 result, to where the party is polling now.  How to interpret the Liberals’ current numbers thus depends on what one makes of the 2000-2011 period and the party’s revival in 2015.  The unpopularity of the Conservatives in 2015 meant that election was always likely to be a high point for the Liberals and that the party would decline at least a bit from their 2015 vote share.  However, if the 2015 election involved a genuine revival of the party, the Liberals may have some easy gains to make in the next couple of weeks.  There may be voters in the electorate that tend to vote Liberal, at least when the party makes a reasonably compelling case for itself, who could be brought back to the party. 

On the other hand, if one views the 2004-2011 period as part of a much larger structural decline in Liberal support, the Liberals may be in more trouble.  There is broader trend, shown in the graph below, in which Liberal vote share has declined over the course of the twentieth century and early part of the twenty-first century.  The increasing difficulty of building broad regional coalitions, simultaneously encroachment of the Liberals left and right flanks by the NDP and Conservatives respectively, and the growing fragmentation of party systems globally can all be seen as party responsible for the Liberals’ decline.  It is not clear that any of those factors are any less present today than they were a couple of decades ago.  Indeed, despite the Liberals’ 2015 revival, their 2019 vote share was right on trend line for the Liberals’ decline over time.  If the Liberals are still in decline, their low poll numbers might not suggest growth potential.  Rather, it may be that the Liberals’ 2015 revival was an aberration and that one should expect the Liberals’ support to fall to a point where the party is closer to its 2008 or even 2011 vote shares.  It is hard to know with any degree of certainty whether it makes more sense to compare the Liberals’ current numbers to the past couple of elections (which would be optimistic for the party) or their longer-term decline (which would be more pessimistic).

What to make of the NDP’s polling numbers depend in large part on what one makes of the Liberals’.  The NDP is polling above their election results for every election except for 2011 (though they are only slightly above their 2015 numbers).  On the face of it, this would suggest the party will have difficulty growing further, as to do so would require reaching out to voters who are less and less predisposed to vote NDP.  Indeed, the party might have to worry about losing support as past Liberal voters move from the NDP back to the Liberals.  This fits with a story where the Liberals are not in long-term decline, but rather underperforming past level of support.

If, however, the Liberals are in longer-term decline, NDP support may stay where it is or even grow a bit.  A long-term Liberal decline is likely driven at least in part by encroachment by the Conservatives and the NDP on the Liberals’ right and left flanks.  This would suggest that the Liberals’ decline should be to the benefit of the NDP as they pick up some of the left voters that are moving away from the party. If this is what is happening, it may not be unreasonable to expect the NDP to outperform their past election results.

Too often during campaigns pundits and other election observers focus only on short term shifts in polls.  This is valuable in capturing the short-term shifts over the course of a campaign, but it can miss a broader context that is useful to understanding the way polls are shifting.  In the case of the Conservatives, ignoring past election results may lead one to miss the how much below their usual election results the Conservatives were and how much the Conservatives could grow by simply winning back voters that had supported them in the past.  Even at 34%, the Conservatives may still have a little bit of space for that easy growth.  Making sense of the Liberals’ and NDP’s polling numbers in broader context is more difficult and depends on whether one interprets the 2015 election as a revival of the Liberals, or an aberration set against a broader declining trend.  That the context behind shifts in Liberal and NDP vote share over time is less clear should lead one to be more uncertain about the potential for growth (or decline) in Liberal and NDP vote share over the last few weeks of this election.

*While climate change is a long-term problem, parties’ climate change policies fluctuate from election to election and so voters’ views as to which party is best able to grapple with climate change may fluctuate from election to election. 

**Berleson, Bernard R, Paul F. Lazersfield, and William N. McPhee. (1954). Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: The University of Chicago Press.

The NDP’s Urban Problem: For a Left Party the Federal NDP is Remarkably Weak in Canada’s Urban Centres

Through the NDP’s rise and fall between 2011 and 2015 much has been made of the party’s successes and failures in Quebec.  The party’s success in the province in 2011 propelled it to official status while the loss of large numbers of seats in the province to the Liberals in 2015 saw the party fall back to third place.  As a province that is often well to the left of others on issues ranging from the welfare state to the environment to LGTBQ rights and gender equality, winning in Quebec is central to success of any left party in Canadian federal politics.  The focus on Quebec, however, has led to an under-appreciation of the NDP’s struggles in Canada’s largest cities.  Urban progressive voters have become a central component of any major left party’s coalition.  Yet, the NDP has struggled to build a strong base of support in Canada’s largest cities.  Indeed, the 2015 and 2019 elections have seen the Liberal open up a large lead over the NDP in major cities.

In this post I look at the vote shares in six of Canada’s larger urban centres between 2000 and 2019: Montreal, Ottawa, Toronto, Brampton/Mississauga/Oakville (grouped together), Winnipeg, and the Greater Vancouver Regional District (Vancouver along with suburbs such as Burnaby, Richmond, and Surrey).*  I excluded Calgary and Edmonton from my analysis because the strength of the federal Conservatives in Alberta makes it unlikely that even urban Alberta would form a crucial component of any left or even centrist party’s electoral coalition.  In this post I include analysis for urban areas as a whole as well as analysis focusing on individual cities.

Analysis that looks at vote shares in large cities as a whole shows the NDP was usually the weakest of the three parties in large cities.  The graph below shows that in every election except for 2008 and 2011 the NDP trailed the Liberals by over 20 points in major urban centres, and in every election the NDP did worse than the Conservatives.  Even in 2011, when the NDP was at its strongest, the NDP only held a small advantage over the Liberals in large cities.  The 30% of the vote that the NDP won in major cities in 2011 matched the party’s 31% nationally.  In contrast, the Liberals’ 2011 collapse was not nearly as steep in cities as in the rest of the country.  While the Liberals won only 19% of the vote Canada-wide in 2011, they managed 28% of the vote in large cities.  When left leaning voters across the country (and particularly in medium sized urban centres such as Hamilton and Windsor) opted for the NDP in 2011, left leaning voters in the largest cities were much more split between the NDP and Liberals. 

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

The Liberals’ resurgence in 2015 saw the party outperform the NDP in major cities by a larger margin than in any election since 2000, holding an almost 30 percentage point advantage.  While Liberal vote share in major cities declined a bit in the 2019 election, so did the NDP’s (with the Green party, and in Montreal the Bloc Quebecois, making gains).  The 2019 election has thus left the NDP with a large gap to make up if it wants to compete with the Liberals, or even the Conservatives, in major cities.

Shifts in vote share between 2000 and 2019 in Montreal follow a pattern that mirrors the rest of Quebec.  The graph below shows NDP vote share in Quebec’s largest city surging in 2011 to match the NDP’s broader growth in the province.  Then, in 2015 and 2019, NDP vote share in Montreal collapses just as the party’s vote in the rest of the province did.  Notably, the Liberals still managed to do well in the city as the party declined in the 2000-2008 period, in part because of their support amongst Anglophone Quebecers.  This may also explain why the Liberals saw slight gains in Montreal between 2015 and 2019 despite experiencing slight losses in the province as a whole.

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Light Blue= Bloc Quebecois, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

Shifts in vote share in Toronto tell a different story.  The graph below shows that to the extent the NDP made a breakthrough in the city in 2011, it was much more limited than their breakthrough in Quebec.  Despite falling to third place nationally, the Liberals were still the preferred party of the plurality of Torontonians in 2011.  The NDP, meanwhile only finished higher than third in the city once during this period (and that was in 2004, largely as a result of the weakness of the Conservatives).  Furthermore, unlike in much of the rest of English Canada, in Toronto the 2019 election saw the Liberals hold on to the gains that they had made in 2015.  The NDP, by contrast, saw their vote share fall slightly between 2015 and 2019.

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

Vancouver and the surrounding area (GVRD) is the NDP’s strongest major city, and the Liberal’s weakest.  Yet even here, the NDP never finished first in the 2000-2019 period, in part because of Conservative strength in suburban GVRD ridings such as Richmond and South Surrey White Rock.  Indeed, the NDP was the third-place party in the Vancouver area in every election I looked at except for 2011.  In Vancouver, however, the NDP may have reason to be more optimistic about Vancouver than about Montreal or Toronto.  Unlike in Montreal and Toronto, the NDP did not see its vote share in Vancouver decline between 2015 and 2019.  Instead, the Liberals’ saw their vote share decline in part due to the growth of the Green party, in part due to slight gains made by the Conservatives, and in part because Jody Wilson-Raybould would have held on to a significant number of her supporters in Vancouver Granville as she went from being a Liberal to an independent candidate.

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

There are a couple points worth noting with respect to the remaining three urban areas that I looked at (all shown below).  The trend in Ottawa matches that of Toronto, though with a stronger Conservative party (which made gains in the Ottawa suburbs a couple of elections before it started to make gains in the Toronto suburbs).  In Brampton, Mississauga, and Oakville the NDP has been quite weak, even in 2011.  The selection of a leader, who represented a Brampton/Mississauga riding (Bramalea-Gore-Malton) in the Ontario legislature from 2011 to 2017, Jagmeet Singh, seems to have done little to help the NDP in the region.  The NDP made only modest gains in Brampton/Mississauga/Oakville between 2015 and 2019, going from 11% of the vote to 13%.  Finally, in Winnipeg the NDP did not see a surge in support in 2011, but did see its vote share decline between 2011 and 2015.  It has seen its support rebound in the city in the 2019 election, but not quite to pre-2015 levels.

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

Blue= Conservatives, Red= Liberals, Orange=NDP, Green= Green Party, Teal= Canadian Alliance, Medium Blue= Progressive Conservatives.

The NDP thus goes into the 2021 election in an awkward position.  Not only does it struggle to win votes in one of Canada’s most left leaning provinces, Quebec, but it also is well behind its Liberal party competitor in the major cities that have become important bases of support for left parties in most other countries.  As the NDP seeks to recover from losses in both the 2011 and 2015 election it needs to consider not just how it might recapture some of the success it had in Quebec in 2011, but also about how it can pose a more serious challenge to the Liberals in Toronto and Vancouver (though the NDP is much closer to challenging the Liberals in the latter).  Without success in Canada’s major urban centres, it is hard to see how a left party like the NDP goes beyond being anything but a third party.**

*A full list of ridings that I included with each city is available here.

**I should note that I wrote a series of posts on the provincial New Democrat parties in Alberta, Saskatchewan, and Manitoba (links here) arguing that the NDP was in better shape in Alberta and Manitoba than in Saskatchewan because of the Alberta’s and Manitoba’s larger urban populations.  Part of the reason the NDP does so well in urban areas in prairie provincial politics is because the prairie provincial Liberal parties are very weak and therefore do not pose the same threat to NDP support in urban areas as the federal Liberal party does.

There Will Be Outliers: How to Think About Polling Outliers and the 2021 Election

With every election comes a renewed focus on polls and on how voters’ views are changing over the course of the campaign.  The 2021 Canadian federal election promises to be no different.  That the election comes after two years of minority government has meant that polls were receiving a great deal of attention even before the election was called.  With the renewed attention to polling comes a fascination with polls that show significantly different results than other polls, sometimes referred to as outliers.  They provide an opportunity to debate whether public opinion is changing, whether the rest of the polls have things “wrong,” or whether the pollster with the outlier has made mistakes in conducting the poll.  Fixation on outliers, or for that matter any individual poll, is a mistake.  Because polls are based off of random samples of the population, it is possible for an individual poll to get results that are very different from the views of the population through random chance.  Rather than trying to determine whether any given poll is an outlier, observes of Canadian politics should focus on polling aggregators treating each poll, outlier or not, as a small piece of a much larger picture.

Even when pollsters do everything right it is possible for them to end up with estimates that are very different from the views of the population.  To estimate the public’s views, pollsters contact a random sample of eligible voters, often weighting responses to compensate for ways in which their sample may have deviated from known demographics such as the regional distribution of voters.  Because polls use random samples, it is unlikely that they will be perfectly accurate.  While sampling Canadians at random should mean that respondents to the poll have views that are similar to those of all Canadians, the chances of any pollster getting a sample that perfectly matches the population is low.  This is why most polls will say that their results are accurate within 2 or 3 percentage points.  The pollsters are acknowledging that a random sample is rarely a perfect match to the population.  But while most random samples with be reasonably close matches to the population, every so often drawing names at random will lead a pollster to a sample that is very different from the population.  This is why pollsters do not just say that their results are accurate to within 2 or 3 percentage points, they also say that they are that accurate only 19 times out of 20.  The pollster essentially expects to get the “wrong” result one time out of 20.

One might think of this the same way one thinks about coin tossing.  If one where to toss a coin 10 times every morning one would expect to get close to five heads most of time, but not all of the time.  Sometimes one would get 4 heads, sometimes 6.  Every so often one would get eight, nine, or ten heads, even with a fair coin.  In the same way, if one takes enough random samples of the population, one is bound to eventually get a sample that has views that are significantly different from the population through pure random chance.  Given the number of polls that are run over the course of an election by the range of different pollsters, there are bound to be a few pollsters that get an odd result thorough random chance.

The possibility that any single poll may be an outlier makes polling aggregators a better way of getting a sense of how well the different parties are doing at any point during an election campaign.  Aggregators such as the CBC’s Poll Tracker or PJ Fournier 338 Canada estimate party support by taking an average of publicly available polls (usually weighting polls by quality and sample size).  By taking an average of polls they include more data than any one given poll has and significantly reduce the chances of showing misleading results through random chance.  An aggregator makes any single poll just one data point amongst a larger picture created by a large number of polls.

Aggregators are important, not because they make it easier to spot outliers, but rather because they make identifying outliers less important.  A poll may be different from others because of bad methodology, because it is picking up on changing trend, because it has randomly ended up with an odd sample, or because there is a great deal of variation in the public’s views producing outliers in multiple different directions.  It is often hard to tell which of these is the reason for an individual poll being an outlier until one has a large number of other polls to compare it to.  Once those polls are available, the outlier becomes unimportant.  If the poll is an outlier, it will have little influence on the trend in the aggregator, little influence on the predictions of those that are following the trend, and thus be of limited importance.  If the poll is the start of the trend, this will be reflected in additional polls and as a result, the trend in the aggregator.

There is further a danger in shaming pollsters for publishing polls that are different from others.  This can lead to “herding” a practice where pollsters adjust the models that they use (pollsters will often weight respondent’s answers to reflect things like the likelihood of the respondent voting or to compensate for under-sampled or oversampled demographic groups) after they get the responses in order to come to an estimate that is closer to other polls.  This is a problem because it is important to see the outliers in polling averages.  If there are a lot of outliers in multiple directions, that would suggest that getting a good sense of what voters are thinking is difficult and that we should treat polling averages with greater uncertainty than we should if there are few outliers.  Further, if a poll is the start of a trend and not an outlier, herding may hide what is in actuality an important shift in voters’ views.  If enough pollsters herd, unaware that others are herding, the change in trend could be missed entirely or look like it happened later than it did in actuality.

None of this is to say pollsters should not be criticized for their methods.  Pollsters certainly can end up with bad or outdated methods that lead their polls to make mistakes when estimating public opinion.  However, critiques of polls should focus on their methods as opposed to whether they get results that are similar to those of other polls.  In the middle of an election when we do not know what the result will be, the possibility of randomly getting a sample very different from the population makes comparing a poll’s estimates to other polls a bad way of judging a poll’s quality.

Getting a good sense of what polling data is suggesting can often be difficult.  That polls rely on random samples means there will inevitably be variation in the estimates that different polls come up with.  An individual poll might differ from others for a wide range of reasons which may or may not be important to those trying to get an accurate sense of what voters are thinking.  This is a reason to use polling aggregators to avoid minimize the influence any one poll has on the way the one sees an election.  Using aggregators allows one to avoid speculating as to why poll is different from others and as to whether that difference is important.