Introduction
The years 2006–2015 marked a remarkable period of transition in Canadian politics, with the Liberals and the New Democratic Party (NDP) trading places as the main competitor to the Conservatives. Much of the analyses on the rise of the NDP have focused on national-level factors such as the relative popularity of party leaders, the extent to which NDP policies and values matched those of voters, and the professionalization of the party (Clarke et al., Reference Clarke, Gravelle, Scotto, Stewart and Reifler2017; Fournier et al., Reference Fournier, Cutler, Soroka, Stolle and Bélanger2013; McGrane, Reference McGrane2019). Cross (Reference Cross2016), however, shows that revitalization of Liberal local organizations and campaigns played an important role in the party's 2015 resurgence. Cross's finding raises a question as to whether there is a similar correlation between NDP local campaign strength and NDP success.
The intuitive conclusion from Cross's (Reference Cross2016) work is that if local organization revitalization helped the Liberals’ rise from third to first in 2015, the NDP should have been hurt by its relative organizational weakness. Further, that Liberal local organizations needed revitalization in the lead-up to 2015 implies that the NDP may have benefited from a local organization advantage in elections between 2006 and 2015. Existing work on Canadian local campaign spending tends to focus on spending's impact on vote shares (Carty and Eagles, Reference Carty and Eagles1999, Reference Carty and Eagles2005; Eagles, Reference Eagles2005). In single member plurality systems, however, votes in marginal ridings have a much greater influence on election outcomes than votes in uncompetitive ridings. Because smaller parties like the NDP are competitive in fewer ridings, understanding the way that the single member plurality electoral system shapes the relationship between NDP local spending and the number of seats the NDP wins is particularly important.
The article finds that overall local campaign spending by the NDP was consistently lower than its opponents’ spending between 2006 and 2015. Local spending by the NDP is correlated with vote shares, but this article finds little evidence that the NDP's persistent spending disadvantage cost the party seats. This is because the NDP outspent its opponents in marginal ridings. The correlation between spending and vote share suggests spending helped the NDP win close seats. The correlation between NDP spending and vote share, however, was not large enough to suggest that the NDP could have won seats it lost by large margins even if it had managed to spend as much as its opponents in those ridings. Further, this article shows that the bulk of the NDP's local campaign revenue in marginal ridings (as well as in lost-cause and safe ridings) came from either electoral district associations (EDAs) or donations to the local campaign. This suggests that local organization strength, not support from the national party, is largely responsible for the NDP's ability to outspend its opponents in marginal ridings.
One must be careful about drawing strong conclusions about causal relationships from this work. The relationship between local spending and riding vote share is likely bidirectional. While parties may win more votes in ridings that they spend more money in, they may also spend more money in ridings where they think they are more likely to win votes. This article uses controls for parties’ previous vote share in ridings to try to control for parties’ expectations of success, but this is only a limited control. Parties use a range of considerations to determine whether a riding is competitive, and without getting into the heads of party staffers, it is impossible to fully control for parties’ expectations of success in different ridings. The endogeneity issues that complicate any analysis of spending's impact on vote share can thus not be completely disentangled in this article.
Nonetheless, the patterns that emerge in this article can tell us important things about local spending and the NDP over the 2006–2015 period. First, the NDP's consistent overall spending disadvantage between 2006 and 2015, coupled with the lack of a clear difference in the way spending affected seat share, suggests that local spending cannot explain much of the NDP's rise or fall between 2006 and 2015. Second, this article demonstrates the potential for parties to mitigate an overall local spending disadvantage if their spending is concentrated in marginal ridings. Third, the high proportion of revenue that campaigns in marginal ridings receive from either EDAs or local donations suggests that high levels of spending in marginal ridings may not always be driven by decisions made by the national party.
Literature Review
The NDP from 2006 to 2015 in context
Understanding local spending by the NDP between 2006 and 2015 can provide important insight into local spending's impact on elections. The period saw two elections (2006 and 2008) in which the NDP made small but substantial gains, one in 2011 in which it rose to official opposition, and one in 2015 in which the Liberals pushed it back to third place. These fluctuations have often been attributed to national-level factors such as the NDP's shift toward the centre (Delacourt, Reference Delacourt2016; McGrane, Reference McGrane2019), the popularity of leaders such as Jack Layton and Justin Trudeau (Fournier et al., Reference Fournier, Cutler, Soroka, Stolle and Bélanger2013; Clark et al., 2017) and the NDP's reaction to Stephen Harper's commitment to ban niqabs from citizenship ceremonies (Stephenson et al., Reference Stephenson, Lawlor, Cross, Blais and Gidengil2019). Cross (Reference Cross2016), however, argues that the revitalization of Liberal local organizations was central to the party's 2015 resurgence. This raises the question as to whether local spending affected NDP success.
A look at local spending and the NDP is interesting for two reasons. First, as a smaller party, the NDP is competitive in fewer ridings, and so how NDP local spending is distributed is likely to be particularly important to the party's success. Second, the party's sudden growth in Quebec in 2011, followed by its sudden decline in 2015, provides insight into how local spending affects parties during elections when party support is volatile. The slow growth of the NDP in the 2006 and 2008 elections provides insight into local spending effects in more stable elections.
EDAs, campaigns and election results
Cross's (Reference Cross2016) work matches an extensive literature on the importance of EDAs and campaigns to election outcomes. Carty (Reference Carty2002) and Cross and Young (Reference Cross and Young2011) consider local campaign autonomy and intensity. Carty and Eagles (Reference Carty and Eagles1999, Reference Carty and Eagles2005) show that local campaign strength affects Canadian elections, while Eagles (Reference Eagles2005) shows the same with respect to spending. This is consistent with work looking at the UK (Cutts and Shryane, Reference Cutts and Shryane2006; Fisher et al., Reference Fisher, Cutts, Fieldhouse and Rottweiler2019; Johnston and Pattie, Reference Johnston and Pattie2013; Johnston et al., Reference Johnston, Pattie, Fisher and Cutts2013; Pattie and Johnston, Reference Pattie and Johnston2003, Reference Pattie and Johnston2009) and even Germany with its proportional representation electoral system (Gschwend and Zittel, Reference Gschwend and Zittel2015).
There is an extensive literature measuring the effect of different components of local campaigns on Canadian elections. Cutler (Reference Cutler2002) shows that voters are responsive to the local economic conditions, while Loewen et al. (Reference Loewen, Heroux-Legault and de Miguel2015) show that the linguistic makeup of an area affects the link between voters’ views on nationalism and their support for the Bloc Québécois (BQ). Belanger et al. (Reference Bélanger, Kenneth Carty and Eagles2003) and Carty and Eagles (Reference Eagles2005) show that leaders’ visits to a riding increase a party's vote share, though the effect they find is limited to the Liberals and NDP. Work on candidates suggests that perceptions of the local candidate are decisive for 4 per cent (Stevens et al., Reference Stevens, Islam, de Geus, Goldberg, McAndrews, Mierke-Zatwarnicki, Loewen and Rubenson2019) or 5 per cent (Blais et al., Reference Blais, Gidengil, Dobrzynska, Nevitte and Nadeau2003) of voters, while Roy and Alcantara (Reference Roy and Alcantara2015) find that a strong local candidate can increase a party's vote share by up to 10 percentage points. Blais and Daoust (Reference Blais and Daoust2017) show that a significant number of voters (though not a majority) will support their preferred local candidate over their preferred party.
While it is clear from the existing literature that local campaign spending has an impact on party's vote share at the riding level, much less work has been done on the impact spending has on seat shares. It is not necessarily the case that because local spending affects a party's vote share that it will affect a party's seat share. In single member plurality systems, a party's success depends on the way that votes are distributed across ridings. The impact of local spending thus depends on where a party outspends its opponents.
This is particularly the case for a smaller party such as the NDP. The NDP is competitive in fewer ridings than the Conservatives or Liberals as shown in Appendix A. That the NDP often had fewer ridings that were close races suggests that the distribution of local spending is particularly important to the spending's impact on its success.
Other elements of local campaigning—such as where parties encourage star candidates to run, which local campaigns are best able to recruit volunteers, and visits by the party leader—are also likely to shape the way that local campaigns affect election results. It is, however, harder to measure these factors, particularly when looking at past elections for which candidate surveys can no longer be conducted. The detailed information available on party spending makes it particularly useful in testing the way the distribution of NDP campaign strength interacted with the single member plurality electoral system to affect the NDP success.
This article also speaks to a growing literature on national party financial transfers to local campaigns (see Colleto et al., Reference Colleto, Jansen and Young2011; Cross et al., Reference Cross2020; Currie-Wood, Reference Currie-Wood2020). It builds upon existing work by comparing the influence national parties and EDAs have on the distribution of local spending between marginal, lost-cause and safe ridings.
Theory
This article considers three possible relationships between local campaign spending and NDP success. The first is that spending mirrored the rise and fall of the party, contributing to its growth from 2006 to 2011 and its decline in 2015. The second possibility considered is that the NDP was consistently outspent by its opponents and that the party's growth was the result of other factors that allowed it to overcome this disadvantage. Last, the article considers the possibility that the NDP faced an overall disadvantage in local spending but that the party's ability to outspend its opponents in close ridings prevented that disadvantage from costing it seats. These are certainly not the only possible relationships between NDP spending and seat shares, but they are the ones I see as most likely and thus consider in this article.
Theory 1: EDAs and campaigns contribute to the NDP's rise and fall
The possibility that follows most intuitively from Cross's (Reference Cross2016) work is that a spending disadvantage cost the NDP seats in 2015 and that an advantage, or a reduced disadvantage, played a role in its success between 2006 and 2011.
Between 2006 and 2011, the growing strength of the NDP and its increased professionalization should have increased the ability of its EDAs to raise revenue. The party's increased popularity should have led to more donors, as well as members and volunteers who could contribute to party fundraising. This should have helped local EDAs, as they would have received at least some of the increased donations and would have had more volunteers and members to help with fundraising. Professionalization under Jack Layton should have also helped the party, as McGrane (Reference McGrane2019) demonstrates that such professionalization involved increased national party support for EDAs and local campaigns. This, in turn, should have allowed the party to spend more, decreasing the gap in local spending between the NDP and its competitors. The existing academic work (Carty and Eagles, Reference Carty and Eagles1999, Reference Carty and Eagles2005; Eagles, Reference Eagles2005) suggests that this decreased gap should have helped the NDP win votes.
It is less clear that there should have been a decline in NDP EDAs’ fundraising capacity between 2011 and 2015. McGrane (Reference McGrane2019) shows that the professionalization that began under Layton continued when Thomas Mulcair took over as leader in 2012. In addition, the NDP's official opposition status and increased number of incumbents going into the 2015 election should have put NDP local associations in a better position to fundraise than in previous elections. Appearing as the most viable alternative to the Conservatives in the lead-up to 2015 should have further helped fundraising. While the party would have been hurt by the loss of the per-vote subsidy, that should have affected the national party more than local EDAs and campaigns.
Nonetheless, it is possible that the revitalization of Liberal associations between 2011 and 2015 discussed by Cross (Reference Cross2016) put the NDP at a disadvantage in local spending even if the NDP's fundraising did not decline. Spending relative to other parties matters more to vote share than absolute spending, and so a party can be hurt by an opponent's improved fundraising even if its own fundraising remains constant.
H1a: Growing NDP local spending between 2006 and 2011 helped the NDP win seats, while an increased local spending disadvantage cost the NDP seats in 2015.
As the NDP professionalized under Layton, one might expect the national party to play a larger role in transferring funds to local campaigns in the most competitive ridings. Indeed, Currie-Wood (Reference Currie-Wood2020) demonstrates that the Conservative's party modern structure involved a significant amount of control by the national party over transfers to EDAs.
Prior to Layton taking over the party leadership, the NDP's federal structure meant that federal EDAs were run largely by members of the provincial party. As a result, the strength of EDAs was more dependent on the strength of the provincial party than on how competitive the federal party was in the riding (McGrane, Reference McGrane2019). Further, the NDP's roots as a mass party (as opposed to the cadre party structure historically present in the Progressive Conservative and Liberal parties) led it to concentrate its spending in marginal ridings less than its competitors in elections prior to 2004 (Sayers, Reference Sayers1999).
Under Layton the party transitioned to a more professional and catch-all structure that involved greater investment in EDAs, replacing the influence that provincial wings of the party had over EDAs with national party influence (McGrane, Reference McGrane2019). Professionalization should have given the national party a greater role in identifying and supporting campaigns in marginal ridings. This, in turn, should have increased national party transfers to local campaigns and made the support that local campaigns received less dependent on whether they were in provinces with a strong provincial NDP, such as British Columbia or Saskatchewan.
H1b: Professionalization led the party to commit a greater proportion of its local spending to competitive ridings in each election year.
This hypothesis assumes the NDP's national organization was both relatively unconstrained in its ability to adjust its transfers to local campaigns and able to accurately predict which ridings would be competitive. The need to appear competitive across the country or satisfy high-profile candidates and incumbents may have constrained the party's ability to focus its support on marginal ridings. Sudden swings in support, as happened in Quebec in 2011 and 2015, may have also hindered the party's ability to accurately identify marginal ridings.
Theory 2: Relatively weak local campaigns were a constant drag on NDP vote share
A second possibility is that NDP local campaigns were consistently outspent by their opponents and that this hurt the party. NDP local campaigns may have had more difficulty raising money than their Conservative and Liberal competitors, and this may not have changed between 2006 and 2015. The NDP's rise to official opposition status during the 2011 election was sudden enough that it may not have given local campaigns the ability to capitalize on the party's newfound competitiveness when fundraising. The existing research (Carty and Eagles, Reference Carty and Eagles1999, Reference Carty and Eagles2005; Eagles, Reference Eagles2005) suggests that such a spending disadvantage should have cost the party votes. As will be discussed in theory 3, the relationship between votes and seats in a single member plurality system is complicated, but one might expect that such lost votes would also cost the party seats.
The corollary of this theory is that the growth of the NDP was a result of national-level factors, such as the popularity of Jack Layton, moderation of the party's platform and the increased professionalization of the NDP's national campaign (Fournier et al., Reference Fournier, Cutler, Soroka, Stolle and Bélanger2013; McGrane, Reference McGrane, Gagnon and Brian Tanguay2017, Reference McGrane2019). Changes in these factors would also be responsible for the NDP's decline in 2015.
That national-level (or at least provincial-level) factors could overcome local spending is most likely in Quebec in 2011. The NDP made a breakthrough in Quebec in 2011, winning 59 of 75 seats on 43 per cent of the vote. It did this despite having just one incumbent in the province and running “paper candidates” in many ridings. While the NDP's success in Quebec may be an outlier, it suggests a need to consider the possibility that local spending had little impact on NDP success.
H2: The NDP was at a constant disadvantage in local spending, and this disadvantage cost the party votes and seats.
Theory 3: Strength in marginal ridings compensate for overall NDP local campaign weakness
A final theory is that the NDP was outspent by its opponents, costing the party votes but not seats. In Canada's single member plurality electoral system, the distribution of a party's votes can matter as much as the number of votes that it wins. For a spending disadvantage to affect the number of seats that a party wins, that disadvantage needs to be in ridings that are marginal enough that local campaigns can affect the winner. A spending disadvantage in ridings where a party is so weak that it has no chance of winning will not affect a party's success.
These dynamics are particularly important for the NDP because it tends to have fewer marginal ridings than the Conservatives or Liberals. Appendix A shows that while at least 10 per cent of ridings in the Rest of Canada (ROC) were marginal for the Conservatives and Liberals in every election between 2006 and 2015, the highest proportion of ridings that were marginal for the NDP was 7.3 per cent (in both 2006 and 2011). In Quebec, the NDP had no marginal ridings in 2004 and only one in 2006. It did, however, have the highest number of marginal ridings in Quebec in 2011 and the second highest in 2015. As a result, overall local spending may do a particularly poor job of capturing the effect of local spending on the NDP's seat share.
H3a: NDP spending disadvantages did not hurt the NDP, because the NDP outspent its opponents in marginal ridings.
Decisions by the national party and the strength of local organizations provide two possible mechanisms that can affect parties’ ability to concentrate spending in marginal ridings. A national party seeking to maximize its seat share should transfer extra funds to local campaigns in marginal ridings to ensure that such campaigns are as strong as possible. Because it has both less money and fewer marginal ridings, the NDP national organization should be particularly careful about which campaigns it transfers money to, as it must use its limited resources as efficiently as possible. That a party would transfer greater resources to marginal ridings fits with work in the UK on local party spending (Johnston et al., Reference Johnston, Pattie, Fisher and Cutts2013; Johnston and Pattie, Reference Johnston and Pattie2014).
H3b: Transfers from the national party prevented the NDP from being at a disadvantage in close ridings.
This theory also has implications for the way that sharp national or provincial changes in support for the NDP affect its ability to concentrate resources in the right ridings. When a party experiences either a sharp rise or fall in support over the course of a campaign, the ridings that end up being marginal in actuality will differ significantly from the ridings that a party expected would be marginal at the beginning of the campaign. Ridings that looked either like lost causes (in the event of a sharp rise in support) or safe (in the event of a drop) could become marginal, confounding national party efforts to identify marginal ridings. Such dynamics should have affected the NDP in 2011 in Quebec when the party experienced sudden growth and in 2015 in Quebec when it experienced a sharp decline.
Local association and campaign strength can also affect a party's ability to have well-funded campaigns in marginal ridings. As Koop (Reference Koop2010) demonstrates with respect to the Liberal party, ridings where a party is stronger tend to have stronger and more professional EDAs. These EDAs should be better able to fundraise and otherwise support local campaigns. The NDP should thus have stronger EDAs in marginal ridings than in lost-cause ridings, and these stronger EDAs should be better able to raise money to support local campaigns. Having incumbents in some marginal ridings should also increase the NDP's fundraising capacity. Donors may also be more likely to give money to campaigns in marginal ridings because they know such donations are more likely to affect an election's outcome.
H3c: The NDP was not at a disadvantage in competitive ridings because of the ability of EDAs in such ridings to raise large amounts of money.
There is a crucial difference between what one would expect to see if hypothesis 3b holds as compared to hypothesis 3c. A national party has little electoral incentive to support campaigns in safe seats. Therefore, national support for campaigns should drop as one moves from marginal to safe ridings. This pattern would be expected by hypothesis 3b. EDAs, however, are likely to get stronger as one moves from marginal to safe ridings. Hypothesis 3c would thus expect EDA support for campaigns to increase as one went from marginal to safe ridings. The proportion of revenue from transfers from the national party, EDAs and local campaign fundraising can also be used to distinguish between hypotheses 3b and 3c.
Data and Methods
Overall approach
The first set of analyses in this article uses descriptive statistics to compare NDP local spending to its competitors’ spending. Regression analysis is then used to estimate the impact of each party's local spending on each party's vote share. These models are used, in turn, to estimate the relationship between local spending and the number of seats the NDP won. Finally, descriptive statistics are used to analyze the revenue sources for NDP local campaigns.
Measurement and variables
Following Carty and Eagles (Reference Eagles2005), I measure local campaign spending using the percentage of the local limit set by Elections Canada that a campaign spends. Elections Canada sets different spending limits for different ridings (based on population and size), and spending limits vary from election to election based on inflation and the length of the election campaign. This makes using spending as a percentage of the Elections Canada limit preferable to raw spending. Spending limits for the 2015 election were unusually high because that campaign lasted 76 days (as compared to 55 for the 2006 election, and 37 for the 2008 and 2011 elections). The particularly high spending limits in that election mean that the average percentage of the limit spent by all major parties is lower in 2015 than in previous elections, even though parties did not spend less money in 2015. The amount a campaign spends and its revenue sources (including donations and transfers from the national party and EDAs) must be filed with Elections Canada, and Elections Canada reviews and makes this data public (found at Elections Canada, 2019).
Figure 1 shows that NDP spending in local campaigns runs from 0 per cent of the limit in ridings where the party had weak campaigns and little chance of winning to near 100 per cent of the limit in more robust campaigns. Because of the significant variation in spending, even a small effect on vote share for a percentage point change in spending can be meaningful. That effect can become quite substantial when one multiplies it by the difference between ridings where the NDP spends next to no money and a riding where the NDP spends half or close to 100 per cent of the limit.
Figure 1. Local NDP Spending by Riding
Note: Each x represents a local campaign.
Spending could also be measured by looking at a party's spending as a percentage of all local spending. Analysis using such measures is conducted in Appendix C and produces similar results to the analysis in the main body of the article. One could also look at donations to a party and donors as measures of local campaign strength. This is done in Appendix B and shows results that are similar to, but less clear, than the results in the main body of the article.
Spending by the NDP's competitors is also likely to affect NDP vote share. To account for this, all analyses in this article include variables for NDP spending as well as Conservative, Liberal and (in Quebec) BQ spending. This is consistent with the approach used by Carty and Eagles (Reference Eagles2005). However, where Carty and Eagles use a single variable for spending by a local campaign's major competitor, this analysis includes a variable for each party's spending. This is done because spending by parties that are ideologically closer to the NDP, such as the BQ and Liberals, are likely to have a larger effect on NDP vote share than ideologically distant parties, such as the Conservatives.
Control variables are included to account for factors that may affect a party's vote share in a riding. These include whether a candidate is an incumbent, ridings’ population density (urban and rural ridings vote differently), ridings’ francophone population, ridings’ visible minority population and ridings’ median income. All demographic variables are taken from the Canadian census closest to the election being analyzed (available through CHASS Reference CHASS2019). Regional differences in party strength are controlled for by breaking the ROC into five regions: British Columbia, the Prairies, Ontario, the Atlantic and the North, each of which is a dummy variable in the analysis, with the North as the base category. The North is used as a base category because it has the fewest number of ridings, though this choice does not affect the estimates for the effect of spending.
The national-level and (in the case of Quebec) province-level conditions vary from election to election. The popularity of Jack Layton in 2011, incentives to strategically vote Liberal in 2015, and numerous other factors shaped the number of seats the NDP won in each election between 2006 and 2015. To account for this, I include dummy variable controls for the election in which each observation takes place. This acts as an election year fixed effect that accounts for each of the national/Quebec level factors that may shape voting behaviour in that election year. It is possible that the effect of local spending on riding vote share varies by election year. To test this, separate models for each election year are run in Appendix D. They find that the effect of NDP spending on NDP vote share is consistent across election years.
Ridings where a party has a history of success are likely to see higher levels of campaign spending and higher levels of party support. To account for this, each model includes a control for parties’ previous vote share in the riding. This has the added advantage of controlling for any effect that spending in a prior election might have had on a party's vote share and is consistent with Carty and Eagles’ (2005) approach. To account for the redistricting that occurred before the 2015 election, transpositions of the 2011 vote share to the 2015 ridings, made publicly available by Elections Canada, were used as the party's previous vote share.
Any analysis of spending's effect on party vote share runs an endogeneity risk. It is possible that, rather than party spending increasing vote share, parties spend more in ridings where they expect to do well. If this is the case, the causal direction would be the opposite of the one expected in the theories outlined in this article, yet the observed relationships would be similar. To the extent that parties’ expectations correlate with past election results, some of this effect is captured by the control for previous vote share. Parties, however, may expect vote share to increase in ridings for other reasons, such as the presence of star candidates. Without getting inside the heads of party strategists, it is impossible to know what their expectations are, and therefore it is impossible to fully account for this endogeneity issue.
The potential for reverse causation should lead to some caution when interpreting the effects of spending on vote share, as it is possible that such findings are a result of reverse causation. A lower estimate of spending's effect on vote share, however, is unlikely to affect the finding that local spending did not cost the NDP seats, though it would suggest that the reason for such a finding has less to do with the NDP's ability to concentrate its spending in marginal ridings. It is further important to note that this analysis can establish correlations but, because it is observational work, cannot prove causation.
When running descriptive analyses comparing levels of spending in marginal, lost-cause and safe ridings, I consider a marginal riding to be any riding that a party is within 5 percentage points of winning or losing. Ridings that a party wins by more than 5 percentage points are considered safe, and ridings that a party loses by more than 5 percentage points are considered lost causes.
The result of the election is used to measure whether a riding is marginal because this article is concerned primarily with whether a riding is marginal in actuality, as opposed to whether parties perceive the riding as marginal. This is because spending in a marginal riding is only likely to affect the number of seats a party wins if the riding turns out to be marginal on election night. This leaves open the possibility that a party may fail to concentrate its spending in marginal ridings because it fails to accurately predict which ridings will be marginal. This is especially likely to be the case for the NDP in Quebec in 2011 and 2015 because of the large vote swings in the province in those elections.
This approach places limits on what can be inferred from the results of the analysis. Because the article does not consider which ridings the NDP thought would be marginal, it cannot draw conclusions about the national party's intent to support campaigns in marginal ridings. It can, however, point toward patterns in spending that have implications for the theories developed earlier.
Methods
Because the vote shares of different parties are related, seemingly unrelated regressions are more appropriate than ordinary least squares models, as they simultaneously model the effect of each variable on each party's vote share. These models are commonly used in analysis of multiparty elections such as British elections (Cutts and Shryane, Reference Cutts and Shryane2006; Russell et al. Reference Russell, Cutts and Fieldhouse2007). For ease of interpretation, linear models are run. Even though election results are not linear (a party cannot get less than 0 per cent of the vote or more than 100 per cent), parties rarely end up at these extremes, making a linear model workable. Only ridings with candidates from all three major parties (four in Quebec) are included in the analysis.
To get at spending's effect on seats won, an estimate for the number of seats the NDP would have won had spending had no impact on the election is compared to the number of seats won in reality. To produce this estimate, the amount of money each party spent in each riding is multiplied by the coefficient for the effect of that party's spending on NDP vote share. These results are summed to produce an estimate of the total effect of all parties’ spending on NDP vote share. This effect is then subtracted from the proportion of the vote the NDP won in reality, producing an estimate for what would have happened had spending not mattered. This process is repeated for the upper bound and lower bound estimates for the effects of spending on NDP vote share to generate a range of NDP vote shares for each riding for a 95 per cent confidence level. The same method is used to generate estimates for the NDP's competitors’ vote shares. Comparing these estimated vote shares allows for an estimation (with error bars for a 95 per cent confidence level) of the number of seats the NDP would have won had spending not mattered.
Analysis and Results
Trends in local campaign spending
Trends in local spending are useful to assessing hypothesis 1a. In the ROC, the average NDP local campaign was outspent by the average Conservative and Liberal campaign in each election between 2006 and 2015. Figure 2 shows that the average Conservative campaign spent the most in each election, the average Liberal campaign spent the second most, and the average NDP campaign spent the least. The gap between local spending by the NDP and its competitors is about the same in 2011 as it was in 2008. This makes it difficult to attribute the NDP's 2011 success to increased overall local spending by the party. In 2015, the percentage of the limit spent by each party declines as a result of the longer campaign and abnormally high spending limit. Nonetheless, NDP spending declined the least. The smaller gap between overall NDP spending and its competitors, however, did not keep the party from falling from official opposition to third place.
Figure 2. Average Percentage of the Limit Spent (ROC)
Figure 3 shows that the NDP was also at a persistent disadvantage with respect to spending in Quebec in every year except for 2015. There is no evidence that the NDP's surge in support in 2011 was accompanied by an increase in local spending. If anything, spending as a percentage of the limit is very slightly lower in 2011 than in the previous election. Even though the NDP caught up to its competitors in 2015, it still lost seats.
Figure 3. Average Percentage of the Limit Spent (Quebec)
There is little evidence from the trends in campaign spending for hypothesis 1a. The party's success in 2011 was not preceded by an increase in local campaign spending. Rather, the NDP was at a persistent disadvantage as compared to each of its competitors. To the extent that the party was able to narrow that disadvantage, it did so in 2015, an election in which it lost seats.
The effect of local spending on NDP vote share
Examining correlations between spending and vote share is useful to assessing hypothesis 2. Figure 4 shows the correlation between each party's spending and each party's vote share in the ROC. A 1 percentage point increase in local NDP spending is correlated with a 0.13 percentage point increase in NDP vote share. By contrast, a percentage point increase in Liberal local spending is correlated with a 0.03 percentage point drop in NDP vote share. A similar increase in Conservative spending is correlated with a smaller drop in NDP vote share that is only statistically significant at the 90 per cent confidence level. As expected, local Liberal spending is positively correlated with Liberal vote share and local Conservative spending is positively correlated Conservative vote share.
Figure 4. Effect of Campaign Spending on NDP Vote Share (ROC)
Note: The margins plot is for the model shown in Table E1 in Appendix E. Bars show the range of effects for a 95% confidence level.
Figure 5 shows that the correlation between local spending and vote share is similar in Quebec. A 1 percentage point increase in local spending in Quebec is correlated with a 0.14 percentage point increase in NDP vote share. The same increase in Liberal spending is correlated with a 0.04 percentage point decline in NDP vote share. A 1 percentage point increase in Conservative spending in Quebec is correlated with a 0.05 percentage point drop in NDP vote share. It is not clear that there is any relationship between BQ spending and NDP vote share.
Figure 5. Effect of Campaign Spending on NDP Vote Share (Quebec)
Note: The margins plot is for the model shown in Table E2 in Appendix E. Bars show the range of effects for a 95% confidence level. Note that there are fewer ridings in Quebec than in the ROC, and therefore one should expect, all else being equal, slightly broader confidence intervals.
These correlations match what is expected by hypothesis 2, insofar as votes are concerned. Not only was the NDP at a persistent spending disadvantage but NDP local spending is positively correlated with NDP vote, and most of the NDP's opponents’ spending is negatively correlated with the NDP vote share. This is consistent with the theory that the NDP's local spending disadvantage was a constant drag on its vote share. One must be cautious about drawing too strong a conclusion about a causal relationship between spending and vote share, however, because of the endogeneity issues noted earlier in the article. Whether this disadvantage could have affected NDP seat share depends on whether the NDP was at a spending disadvantage in marginal ridings.
Spending and seat shares
To get at hypothesis 3a, one needs to look at how the correlation between spending and vote share fits into a model that estimates the relationship between spending and number of seats won by the NDP. There is little evidence that the NDP would have won more seats had spending not affected election results. Figure 6 compares the number of seats that the NDP won in the ROC with estimates for the number of seats they would have won had there been no correlation between spending and vote share. In every election the NDP is estimated to have won fewer seats than they won in reality, and in every election but 2006 that difference is statistically significant at the 95 per cent confidence level. I estimate that in 2015, the NDP would have only won 17 seats in the ROC had spending not mattered, as compared to the 28 seats they won in reality; 34 compared with 44 in 2011; 22 compared with 36 in 2008; and 18 compared with 29 in 2006.
Figure 6. Seats Won without Spending Effect Compared to Seats Won in Actuality (ROC)
Note: Xs show the predicted number of seats based on the effects detailed in the previous section. Bars show the ranges for a 95% confidence level.
Figure 7 shows that the story is more complicated in Quebec. I estimate that in 2011, the NDP would have won more seats had spending not been correlated with vote share, while in 2015 they would have likely won fewer (though the difference in 2015 is not quite statistically significant). I estimate that in 2011, the NDP would have won 68 seats had spending not mattered, as compared to 59; and in 2015, I estimate they would have won 9 seats as compared to 16. Because the NDP was very weak in Quebec prior to 2011, it is unlikely that spending could have affected the NDP's success in 2006 or 2008.
Figure 7. Seats Won without Spending Effect Compared to Seats Won in Actuality (Quebec)
Note: Xs show the predicted number of seats based on the effects detailed in the previous section. Bars show the ranges for a 95% confidence level.
It is particularly noteworthy that 2011 in Quebec is the only instance in which there is evidence that the NDP was hurt by its local spending disadvantage. As will be discussed in detail below, the NDP's overall disadvantage is not correlated with a loss in seats because the party was not at a disadvantage in marginal ridings. The exception to this is 2011 in Quebec, where the NDP's surge would have made it difficult to predict which ridings would end up being marginal. Nonetheless, the model's estimates of NDP losses in Quebec in 2011 are cancelled out by estimates of NDP gains in the ROC. Taking Quebec and the ROC together, the model estimates that the NDP would have won 102 seats in 2011 had spending not been correlated with vote share, as compared to the 103 seats the NDP won in reality.
Analysis of marginal ridings shows that the NDP was not at a spending disadvantage in such ridings, at least in the ROC. Figure 8 shows that the party consistently outspent the Liberals in these ridings, even in the 2015 election when it was facing revitalized Liberal associations. The NDP was outspent in marginal ridings by the Conservatives in 2008 and 2011 in the ROC.Footnote 1 Conservative spending, however, has a weaker correlation with NDP vote share and thus has less influence over NDP success.
Figure 8. Average Spending in Marginal NDP Ridings (ROC)
Note: The number and, in brackets, proportion of marginal NDP ridings in the ROC in each election is as follows: in 2006, there were 17 (7%) marginal ridings; in 2008, there were 13 (6%); in 2011, there were 17 (7%); and in 2015, there were 18 (7%).
The NDP's spending in marginal ridings in Quebec is more complicated. Figure 9 shows that the NDP was outspent in marginal ridings in 2011. This is to be expected given the NDP's sudden surge in support in the province in 2011. A sudden surge in support over the course of a campaign should lead to lower spending in marginal ridings regardless of the mechanism that affects spending in marginal ridings. If the national party is influencing the concentration of spending in marginal ridings, a sudden surge in support should make it difficult for the national party to identify the marginal ridings it ought to transfer extra funds to. If EDAs are responsible for high spending in marginal ridings, a sudden surge in support should make ridings with weak EDAs marginal, decreasing the amount of support that campaigns in marginal ridings get from EDAs. Regardless of the mechanism, the case of the NDP in Quebec in 2011 shows how a sudden surge in support can reduce the degree to which a party's local spending is concentrated in marginal ridings.
Figure 9. Spending in Marginal NDP Ridings in Quebec
Note: In 2011, there were 11 marginal ridings for the NDP in Quebec (15% of ridings in Quebec), and in 2015, there were 18 (23%).
Unlike in 2011, the NDP was able to outspend its opponents in marginal ridings in Quebec in 2015. In such ridings the NDP's average spending was more than double that of the Liberals and substantially larger than that of the BQ or Conservatives. This fits with the estimate that the NDP would have won fewer seats in Quebec had spending not mattered. The sudden drop in support for the NDP in Quebec in 2015 may have played a role in this. The drop would have made what initially looked like safe ridings with stronger EDAs marginal, and such EDAs would have likely been able to provide particularly strong support to their local campaigns.
These results are consistent with what is expected by hypothesis 3a, at least in the ROC. The NDP was at a disadvantage in overall local spending but was not at a disadvantage in marginal ridings. This makes it possible for the NDP's spending disadvantage to cost the party votes but not seats. With the exception of Quebec in 2011, in marginal ridings the correlation between NDP spending and NDP vote share did not hurt the NDP because the NDP outspent its main opponent (the Liberal party). In ridings that were not marginal, the correlation between NDP spending and NDP vote share was not large enough to make a difference to the NDP's ability to win ridings. These correlations and patterns suggest hypothesis 2 applies to the relationship between NDP spending and vote share but not seat share. There is little evidence that the NDP's local spending disadvantage was a persistent drag on its ability to win seats.
Party transfers and donations
Understanding where campaigns got their revenue is necessary to assessing the plausibility of hypothesis 1b and hypotheses 3b and 3c. If strategic decisions by the national party drove the NDP's advantage in marginal ridings, transfers to local campaigns should go up as one goes from lost-cause to marginal ridings and then drop as one goes from marginal to safe ridings. Figure 10 shows that national party transfers in the ROC did not fit this pattern. Instead, transfers from the national party to local campaigns consistently increase as the party's vote share increases, even when one goes from marginal to safe ridings. The closest national transfers come to the predicted pattern is in 2015 when transfers level off (but do not drop) in safe ridings.
Figure 10. Party Transfers by NDP Margin of Victory or Defeat (ROC)
Explaining the lack of a decline in transfers from the national party to local campaigns as one goes from marginal to safe ridings is beyond the scope of this article and would require qualitative research into the motivations behind party transfers. It could be that incumbents in safe seats can use their influence over the party to secure extra funding for their campaigns. It may also be the case that the NDP feels it does not have safe ridings and therefore transfers extra funds to local campaigns in safe ridings out of an abundance of caution. That transfers to safe ridings level off in 2015—the election in which the party started from its strongest position—fits with this latter explanation.
The dynamics of party transfers in Quebec are more complicated. Figure 11 shows that in every election except for 2015, transfers to local campaigns in Quebec increased as the NDP got stronger. From 2006 to 2011 the NDP had no reason to believe that it had safe seats in Quebec and thus it could not afford to reduce its transfers to campaigns in its strongest ridings in the province. Notably, in 2011 NDP national party transfers to local campaigns in Quebec were greatest in the party's four strongest ridings. This may have been a result of the party's low expectations about how many seats that the party could win. Only when the party's support in Quebec surged in the middle of the campaign did those ridings become safe. In 2015 the NDP's national transfers fit the expected pattern, dropping off as one gets to the strongest ridings for the party. However, the NDP's decline in support over the course of the 2015 campaign meant that ridings that looked safe became marginal. As a result, the national party underinvested in marginal ridings.
Figure 11. Party Transfers by NDP Margin of Victory or Defeat (Quebec)
It is does not appear that local campaigns in marginal ridings were particularly reliant on national party support for their revenue. Figure 12 shows the average percentage of ROC local campaign revenue that came from national party transfers, EDAs, donations during the campaign and loans.Footnote 2 The vast majority of local campaign revenue comes from EDAs and donations during the campaign. In no election did transfers from the national party make up more than 20 per cent of a campaign's revenue. Marginal campaigns are not an exception to this. Depending on the election, between 83 per cent and 90 per cent of revenue for the average marginal campaign came from either the EDA or donations to the campaign.
Figure 12. Proportions of Transfers and Donations to Local Campaigns (ROC)
Figure 13 shows that marginal campaigns in Quebec were more reliant on transfers from the national party than those in the ROC. Marginal campaigns in Quebec received 35 per cent of their revenue from the national party in 2011 and 23 per cent in 2015, with the rest of campaign revenue coming from EDAs and donations. The national party was thus more important to local NDP campaigns’ ability to spend in Quebec than in the ROC. Nonetheless, EDAs and donations still accounted for over three-quarters of local campaign revenue in Quebec in 2015, the election in which the NDP was able to outspend its competitors in marginal ridings in Quebec. Even in Quebec, EDAs and local campaign fundraising played a much larger role in the NDP's ability to outspend its opponents in marginal ridings than transfers from the national party did.
Figure 13. Proportions of Transfers and Donations to Local Campaigns (Quebec)
While transfers from the national party to marginal campaigns were non-trivial, especially in Quebec, marginal campaigns got the bulk of their revenue from their EDAs and donations that they were able to raise. This provides little evidence for hypothesis 1b (that transfers from the national party to the local campaigns became more effectively targeted over time) or for hypothesis 3b (that the NDP's spending advantage in marginal ridings was a result of carefully targeted transfers from the national party). Indeed, the national party transfers to campaigns in Quebec in 2011 and 2015 suggest significant limitations to the ability of the national party to identify marginal ridings when there is a dramatic swing in party support. The findings are consistent with hypothesis 3c: that the NDP's ability to outspend its opponents in marginal ridings is largely determined by the strength of the EDA and local campaign fundraising. Table 1 presents a summary of the hypotheses and findings.
Table 1. Summary of Hypotheses and Findings
Conclusion
Over the 2006–2015 period, NDP local campaigns were consistently outspent by their opponents. While lower local campaign spending is correlated with lower vote shares, there is little evidence that this lower spending cost the NDP seats. Because the NDP outspent its opponents in marginal ridings (with the exception of 2011 in Quebec), a local spending disadvantage and a correlation between spending and vote share does not necessarily mean that spending cost the NDP seats. Quite the opposite: the NDP's spending advantage in marginal ridings suggests that the correlation between spending and vote share contributed to their ability to win marginal seats. The main source of the NDP's advantage in marginal ridings was revenue from EDA and local fundraising, suggesting that the national party had a limited ability to influence spending in marginal ridings.
One must be careful about drawing strong conclusions about the presence of a causal relationship between spending and vote share from this work, or about the motivations behind national party transfers to local campaigns. The endogeneity issues that exist as a result of the possibility that parties spend more in ridings in which they expect to win more votes complicate the conclusions one can draw from the correlations between spending and vote share. The difficulty in understanding the motivations behind the decisions made by party strategists makes it difficult to know with certainty the extent to which the NDP's national organization was trying to identify marginal local campaigns and provide them with additional support.
Nonetheless, this article's findings have important implications for understandings of local spending in Canadian elections. They demonstrate that in single member plurality electoral systems, where a party spends money can matter as much as the amount a party spends. A party can be at a disadvantage in local spending and not lose seats even if local spending is correlated with vote share. This is particularly important for small parties. One cannot understand the impact that local spending has on such a party's success without understanding whether its spending is concentrated in marginal ridings. This is likely true not just for the NDP but also for other smaller parties, such as the federal Green party, third parties in provincial elections, and small parties in other countries with single member plurality systems (for example, the British Liberal Democrats). Notably, concentrated spending in marginal ridings can only compensate for smaller parties’ spending disadvantages in single member district electoral systems (single member plurality systems and possibly alternative vote systems). In a proportional system, where the geographic distribution of votes matters less to a parties’ seat shares, concentrating spending matters less.
This article further suggests that national party organizations, or at the very least the NDP, are limited in their ability to concentrate spending in marginal ridings for two reasons. First, the bulk of local revenue for local campaigns comes from EDAs and local campaign fundraising. National parties certainly can try to build up local EDAs between elections and encourage them to fundraise, but the extent to which an EDA fundraises and supports a local campaign ultimately depends on the EDA itself. Second, as is demonstrated by the NDP's experiences in Quebec in 2011 and 2015, it can be difficult to predict which ridings will be marginal if there is a significant change in national or province-wide vote intentions in the middle of a campaign. A national party that targets its transfers to local campaigns too narrowly may end up investing in the wrong local campaigns if a mid-campaign swing in voter intentions changes which ridings are marginal. A small party's concentrated spending in marginal ridings may thus be more a result of having strong EDAs in the right places than of careful strategic planning by the national party.
Without a fuller analysis of the extent to which the Conservatives and Liberals are able to concentrate their spending on marginal ridings, it is hard to say whether the NDP is the only party that benefits from having its spending concentrated in marginal ridings. That the NDP was able to outspend its opponents in marginal ridings suggests that its mass-party roots do not prevent it from spending more in marginal ridings than lost-cause ridings. Though I do not find evidence that NDP local spending became more concentrated over the 2006–2015 period, my findings broadly fit with McGrane's (Reference McGrane2019) argument that the party adopted a more professionalized structure for the entirety of the 2006–2015 period. That EDAs were more responsible for the NDP's spending in marginal ridings than was the national party suggests that building strong grassroots organizations was an important part of that professionalization. Importantly, my findings fit with a broader argument that the NDP has moved away from a mass-party structure to a more professionalized catch-all structure.
Finally, this article's findings provide some nuanced evidence addressing the question of whether local campaigns matter. The correlations between spending and vote share and spending patterns that see the NDP outspend its opponents in marginal ridings fit with the theory that NDP spending in marginal ridings helped it win seats in 2006, 2008 and 2015. One has to be careful, however, about drawing conclusions about a causal relationship from this evidence because of the endogeneity issues inherent in tests of spending's effect on vote share. The work here, nonetheless, fits with a broader literature suggesting local campaigns matter (Blais et al., Reference Blais, Gidengil, Dobrzynska, Nevitte and Nadeau2003; Blais and Daoust, Reference Blais and Daoust2017; Carty and Eagles, Reference Carty and Eagles1999; Carty and Eagles, Reference Eagles2005; Stevens et al., Reference Stevens, Islam, de Geus, Goldberg, McAndrews, Mierke-Zatwarnicki, Loewen and Rubenson2019). Local campaigns do not, however, matter in all cases. The correlations between spending and vote share found in this article are strong enough only to affect who wins seats in marginal ridings. This suggests that, at best, local campaign spending only matters in ridings that are already close races. When there is a large national or province-wide shift in vote share, as was the case for the NDP in Quebec 2011, local campaigns and spending play less of a role in the success of political parties.
Supplementary Material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0008423922000014.
Acknowledgments
I am grateful for comments provided by Elizabeth Baisley, William Cross, Kyle Hanniman, J. Scott Matthews and Elizabeth Schwartz on earlier versions of the paper. I am also grateful for comments provided by many members of the Queen's Political Studies department during presentations of this work in the department and for those by the anonymous reviewers.
Open access
