Measuring Campaign Mobilization Effects Canvassing and the Vote in a Canadian Electoral District* - A Preliminary Assessment June 2005

Introduction:

This report looks at the election campaign undertaken by a federal party campaign team in an electoral district (also known as “ridings” in Canada) in the province of Ontario during the context of the June 2004 federal election. The seat had been held by the party during the 1980s, but it changed party hands in the 1993 campaign. The 2000 campaign for the party whose effort we assess in this report was a bare-bones effort and attracted less than 10% of the vote. However, the  incumbent Member of Parliament retired just prior to the 2004 election, thereby opening the seat up for a tight three-way race. The party mounted a much more aggressive campaign in 2004 than in the previous election. The present objective is to assess in very general terms the effectiveness of the several grassroots campaign activities that it undertook.

For the analysis, we use information supplied by the party’s campaign manager, for the 213 polling districts (6 polling districts had incomplete data and were dropped from the analysis) that comprise the federal riding. For each polling district we have the party’s 2000 vote share and four measures of the 2004 campaign effort. For each of the riding’s polling districts, we have coded dummy (or indicator) variables scored “1” when the candidate made door-to-door canvassing visits to all homes, when advertisements were included as an insert in the Pennysaver that was delivered to each home, when leaflets were dropped off at each home by party canvassers. In addition, we have a count of the number of lawn signs for the party in each of the riding’s polls.

This is a simple test but it is one that has some utility in guiding future campaigns. No attempt was made to incorporate socio-demographic controls from census data. Since we have included a measure of the party's prior support, however, some of these socio-demographic effects would be subsumed by this variable. No evidence on the campaign effort of rival candidates was available. Barometrics Research is in the process of preparing data for a more extensive evaluation. Notwithstanding the simplicity of the test, however, we would expect that the party’s vote share in 2004 should be enhanced over its 2000 result in those polls where it concentrated its campaign effort. Was this the case?

Analysis:

The 2004 campaign successfully advanced the party’s vote share, from the 7% won in 2000 to approximately 24% of the vote share. Table One presents a statistical analysis of the impact of these various campaigning efforts across the district, controlling for the party’s 2000 performance. As is usually the case, since past performance is the best predictor of future performance, the foundation for the party’s growth was substantially related to its prior base – i.e., the party did best in 2004 where it did best in 2000. Our analysis suggests that once the campaign efforts are statistically controlled, for every 1 percent of the vote the party won in 2000, in 2004 the party attracted 1.597 percent. The improvement in the party’s performance over these two elections can be seen as the joint product of a more competitive local race (as a result of the retirement of the incumbent), a stronger and more active candidate for the party, and the party’s more aggressive mobilization effort during the 2004 campaign. How much of the party’s growth in support can we attribute to the efforts of the campaign team and candidate?

The results strongly suggest that the campaign was quite effective in attracting additional support to the party. For example, the analysis suggests that each lawn sign in a poll added an additional 0.68% to the party’s vote in the poll. The Pennysaver insert added a further 2.176%, and the door drop of leaflets added an impressive 3.247% to the party’s performance. Cumulatively, this suggests that upwards of 5 to 6 % in some polls was added directly as a result of the campaign efforts. This is impressive evidence of the difference that aggressive local campaigning can make.

Surprisingly, the only activity not contributing to the party’s improved performance was the candidate’s doorstep visits. This is surprising because it is in some respects the most intensive form of grassroots campaigning, and obviously the most demanding for the candidate. Interpreting this result is difficult with the data at hand. It is possible that the effects of the personal door-to-door campaign by the candidate were overwhelmed by her participation in a variety of constituency-wide campaign events – i.e., participation in all-candidates debates, media interviews, etc. When given an active and credible candidate, the electorate appears to have responded positively, but the geography of this response was relatively uniform across the riding and was not directly associated with the candidate's specifically personal activities in canvassing.

In addition, this unexpected finding about the absence of evidence for an effect associated with candidate visits also reflects in part the strategic decision made about where in the district to send the candidate. In this respect, the campaign manager reported that the candidate’s visits were targeted at polls in which a rival party had traditionally dominated, but where the campaign expected that this support was "soft" and possibly winnable. In other words, the candidate was sent into hostile territory to attempt to win new recruits.  While we cannot say with certainty, it is entirely possible that these areas responded favorably to the visits but the impact was not strong enough to overcome their traditionally negative disposition toward the party. Had the strategy been to deploy the candidate to mobilize the party's base rather than to win new voters to the party, the coefficient would almost certainly have been in the expected (significant positive) direction.

In Table Two we identify the handful of polling districts in which the party does much better than the relationships captured by the model (prior vote and all measures of campaign effort) would lead us to expect (high positive residuals). Here whatever you are doing – or whatever exists in these environments – is very conducive to support for this party. By the same token, we include a handful of polling districts in which the level of party support observed in 2004 was BELOW that which the model would have led us to expect. These are polling districts in which an aggressive campaign next time around might pay especially high dividends. Bringing these “up” to the general level of support your party drew across the riding as a whole would add considerably to the party’s chances at winning.

Conclusion:

Campaign effort clearly mattered for the party in 2004. In polls where the party made a concentrated effort in distributing its literature, the electorate responded positively. While a 6% increase in vote share may not sound large, in the context of a tight three-way race vote shifts of this magnitude could be crucial. This finding should energize the party’s campaign in the next election, and suggest that there is a return to be expected in vote share for the effort expended.