Outcomes of the September 6 byelections in Ontario focus a great deal of attention on the "swing voter," that is, the voter with loose partisan ties who can potentially be swayed from one party to another. This is the prize all three parties pursue, especially in Kitchener-Waterloo where there is nothing to suggest one candidate has a comfortable and wide lead. It's anybody's game, hence the race to sway the swing voter.
On this theme, I raise four questions, and locate answers based on a preliminary analysis of some data in our collection.
First, how do we define a “swing” voter, methodically? This is certainly not a new topic. For instance, Mayer used the American National Election Study’s 100-point thermometer scales to identify the American swing voter, defined as a voter with equally favourable views of the Democratic and Republican parties. Read the article for details on the actual methodology. However, at present, it cannot be easily applied in the Ontario case, at least not with the data I have on hand.
I’m using the 2011 Ipsos Reid online “exit poll” conducted on the day of the Ontario general election (available at the LISPOP data portal), and it does not contain thermometer rating scales. So I had to improvise using other items available in this survey.
Let’s start with some assumptions about a swing voter. They probably hold the following attributes:
- They are less likely to remain loyal to one party. Such partisans are, by definition, the opposite of a swing voter.
- They likely make their final vote decision later in the campaign, given their propensity to “swing” back and forth along with the oscillations of campaign dynamics.
Either one, on its own, may not be sufficient grounds to identify the swing voter. One can vote for one party in one election and another party in a later election because of a change in loyalties. Or, someone could wait until near the end of a campaign to make a "final" decision, but it may be little more than an affirmation of the party which they tend to support quite regularly. They are not moved by any partisan loyalties, per se, but by habit (which, I guess, is a form of partisan attachment). In any case, it seems that from this improvised two-dimensional typology, the "swing" voter would be one who decided late and also decided to support a party that is different than one chosen in a previous election.
My second question is: How large is the "swing" vote? According to some American observations, it's marginal. A New York Times article (brought to my attention by a LISPOP colleague Geoffrey Stevens) reports on research that finds a very small percentage of the American electorate considered true swing voters, identified as those who are "independent" or not partisan to any one party. Furthermore, most swing voters appear to reside in non-battleground states, such as California and Texas.
But when my two dimensions are applied to the Ipsos survey, we see different picture.
As the table shows, 30% of respondents voted differently between the two elections. The same applies for the Kitchener-Waterloo riding, although this is based on a much smaller subsample (n ˜ 80). The numbers from Vaughn, while available, are too tiny to render any meaningful insight. But regional data shows the "416" and "905" as having the largest percentage of consistent voters (more than 70%).
Furthermore, 52%, appear in the top-right section: voted the same in 2007 and in 2011, and made this decision early in the campaign, defined here as before the televised leaders' debate. These voters may have been unaffected by campaign activities as their decision to vote for the same party they had supported in 2007 was reached before the campaign actually began. We can call this cell the "partisans."
The least populated section is in the top-left, switchers who made their final decision before the campaign. This is not surprising at it takes some effort to move someone from supporting one party to another. What’s a bit surprising is this cell still accounts for a substantial proportion of the sample, 14%. Perhaps these are voters whose partisan ties have realigned.
The bottom right-hand comprises 18% of respondents. Their vote from 2007 to 2011 remained the same, but for some reason the final vote choice arrived late. Perhaps these are the more "habitual" partisans who were not entirely engaged with the campaign, so their final decision was to remain loyal to one party. It could very well be that these are swing voters who may have at one time during the campaign moved to another party, but were, for whatever reason, swayed back to their original leanings.
Let’s now turn our attention to the bottom left group of cells. These respondents, which account for 16% of the total sample, voted in 2011 for a party different than the one they had supported in 2007, and also made their final vote decision sometime after the leaders’ debate, i.e., the “swing” voter according to the formulation outlined here. It is not a humungous proportion of the electorate, but not insubstantial.
Provided we accept this formulation, then just how practically significant is this group of swing voters? In other words, among this group of 1300 respondents, how did their vote change? Which party benefited? That's my fourth and final question.
From the above table, two things are clear. First, the biggest group of swing voters are those who voted Liberal in 2007. Second, the biggest beneficiary was the NDP, which gained most from swing voters across all parties. Even among those who voted PC in 2007, 46% moved to the NDP, which is higher than the 43% who moved to the Liberals.