Election analysis: Post election blues: Splitting hairs on vote splits

One of the outcomes of the 2011 Federal election has been interest in how Liberal and NDP voters split the progressive vote, and thus paved the way for a conservative majority.   Shouldn't it be easy to understand what role vote splitting played in the 2011 election? Well, yes. And no.

As with most things, it all depends on how you define vote splitting. Traditionally, vote splitting refers to a situation whereby voters torn between two or more similar candidates effectively split their vote, reducing the chance that either candidate will win while increasing the chances for a dissimilar candidate. Brian Topp and Ken Boessenkool offer another useful set of assumptions. They suggest one look in more detail at victorious candidates winning with less than 50 per cent of the vote, who won seats with small margins over the second place finisher. In the 1990s it was the conservative vote that was split. More recently, it is the progressive vote that appears split. The answer, according to some, is a merger between the New Democratic Party and Liberal Party of Canada.

Those who spend time gathering and analyzing data reject this simplistic definition. They argue you cannot simply add LPC and NDP vote totals together and then assumethese voters would work together to stop a Conservative MP. Both parties are made up of a variety of voters and more rigorous approaches are needed to understand this issue.

One is by Éric Grenier who tried to show how New Democratic gains contributed to Liberal defeats. He considers vote splitting by looking at 2008 data and then comparing how candidates did in 2011. His analysis suggests, especially in Ontario, that NDP gains were greater than the gaps between Liberal incumbents and Conservative challengers. While the Conservatives earned their majority, Grenier concludes the vote shift played an important role in electing a Conservative majority government.

Another is by Alice Funke who has challenged this conclusion. She considers that vote splitting occurs when the party who wins the seat gains fewer votes than the party losing the seat, or when a third party gains more votes than the losing party loses. On this view, she argues that in only 6 cases did the so-called NDP surge cost the Liberals their seats. Furthermore, even if the LPC or NDP simply disappeared, Funke contends the Conservatives would likely still have won unless either of the parties could hold on to 100% of the other party’s votes.

While useful, each of these approaches must contend with some methodological challenges associated with some of the assumptions they rely upon. Both use the results of 2008 as the starting point. While this is logical, it also begs a few questions: Can we really assume past elections are a good prediction for the future? How do we account for broader trends that might change a voter’s choice between elections? Are provinces, regions, or other distinctions relevant? What about new voters?

At least some of these questions will be answered once the Official Voting Results, including the poll-by-poll results, are released over the summer. In the meantime, the pundit class ought to hold their powder and wait to theorize until we have more data. At the very least they should, as Grenier and Funke have done, clearly spell out their assumptions, and present their methodology. It does appear though that in Ontario, there were 16 seats won by the CPC in which the difference was 7% or less. It is these sorts of outcomes that have driven some to develop new approaches to strategic voting. TheCatch 22 campaign is one example and their experience offers some important lessons about what may work and what may not going forward. Clearly, strategic voting cannot occur without reliable data and good local polling at the riding level. Informed decision-making requires information – one lesson from 2011 is that it is local riding information, not national trends that will allow progressives to make their voice heard in four long years. The planning starts now.