So long as I've been involved in politics, we've always measured performance by "percent of the vote", aka "vote share", which is the candidate's or party's number of votes expressed as a percentage of the valid ballots cast.

We say that a party's vote share went up or down, or that party X lost a certain percent of the vote to party Y; and define the gap between first and second place (the "percent margin") as a percentage of the valid ballots cast as well.

The unspoken assumptions behind this analytic shorthand were two-fold: that turnout was reasonably constant, and that in any event the party leanings of those who didn't vote for whatever reason in a certain election were insignificantly different from those who did.

I'm no expert in turnout, although André Blais and Peter Loewen who are have just published a working paper estimating it, available through the Elections Canada website if you missed the citation at Aaron Wherry's blog the other day; they are also the same two who've written about youth engagement for Elections Canada here.

But I think anyone who's active in politics these days can agree that turnout levels are dropping, and there's good reason to suspect that voters and non-voters are now diverging in a number of ways. We would probably also draw a distinction between long-term non-voters and temporarily discouraged voters, in order to round out the picture.

Karl Rove originated the strategy in the United States of finding wedge issues to motivate targeted groups of hitherto unlikely voters to support his candidate, while employing other (primarily "air-war") tactics to discourage the supporters of his candidate's opponents in the hopes that they'd "stay home" (i.e., not go out to vote). As I've written before, there are two paths to winning: one by increasing your own share of the vote, whether from non-voters or former supporters of your opponent; and one by causing your opponents to stay home. In response to the Rove-ian approach, David Plouffe and David Axelrod advocated a different strategy for their candidate's campaign, where they sought to increase vote-share from long-term non-voters by promoting voter registration and running fully-funded ground campaigns in all 50 states, rather than just targetted efforts in swing states as had been the usual practice in the past.

As you've no doubt figured out by now, the usual metric of "vote share" is just not up to the task of measuring or properly describing these movements. In a situation where no other votes change, but party A's votes stay home, using percentage of the vote as the indicator can make it look as though party A's supporters switched to party B and C, when they did no such thing.

This is why I've added "percent of the electorate" (aka "% Elec") to the riding profile pages here, and will be slowly adding it elsewhere in the database as well. For greater clarity: a candidate's or party's "percent of the electorate" is their number of votes expressed as a percentage of the eligible voters, rather than the valid ballots cast. Then the number of non-voters (NV) can also expressed as a percentage (plus it also works out to 100% - the turnout rate), and added to the bottom of the list. In a typical riding profile these days, the NVs are winning, but their numbers do vary from election to election. I show all the % Elec calculations in an italic font, so it's easier on the eyes to distinguish them from the "percent of the vote" column.

So, with this new metric available, what new insights can we get into some of the pivotal strategic questions of the next campaign? Let's take a second look at some of the seats that changed hands in earlier campaigns, and see if we can find out why:

• Welland, ON and Sudbury, ON - These two seats switched from the Liberals to the NDP last time. If you take a look at the results expressed as a percent of the electorate: the Conservative vote stayed constant, the NDP vote (surprisingly) declined very slightly by a single point in both case, but the Liberal vote dropped substantially (by 7 and 10 points respectively), ALL of it switching to the non-voting (NV) camp. Suddenly it becomes very clear that the NDP strategy to keep the seats must involve increasing their vote from longer-term non-voters, while the Liberal strategy will be focused on re-motivating their earlier supporters to return to the polls in the hopes of regaining their former seats. Going back one more election in both ridings, the NDP vote did increase in 2006 from the ranks of previous non-voters, but it did not increase further between 2006 and 2008.
• Kitchener Centre, ON - In this case, the NDP and Green vote stayed constant from 2004 to 2008, while the Conservative vote increased between 2004 and 2006, but stayed put in 2008. The difference in 2008 is that the Liberal vote dropped by 7.5 points, all of it switching to the non-voting camp. This was enough to cause the seat to change hands by 339 votes, just 1.6 per poll. That's 0.8 percent of the vote, and 0.4% of the electorate. So, turnout dropped from 64.7% to 57.0%, all at the expense of the Liberal incumbent.
• Toronto – Danforth, ON in 2004, and Trinity – Spadina, ON in 2006 - In both these races, the NDP won the seats away from the Liberals in a different way than the first case: Jack Layton won in 2004 by picking up votes from previous non-voters, a feat not achieved by his wife Olivia Chow until 2006 in her riding. A similar pattern can be seen in Timmins – James Bay, ON in 2004. Here the other parties' votes held constant, or else the Liberals declined slightly, while the NDP posted big gains as turnout fell. In subsequent elections, with the incumbency tables turned, the Liberal vote continued to fall, with some of it switching to the Conservatives and Greens as well as staying home, leaving Layton and Chow with increased margins, and a more divided opposition, even as their own vote abated somewhat.
• Saint Boniface, MB - Of course, not every riding that changed hands fit one of the above patterns, as in this case where turnout did not change too much and it appears that votes actually moved from the Liberals to the Conservatives.

I think it's important to examine turnout and the parties' shares of the electorate much more so than vote-share when trying to explain how votes shift between elections.

It's *particularly* important when various schemes are multiplying in the commentariat and blogosphere about how parties could or should combine their efforts or step aside in favour of others, whether from my old Carleton poli sci prof Reg Whitaker and his colleague Philip Resnick, writing in TheTyee.ca about how to defeat the government, or L. Ian MacDonald writing in the weekend's Montreal Gazette about how to block the Bloc in Quebec.

What I hope I've demonstrated is that just looking at changes in a party's vote-share from one election to the next, and comparing it to changes in another party's share, can lead to some erroneous conclusions, and cause you to miss other movements which actually occurred ... with potentially serious strategic consequences.

Most of the examples above come from Ontario, because I've been working on Ontario ridings for another reason the past few days, but the same reasoning when applied to seats in British Columbia will allow you to see more clearly how votes shifted over the past few elections.

Next time: a new tool to examine regional party vote swings in comparison with one another.

Labels: Percent of the Electorate, Turnout