North America: Polling and Politics

First Past the Post is the dominant electoral system across North America. The country is divided into single-member constituencies, each electing a single representative by simple plurality. The party that wins the most constituencies forms the government — even if it is short of a majority of votes nationally. This has two recurring consequences. First, parties whose support is geographically concentrated tend to win seats far in excess of their national vote share, while parties whose support is geographically spread can be heavily penalised. Second, swings in vote share do not translate cleanly into swings in seats: a two-point national lead can produce a comfortable majority in one election and a hung parliament in another, depending entirely on where the votes pile up.

For polling, the practical implication is that national vote intention is only weakly informative about seat counts. Sophisticated analysts in FPTP systems work with constituency-level swing models, regional polling, and seat-by-seat probabilities. A party leading by three nationally but trailing in twenty marginal seats can lose the election outright. Readers who focus only on the headline number can miss the real story.

Two-bloc dynamics dominate the surface of North American politics, but the picture beneath is more complex. Both Canada and the United States have seen significant shifts in regional voting patterns over the past two decades, with previously safe seats becoming competitive and previously marginal seats becoming safe. Polling that ignores this geographic realignment systematically misreads the state of play.

Polling in North America has grown more difficult, not easier, over the past decade. Phone response rates have collapsed; online panels are now dominant; and at least one or two of the most prominent firms in each country have had embarrassing misses in recent national elections. The industry has worked hard to adapt, but the careful reader should treat individual polls with humility and pay more attention to averages and trends than to the latest survey.