The effects of election timing on national and subnational turnout in Canada

ABSTRACT

Where a considerable volume of work has been spent looking at the effects of voter fatigue in multi-level democracies, few work has looked at Canada as a unique case. Leveraging the specificity of the Canadian federation – level-specific party systems and electoral cycles – we show that there exists a strong and significant voter fatigue effect such that electoral turnout is lower in early, ‘snap’ elections and increases as does the time in weeks between elections, wherever the level at play. This is particularly important given that the stochastic nature of provincial and federal elections in Canada induces Canadians to go to the polls frequently – perhaps in part explaining the country’s turnout decline since the mid-1940s.

KEYWORDS:

• Elections; election timing; turnout; federalism

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Scarrow, himself, relies on the official elections’ reports for these years. A caveat is required for the data from certain provinces. First, we report turnout from British Columbian elections with MMD systems by dividing the number of voters who casted a ballot over the number of registered voters. Second, elections conducted from 1967–1978 in Saskatchewan only report turnout as a percentage of eligible voters. Third, Prince Edward Island and New Brunswick only began reporting turnout in provincial elections in 1966 and 1967 respectively. We exclude municipal elections for two reasons. First, municipal electoral data going back to the mid-twentieth century is difficult to aggregate given the volume of elections held (there were, for example, roughly 3,000 elected positions in the 2021 municipal elections in Quebec (Government of Quebec, 2021)). Second, municipal elections vary in terms of electoral and party system. And where political parties do exist (e.g., Vancouver), they are independent of the provincial and federal parties.

2 Our two covariates of interest are correlated. That might explain why the addition of the number of weeks passed since the last election (Model 3) reduces the magnitude of the coefficient associated with snap elections.