Byelection performance can be predictive of future results

After the votes are counted in Monday's five federal byelections, only the parties that make gains will want to talk about the results.

The others will point out that byelections are local affairs that tell us nothing about broader national trends; that they don't really matter. 

But that's a myth.

In fact, there is a relationship between how a party does in a byelection and how that party performs nationwide in the next general vote.

• Trudeau hits campaign trail as opponents cry foul
• 5 byelections mark 1st electoral test for Trudeau government

Byelections will be held in the ridings of Calgary Heritage, Calgary Midnapore, Markham–Thornhill, Ottawa–Vanier and Saint-Laurent. The Conservatives are the incumbents in the two Calgary ridings while the Liberals are in the other three. All are considered safe seats.

An upset in any of them is unlikely. But the results from the byelections will nevertheless provide clues as to the future prospects of each of the parties.

An analysis of 103 federal byelections held since 1978 shows that a party's change in support in a byelection is an indication of how that party will perform in the next general election, both at the provincial and national level.

Since 1978, a party that has lost (or gained) vote share in a byelection has gone on to lose (or gain) support provincewide in the next general election 65 per cent of the time.

Parties have repeated nationwide either a gain or a loss in a byelection in the subsequent general election 66 per cent of the time. 

The chart below shows the relationship between the results of a byelection and a party's change in support in the next general election in the province in which the byelection was held.

In short, this means that a party that is, for example, on a negative trajectory in a byelection, will be more likely to perform worse in the subsequent general election — both in the province where the byelection took place and in the country as a whole.

Pre-2015 byelection hints

The 15 byelections that occurred between the 2011 and 2015 federal elections are a demonstration of this relationship. The Conservatives lost support in every byelection held over that time. The party then lost support in every province where the byelections had been held (with the exception of Quebec) in the 2015 federal election, as well as nationwide.

The New Democrats, another party to experience across the board decreases in support in 2015, saw their vote share drop in 13 of 15 byelections held during the last parliament. The Liberals, who made gains nationwide in 2015, increased their vote share in 13 of the 15 byelections.

Since 1978, parties that dropped vote share in byelections lost, on average, three points both at the provincial and the national levels in the subsequent general election. A party that increased support in a byelection also saw a three-point gain in provincial or national support in the next vote.

Time did not have a significant impact on the likelihood of a byelection's trend line being continued into the next general election — byelections held two years before the next general election (like the five that will be held on Monday) were just as likely to predict a party's subsequent performance nationwide.

Past performance not a guarantee

Still, a byelection is not a perfect predictor of what to expect a few years into the future. Roughly one-third of the time, a party that, for example, loses support in a byelection, subsequently makes provincial or national gains in the next national vote.

On Monday, it is possible that one or more parties will experience gains in some ridings and losses in others. In those cases it would be more instructive to look at the party's overall performance in byelections, as was the case for the NDP and Liberals between 2011 and 2015.

And it is important to note that there isn't necessarily a causal relationship between byelection and general election performance.

Nevertheless, a party is twice as likely to replicate byelection trends in the subsequent national election than to reverse them.

The notion that governments do badly in byelections is inaccurate — they retain their seats as often as the opposition. The line about byelections telling us nothing about broader trends is also incorrect.

But that won't stop the parties who take a drubbing at the polls from using that line.