On James Morton's blog, a commentator noted that "by-elections usually punish the party in government," a sentiment expressed by many Conservatives it seems, usually as a last-ditch reasoning against any sort of positive sentiment coming from their opposition. This, frankly, is a horribly incorrect notion that is born out of the idea that if the party in power wins in a by-election, they're doing good; if they lose, they're facing the wrath of an electorate.
While anyone could easily discredit this notion based on simple electoral politics, namely that singular elections in small districts more often than not end up locally-based rather than as a referendum on the government, I thought I'd just do a quick rundown of by-elections in Canada over the past two decades, who won them, and of course, the governments of the time.
Since 1993, there's been quite a few by-elections, but only 11 of those by-elections have resulted in the government party either winning or losing a seat. This list includes:
Brome--Missisquoi (1995, Bloc to Liberal)
Port Moody-Coquitlam-Port Coquitlam (1998, Reform to Liberal)
Windsor West (2002, Liberal to NDP)
Gander-Grand Falls (2002, Liberal to Progressive Conservative)
Perth-Middlesex (2003, Liberal to Progressive Conservative)
Temiscamingue (2003, Bloc/Independent to Liberal)
Levis-et-les-Chutes-de-la-Chaudiere (2003, Bloc Quebecois to Liberal)
Roberval--Lac-Saint-Jean (2007, Bloc Quebecois to Conservative)
Desnethe--Missinippi--Churchill River (2008, Liberal to Conservative)
Montmagy--L'Islet--Kamouraska--Riviere-du-Loup (2009, Bloc Quebecois to Conservative)
Vaughan (2010, Liberal to Conservative)
Looking at these we results, we see that 8 of the 11 by-election contests have been won by the governing party, while 3 were lost. This includes four Harper by-elections, and four Chretien by-elections. The three losses occurred on Chretien's watch, but during his comparatively more popular years just after the 2000 election (though Chretien was never that popular, he was hated more during the latter half of the 1990's than he was in the 2000's).
However, a good point to note here is that in the 16 by-elections during Harper's tenure, only one has been a Conservative-held riding (not counting Bill Casey), and it could hardly be considered a "toss-up," given that its Dauphin-Swan River-Marquette. That number will be up to 3 out of 19 with the resignations of Jim Prentice and Jay Hill, but their ridings are safe for the Conservatives as well.
Compare this to the Chretien/Martin era, where of the 32 by-elections, 22 of them were in Liberal-held ridings. The fact that they lost only 3 of them is amazing, and not all of them in safe ridings. Harper has yet to face such vulnerabilities.
But, nevertheless, the point is that it doesn't seem like governing parties, at least since 1993, do too badly in by-elections. If, of the 48 by-elections since 1993, governing parties have won 30 of them, how exactly do by-elections "usually" end up in government losses? That's nearly a 2-1 split. Heck, even of the 3 losses during that messy Chretien/Martin infighting period, the losses occurred in provinces (Ontario, Newfoundland) where the Liberals in 2004 either gained or maintained most of their vote share, yet the two wins in Quebec during the same period from the Bloc were counter-intuitive, because the Liberals lost most votes in 2004!
This isn't just easily demonstrated on the federal level. I can show the same thing occurring in the provinces, in the UK, and even in the United States. The fact is that by-elections usually end up turning on local conditions and the candidate, a lot more than on whatever the national scene is like at any given moment.
But, if the narrative helps the Conservatives, however false it may be, then it must be true, right?