If it were up to me, I would eliminate special elections for the House of Representatives entirely. They make sense when it comes to the Senate, where every state has only two senators and terms run six years, meaning a vacancy can leave a state without significant representation for an extended period of time. But when a congressman dies or retires and there's another election to fill that critical 1/435th portion of the lower house's lawmakers in a few months, do we really need to mobilize the state's electoral resources, spend millions of dollars, and get a bunch of retirees to haul themselves down to the polls, only to do it all again before you know it? Hardly.
The other objectionable thing about special elections is that because they're almost always the only election happening at that moment, they not only get an inordinate amount of attention, the results also get absurdly over-interpreted. This is a symptom of what we might call the Consequential Imperative among the press (note: if you have a better moniker for this that could propel me to the front rank of contemporary neologism-coiners, hit me up on Twitter). The Consequential Imperative is the impulse, the desire, the need to assert that whatever a journalist happens to be reporting on is very, very important. So for instance, if your editor sent you down to Florida to do a week's worth of stories on the special election that just concluded there, you are extremely unlikely to write that this election was a contest between a couple of bozos, and means next to nothing for national politics (unless you're Dave Weigel, who for some reason seems to be almost the only reporter capable of saying such a thing). It's the same impulse that causes every gaffe, polling blip, and faux-controversy of every campaign to be presented as though it could dramatically alter the outcome of the election, despite all the experience telling us it won't.
What happens after every special election is this: The losing side says, "This means nothing!", while the winning side says, "This is a bellwether, signifying more victories to come for us!" And the press almost always agrees with the winning side, whichever party that happens to be, because the Consequential Imperative dictates that, like every other political event, this one must be of great consequence.
So in the case of yesterday's special election in Florida, we get articles like "Why a Republican Wave In 2014 Is Looking More Likely Now" (National Journal) and "Florida Loss Big Blow to Democrats' 2014 Hopes" (Politico), explaining that the results of this low-turnout election in one district in Florida can reasonably be extrapolated to tell us what will happen in the November 2014 elections.
As it happens, this race was decided by less than 3,500 votes. To believe that it emphatically means one thing for election outcomes all over America eight months from now, whereas if those 3,500 votes had gone the other way it would have just as emphatically meant the exact opposite, is just absurd. But, you may be saying, that's because the Republican won! And if the Democrat had won, I'd be saying it really was significant! Well, no. Special elections don't mean anything beyond deciding which person is going to represent that district until the next election. They may be interesting for one reason or another in and of themselves, but they're never a harbinger or a bellwether of any national trend. If you ever catch me saying otherwise, feel free to call me a hypocrite and a fool.
