If I were the DNC, I would choose Columbus, Ohio, for the 2016 convention, if only to neutralize any statewide advantage that the Republicans stand to gain from holding their convention in Cleveland. Not that it would be a home run for the Democrats. More like a throw to first, to keep the runner honest — and maybe pick him off, which could change the outcome if the score is close. Most wins do not involve walk-off grand slams. It’s a game of inches.
“Election scholars do not find parties more likely to win the states in which they host their nominating confabs, and may in fact do slightly worse than usual in them,” Kevin Williamson writes, but I wonder. He pointed me to this piece by Harry Enten, at FiveThirtyEight, who found on average a 0.4 percent boost going back to 1964.
It’s too small to be significant, according to Enten, but states have swung by smaller margins, and in any case what’s at stake in battleground states, including Ohio, are typically slivers of one or two or three percentage points. In 2012, Obama won Ohio by a hair less than three points, and 0.4 percent would be a sizable chunk of that. The approximately full-point boost that Fred Schwarz found when he crunched the numbers the same way Enten did (Fred looked at conventions going back only to 1988) would go a good way toward flipping Ohio to the Republicans in 2016. Enten is correct, though, that his method — which is also Fred’s — isn’t nearly granulated enough to ensure that the tiny figure he arrives at is more than statistical noise.
For an illustration of how resistant the data are to easy interpretation, consider 1980, when the Republicans convened in Detroit and then carried Michigan. The Mitten State went more Republican than the nation did but by a much smaller margin than in 1976. But the Republican candidate in 1976 was a native son, Gerald Ford. This would suggest that the native-son effect is stronger than the convention-site effect, not that the convention-site effect is zero or negative.
By Enten’s reckoning, when a party meets two conditions in a presidential election — it had a good year nationally, improving its performance since the previous election by a wide margin, and it held its convention in a stronghold, a state that last time voted for the party’s presidential nominee by an overwhelming margin — that almost guarantees a score that will indicate, wrongly, that the party’s performance fell off in the convention-site state.
For example: If party R in a presidential year gains ten points nationally, going from from 45–55 to 55–45, it picks up 18 percent of the vote that went to its rival in the previous election — quite a feat. But take a state where party R already got 65 percent of the vote last time. If it it wins over a full 20 percent of that remaining 35 percent, that would register as a gain of only seven percentage points. Enten would subtract 7 from 10 and conclude that party R performed three percentage points worse in that state than it did nationally, but we would come closer to the truth by subtracting 18 from 20 and concluding that the party performed better in the state by two points.
The size of the state and the city’s size relative to it matter, too. A GOP convention in Pittsburgh, for example, could succeed in turning some independents in Wexford from the D to the R side of the fence, while their counterparts in Yardley were untouched by the convention, which from their point of view might as well have been held in Honolulu. Philadelphia and Pittsburgh share a state in the same sense that Brooklyn and Buffalo do. We look for statewide bounces because those are all that matter for the Electoral College, but to identify the specific effect of a convention site, we would have to confine our search for it to the metropolitan area of the city in which the convention was held. Define it as that city’s media market, if you like.
And a small market should be more susceptible than a big market to a convention-site effect. New York? Fuggedaboutit. Don’t think that with a mere convention you could win it over even a little. You wouldn’t even register on its radar. You would on Cleveland’s, though. Cleveland proper, where some precincts went 100 percent for Obama in 2008, will probably not budge from its place at the deepest blue end of the spectrum, but the outer suburbs and the exurbs may be capable of turning a redder shade of purple or darker shade of red.