E.J. Dionne makes much of the apparent shift in party identification in his column this morning, and there’s no question the GOP has fallen from the great heights it had reached earlier in this decade.
But it’s worth keeping a little perspective on just how unusual those heights really were. Dionne says “Democrats now outnumber Republicans 37 to 28 percent.” That’s certainly an improvement for the Democrats over the 2004 split (when Democrats only outnumbered Republicans 32 to 29 percent, according to the University of Michigan’s American National Election Study series, which is the gold standard for historical party ID figures). But that 2004 number, and indeed Republican party identification from 2000 until 2005, was very far above the historical norm for Republicans. In 1984, when Ronald Reagan won 49 states, the University of Michigan survey found a party split of 37 to 27 in favor of the Democrats — roughly the same as today. In fact, according to the Michigan series, the GOP has only been above 28 percent in one presidential election year since 1968 (when they started measuring party ID with a standard question). That one time was 2004. The GOP has been above 28 percent in non-presidential years only twice: in 1994 and in 2002. The average GOP party identification over the last forty years has been 24.9 percent. The average Democratic Party ID figure has been 38.1 percent.
(For the non poli-sci nerd community, to reach the historical party ID table, go here and enter “VCF0302” in the “row” box and “year” in the column box).
Party ID statistics themselves are not great predictors of electoral outcome (for instance, 41 percent of voters labeled themselves Democrats in 1980, and 35 percent did in 1992), but obviously trends in the numbers do tell us something important. The trend in the past two or three years has been downward for the GOP, and that surely spells trouble. But what kind of trouble and how much of it is very hard to say from a look at party identification numbers. The key is which party, and which candidate, does better with the very large portion of voters who identify as independents, and also whether a particular candidate can win a significant number of voters who nonetheless identify themselves as supporters of his opponent’s party. In the presidential race this year, both questions are wide open, and both candidates have unique opportunities. That makes the party ID figures less useful than usual, not more.