Patrick Egan in the Washington Post argues that presidential approval ratings is a better predictor of election outcomes than polling data about party identification.
“The figure at below left displays the relationship between party identification in the electorate (or “macropartisanship,” as we political scientists like to call it) in the year before a presidential election year and the Election Day outcome. For each of the 16 presidential elections extending back to 1948, the Democratic presidential candidate’s share of the popular vote is plotted against the Democratic Party’s average affiliation advantage in the year before the election year.”
“The result? Party identification more than a year in advance of the election predicts nothing about how the election will ultimately turn out. (To be wonky about it, party identification insignificantly predicts the opposite of the final election outcome. A similar analysis, which I don’t show here, finds no relationship between change in party identification the year before an election and the election’s ultimate result.)”
“Compared to any other polling data regularly covered by the media right now, Obama’s approval rating is the only number that history demonstrates tells us something meaningful about which party may win next year’s presidential election.”