One argument for why Democrats should be relatively pessimistic about 2016 is that voters get fatigued about the incumbent party, so getting the third term is generally a grind at best. And there is a historical pattern of these elections being toss-ups even in decent economic circumstances. One problem with inferring things from these patterns is the small n. But the even bigger problem, as Chait says, is most of these elections took place in a context of heterogeneous parties that no longer exists:
The logic that predicts a toss-up election is rooted in the perfectly sound assumption that the historical models give us the best guide to the future. A third straight term from a party whose president has middling approval ratings sits right on the probability fault line, historically. As Nate Silver writes, “these cases default to being toss-ups.”
The trouble is that almost all those cases are drawn from a historic period that is very different from the current one. During the 20th century, the two parties were extremely heterogeneous. The Republican Party had a moderate wing that dominated its presidential elections for most of the postwar years until Ronald Reagan. Democrats had a powerful southern conservative wing. In that environment, the old folk wisdom, “Vote the man, not the party,” made a great deal of sense. In that environment, large chunks of the electorate swung easily from one party to the other depending on transient factors, like the current state of peace and prosperity, rather than deeper values.
The splitting of American politics into two coherent ideological parties with very little programmatic overlap changes things. Voters who are fundamentally attached to one party or the other are not going to abandon their team merely because their party has held onto office for too many terms, or because the other party’s president is presiding over a nice recovery. Those factors are not meaningless because some swing voters do still exist. And performance can change voter perceptions to a degree; a deep recession might make some Democrats doubt their party’s economic program. But these temporal effects are muted.
Incumbent fatigue is a plausibly significant factor in a context in which most of the electoral map was theoretically in play. 45+ state landslides used to occur with some frequency. In 1988, George H.W. Bush — winning a third consecutive Republican term! — won California, Maine, Vermont, Maryland, and New Jersey, and lost New York by only 4 points. Those days are gone. It’s hard for people to internalize the consequences of realignment and entrenched partisanship — my favorite dumb argument about the 2000 election, “Durr, Al Gore was so bad he couldn’t even win his own state, durr [drools],” reflects among other things a failure to grasp what really should be obvious about the contemporary electoral map.
I don’t know what the structural context of the 2016 elections will be. But I don’t think that superficial patterns derived from a pattern of partisan norms that is deader than John Lindsay tell us much of anything.