Kevin Drum suggests that Democrats should calm down about the possibility of a long nomination fight hurting their chances in the fall:

The hot topic of conversation right now is the proposition that a long, drawn-out Democratic primary runs the risk of destroying the party and putting John McCain in the White House. So for the good of the country, Hillary should withdraw.

Now, this might be true. But I’d like to offer a historical counterexample: 1968. Consider. The Democratic incumbent president was forced to withdraw after a primary debacle in New Hampshire. The Vietnam War had split liberals into warring factions and urban riots had shattered the LBJ’s Great Society legacy. A frenzied primary season reached all the way to California in June, culminating in the assassination of Robert F. Kennedy. The Democratic Convention in Chicago was a nationally televised battle zone. Hubert Humphrey, the party’s eventual nominee, had never won a primary and was loathed by a significant chunk of the liberal community. New Left radicals hated mainstream Democrats more than they hated Republicans.

In other words, this was the mother of all ugly, party-destroying campaigns. No other primary campaign in recent memory from either party has come within a million light years of being as fratricidal and ruinous. But what happened? In the end, Humphrey lost the popular vote to Nixon by less than 1%. A swing of about a hundred thousand votes in California would have thrown the election into the House of Representatives.

If long, bitter, primary campaigns really destroy parties, then Humphrey should have lost the 1968 election by about 50 points. “Bitter” isn’t even within an order of magnitude of describing what happened that year. And yet, even against that blood-soaked background, Humphrey barely lost.

This is all true, but it’s important to remember that Humphrey drastically underperformed in 1968 relative to what we would expect given the state of the economy at the time (a result that is often attributed to Vietnam War deaths). We can’t quantify what damage was done by the polarizing primary campaign, but it’s hard to see how it would help.

Democrats risk a similar scenario — a destructive primary campaign could turn a possible rout in November into a 50/50 coin flip a la 1968. And even if they do win, any significant reduction in the popular/electoral vote margin could have significant legislative and political consequences going forward.

On the other hand, as John Sides points out at The Monkey Cage, the most prominent political science article studying this question did not find a significant negative effect (gated JSTOR link):

Divisive Primaries and General Election Outcomes: Another Look at Presidental Campaigns

Lonna Rae Atkeson

American Journal of Political Science, Vol. 42, No. 1. (Jan., 1998), pp. 256-271.

Theory: The divisive primary hypothesis asserts that the more divisive the presidential primary contest compared to that of the other party the fewer votes received in the general election. Thus the party candidate with the most divisive primary will have a more difficult general election fight. However, studies at the presidential level have failed to consider candidate quality, prior vulnerability of the incumbent president or his party, the national nature of the presidential race, and the unique context of each presidential election campaign. Once these factors are taken into account presidential primaries should have a more marginal or even nonexistent effect in understanding general election outcomes.

Hypothesis: Including appropriate controls for election year context in a state-by-state model and creating a national model that controls for election year context, candidate quality, and the nature of the times should diminish the effect of nomination divisiveness on general election outcomes.

Methods: Regression analysis is used to examine the effect of presidential divisive nomination campaigns on general election outcomes.

Results: Once election year context in the state-by-state model is controlled for, divisiveness has a much more modest effect. This modest effect does not appear to change general election outcomes. In addition, the election year model, which posits that presidential elections are national elections and not state-by-state elections, indicated that divisiveness was not significantly different from zero.