Kennedy’s Decision to Back Obama Not So Decisive After All

by Stuart Rothenberg February 13, 2008 · 11:05 PM EST

Massachusetts Sen. Edward Kennedy’s endorsement of Sen. Barack Obama (D-Ill.) for president was a huge story in the national media.

Dave Espo of The Associated Press said the Senator “is in a position to help Obama court voting groups who so far have tilted [Sen. Hillary Rodham] Clinton’s [D-N.Y.] way.”

Almost everyone who commented on Kennedy’s Jan. 28 endorsement said it was a big deal. Admittedly, I disagreed, preferring to stay consistent with my column of July 16, 2007, in which I argued that most endorsements in high-profile presidential contests don’t matter much.

Now that last week’s primaries are over, we can see whether the endorsement mattered. We can also evaluate the exit poll question that asked about the impact of the Kennedy endorsement.

Establishing (or disproving) a causal connection between an endorsement and a vote choice is difficult, leaving a lot of room for differing interpretations. But that should not dissuade us from taking a look at the available data.

In a piece published before Kennedy’s formal endorsement, Susan Milligan of The Boston Globe said Kennedy would campaign for Obama by focusing on “Hispanics and labor union members, who are important voting blocs in several Feb. 5 states, including California, New York, New Jersey, Arizona and New Mexico.” Espo made a similar point, though he added “lower-income, older voters” to his list.

Kennedy did indeed travel to those states and try to woo those voters for Obama. In the five states that Milligan mentioned, Clinton won four and is narrowly leading in the fifth (New Mexico). It’s possible, of course, that she would have won them even more convincingly had Kennedy not endorsed and campaigned for Obama, but the outcomes certainly at least suggest the endorsement’s effect was minimal, at best.

The Massachusetts Senator’s greatest potential impact was widely viewed as among Hispanics. Yet it was Clinton who generally won Hispanics, often overwhelmingly.

Clinton carried Hispanics in California (with 69 percent), Arizona (55 percent) and New Jersey (68 percent). In Illinois, Obama’s home state, Clinton drew “only” 49 percent of Hispanics. She ran 8 points better among them than among white voters in the state. And in New Mexico, the site of one of Kennedy’s rallies for Obama, Clinton won Hispanics by 20 points (56 percent to 36 percent), but lost white voters to Obama.

Clinton also won older voters and union members in most states, two other groups that supposedly might be swayed by Kennedy’s support for Obama. Given all of these results, it’s difficult to make the case that Kennedy’s endorsement was important.

Finally, there is the exit poll question that asked Super Tuesday respondents how important the Kennedy endorsement was, and it is to this question that I’ll now turn.

I checked 10 Super Tuesday states with primaries and found that in two of them (Illinois and Arizona), at least half of those polled said that the endorsement was “very important” or “somewhat important.” In another six states (California, Massachusetts, Georgia, Connecticut, New Jersey and New Mexico), 43 percent to 49 percent described the endorsement as very or somewhat important.

At face value, those responses suggest that the endorsement was significant. But if you look deeper, it appears that instead of the endorsement prompting people to vote for Obama, the causality worked in the other direction.

I believe that people who voted for Obama — and would have voted for him anyway — told pollsters that the endorsement was important, even if it actually had no impact on their vote choice.

As Obama supporters, they were excited by the endorsement. Moreover, journalists and talking heads told them the endorsement was a big deal. Given that, when pollsters asked about the endorsement, many respondents said it was important.

Here’s the evidence: In California, according to the exit poll, 19 percent of Democrats said Kennedy’s endorsement was “very important.” Yet almost half of those people (45 percent) voted for Clinton. Another 50 percent in California said Kennedy’s endorsement was “somewhat important,” yet almost half (47 percent) voted for Clinton.

In other words, almost half of voters in California who said the Kennedy endorsement was important didn’t vote for Obama. And that trend appeared everywhere else as well, thereby undermining the responses.

In only two states, Georgia and Illinois, did voters who said the endorsement was “very important” go overwhelmingly for Obama. And in both cases, other factors than the Kennedy endorsement (race and home-state loyalty) seemed far more important.

It’s wise to be very skeptical about survey questions that ask voters to identify their own motives, and the Kennedy question is a perfect example of why.

Ultimately, the Kennedy endorsement wasn’t important. It didn’t change things. That doesn’t surprise me. As I have argued before, this is the kind of race where voters look at the candidates and make up their own minds.