The Bradley Effect -- A Bad Analogy

A lot of pundits today, in their scramble to hypothesize, have suggested that Hillary Clinton's victory in the New Hampshire primary may have been, at least in part, the result of the Bradley effect.

The problem with the analogy is that it misunderstands what the Bradley effect was.  When Tom Bradley was running in a GENERAL election in California, pollsters were calling Democratic and Democratic-leaning voters to ask them about their preference for governor.  There were scores of Democratic voters who didn't want to vote for a black man but were afraid of being judged for uncharacteristically supporting a Republican.  They had incentive to lie.  And they did.  Bradley, who had been expected to win by 7 points, lost by 1 point.

Unlike New Hampshire, every other example of the Bradley effect involves a general election, not a primary.  In a situation like last night in New Hampshire, voters had no reason to lie to pollsters.  If someone wasn't willing to vote for Barack Obama because of his race, they could have said they were voting for Hillary.  There would be no need to hide, no need to lie.  When a Democrat tells a pollster that they're voting for Hillary, no one raises an eyebrow toward racism.  There just doesn't seem to be any rationale for a New Hampshire Democratic voter to lie to a pollster by saying that they were going to vote for Obama.  

•  I'm exceedingly uncomfortable

  with the whole racial tenor of the discussion here at Daily Kos today (which I'm not necessarily lumping the diarist in with).

  Way too many diaries suddenly going on and on about this "Bradley Effect" and "Wilder Effect," seeking to explain away Hillary's success by essentially calling her supporters closet racists.

  The surprise N.H. loss seems to be bringing out a nasty and disturbingly familiar undercurrent in the camp of "hope" and "change."

  by Jim J on Wed Jan 09, 2008 at 02:30:05 PM PDT

  •  RE:

    You're talking to an Obama supporter here.  I'm not trying to say race played a role.  I'm trying to say it didn't.

    by dloewe on Wed Jan 09, 2008 at 02:32:42 PM PDT

    •  I'm not really directing that at you

      Though I do wish this whole thing would be examined here a lot less hastily, if we need to go there at all.

      by Jim J on Wed Jan 09, 2008 at 02:36:10 PM PDT

      •  We definitely need to go there if it's a myth...

        ...because it needs debunking.

        I tend to think it's much less true than people are saying.

        by JedReport on Wed Jan 09, 2008 at 02:40:29 PM PDT

    •  dloewe...

      ...look at the data i put on my site http://www.jedreport.com about the #s.

      also find out which pollsters are robo versus non robo, and if their results differed. if they were the same, tends to discount bradley effect, because why would someone lie to a machine?

      also, hillary led obama by 1.5 inthe post-xmas polls in iowa, yet obama won by 8 or 9.

      as you'll see i'm not a huge obama fan, but if this bradley effect is a myth, you'd be doing a great service to the nation by helping to debunk it, because if it's a myth but people believe it, it will have a very harmful effect on race.

      i think if you did a more data-driven diary in a day or two people might also be in a more receptive frame of mind.

      by JedReport on Wed Jan 09, 2008 at 02:39:33 PM PDT

•  I totally agree

  There just doesn't seem to be any rationale for a New Hampshire Democratic voter to lie to a pollster by saying that they were going to vote for Obama.  

  There were other choices in the field for a race based preference which I conceded does exist on the part of some in the electorate including Democratic voters. I don't know what happened in NH but it wasn't due to the Bradley Effect.

  by mbair on Wed Jan 09, 2008 at 02:33:33 PM PDT

  •  potential counterpoint

    exit polls showed no effect, but obama led by 8 in the 15 poll average of monday/tuesday released polls.\

    there might have been a slight bradley effect with phone poll takers whose voices were identifiably african-american; that would explain the gap between exit polls and phone polls.

    question is whether the same gap was in robo polls, where there would have been no bradley effect at all.

    does anybody know which polls are robo polls?

    by JedReport on Wed Jan 09, 2008 at 02:36:06 PM PDT

    •  I disagree

      You're assuming that a person would tell a live human being that they would vote for Obama because they didn't want that person to think they were racist. That's not an accurate way to look at a the poll. The respondent can just say, "He doesn't have enough experience, I like him but he's not ready to be president." "I like Edwards better." "Clinton has more experience." Whatever.

      They don't have to justify their choice to a pollster over the phone based on racial preferences which do exist.

      The Bradley Effect was, as is pointed out in the diary, an anomaly in a general election with only two candidates. Indes in CA usually voted a certain way until a black man represented the Democratic party. That is not analogous to a multi-candidate field full of white candidate choices. It just isn't the same thing. Besides I think the candidate that would have benefited the most from a Bradley Effect would have been John Edwards. He did not. Only the Obama-Clinton numbers were off. The Edwards number was right on the money in the polls leading up to the vote.

      Right?

      by mbair on Wed Jan 09, 2008 at 02:49:00 PM PDT

      •  Yes you are right.

        I was proposing something that would make the argument seem better but actually isn't relevant now that I think about it.

        I've got a meta question for ya', since you've been around here longer than I have.

        In the midst of this diary, my TU status seemed to be revoked.

        It's not really a problem because I've only dropped 2 TRs ever, but I can't figure out why my TU status would have been revoked. It's nice being able to see hidden comments.

        Any idea?

        by JedReport on Wed Jan 09, 2008 at 02:58:08 PM PDT

        •  Happened to me, too, yesterday,

          and I am a good little Kossack who posts regularly, and doesn't participate much in candidate diaries.

          by blindyone on Wed Jan 09, 2008 at 03:10:48 PM PDT

          •  Now, you mention it....

            I hadn't noticed, but mine is gone too.

            No doubt some glitch in the system, as I rarely use mine, either.

            by kpardue on Wed Jan 09, 2008 at 03:19:38 PM PDT

            •  I was just going to give you a "courtesy"

              rec for responding to me, and lo and behold, my TU status is back. Nice. On the other hand, the site is slow and dragging for me, so I can't read the comments on diaries, and some of my comments aren't posting normally. I guess DK is drowning in traffic right now. I tried to rec your reply and I got a "failed" notice.

              by blindyone on Wed Jan 09, 2008 at 03:41:49 PM PDT

              •  ct has a diary on this

                everything seems fixed now!

                http://ct.dailykos.com

                by JedReport on Wed Jan 09, 2008 at 04:08:00 PM PDT

                •  I still don't have mine back

                  I should have read the rest of this stuff here before commenting, but TU is time dependent.

                  by mbair on Thu Jan 10, 2008 at 05:48:35 AM PDT

        •  I just lost TU this morning, too

          Don't know. TR's can pull your average rating number down, but I usually lose it because I don't post enough. The average comment rating is weighted by time so recent comment ratings count for more than stuff in the distant past.

          There could also be a new threshold for this site on TU. I don't know if it's just a software thing for the overall site or a TR thing in your case. Doesn't sound like a TR thing in your case because those are time weighted in your average as well. Unless those TR's were very recent they wouldn't cause you to lose TU.

          But overall, if you were borderline and got a lot of TR's recently they count more than a tip jar or comment you might have gotten a week ago. Does that make sense?

          I think the rule of thumb is that you need a comment with about 30 recs once a week to maintain TU. Then everyday that average comment rating declines just because it's a new day and you have to build up your rating daily by getting recs for your comments to maintain TU.

          Back on point, I think there is a racial preference to votes on both side, our side less so but still. But in this case with a multi-candidate field and a John Edwards right there in the pack that whatever racial bias was in the electorate was already built into the pre-election polling. Because basically, in this situation, those voters had to be racist and not sexist. They had to be against a black man for president and for a woman who was also a Clinton for president. I think that's highly unlikely.

          by mbair on Thu Jan 10, 2008 at 05:47:06 AM PDT

•  The exit polls showed no effect

  according to my analysis of the exit polls, they showed hillary winning.

  (see http://www.jedreport.com/...

  NH exit poll takers are almost certainly mostly white. OTOH, phone poll callers may be more racially diverse because they can be from anywhere int eh country. Slate had a good article about this.

  So the exit polls might not have been bradley effected, but some of the non-robo pollsters may have been.

  by JedReport on Wed Jan 09, 2008 at 02:33:49 PM PDT

•  I don''t know.

  My initial response was that there was no Bradley effect. However, I don't quite buy the argument that someone would deny voting for a black person in the general election out of embarrassment in the GE but not in the primary. The reason would be the same for both. One would not want to appear racist.

  Now I don't know what caused the last minute shift, but Larry Sabato was just on Hardball and he was involved in the polling for the governor's race in VA when Doug Wilder barely pulled off a win after having held a large lead prior to the election. He did not say that this was the explanation but could not rule it out.

  As they noted, the exit polling of Democrats showed Barack with a 5 point lead, yet he lost by 2-3 points. That apparently does not typically happen. By the way, they also noted that the polling on the GOP side was entirely accurate.

  That said, I absolutely hope this is not the case.

  by smartdemmg on Wed Jan 09, 2008 at 02:34:26 PM PDT

  •  saw that same roundtable on Hardball

    very interesting. I don't know what to think. these talking heads are uncontrollable

    by PeckingOrder on Wed Jan 09, 2008 at 02:42:24 PM PDT

    •  I know, that's why I was left

      with a question mark on the question. There was no agreement one way or the other, just a number of hypotheses put forth.

      by smartdemmg on Wed Jan 09, 2008 at 02:44:48 PM PDT

    •  RE:

      yes, but larry sabato is barely saying anything at all.  he's saying these things need to be investigated and until then, we shouldn't draw conclusions.  this may be good advice, but it has very little to do with whether the bradley effect happened here.

      bottom line is, if you can't explain the incentive for democratic voters to lie to pollsters, you don't have it.  and if the best explanation we get is that, maybe there was polling companies using African American callers, we're starting to flock a little close to the conspiracy theory flame.

      by dloewe on Wed Jan 09, 2008 at 02:46:26 PM PDT

•  I doubt there was a Bradley effect

  but to play devil's advocate, consider independents, it is possible that some of them might have done that.

  But I think it is a stretch, much more likely was the swing of young women and independent women to Hillary (they may have been with Hillary most of the way, but may have leaned Obama on 1/5-1/7 as the Iowa results were being digested).

  by IhateBush on Wed Jan 09, 2008 at 02:36:02 PM PDT

•  The Bradley effect is not limited

  to the general election.

  See Wikipedia entry on the Bradley effect specifically mentioning that the same occured in an all-D 1988 Presidential primary in Wisconsin between Jesse Jackson and Mike Dukakis.

  by Drgrishka1 on Wed Jan 09, 2008 at 03:40:11 PM PDT

•  "The Hillary Effect"

  I think there may be something new going on here, and thats that people are embarrassed to admit that they voted (or will vote) for Hillary.

  So its not a racial thing, its that everybody thinks youre suppose to hate Hillary so people lie that they actually voted for her or that they will vote for her.

  Obama supporters better hope theres no "Bradley effect" and they better stop arguing that there is, because if theere really is a Bradley effect then that means he has alot less chance of winning the general election.

  The Hillary effect, if it exists, suggests the opposite. There's hidden support for her that doesnt show up in polls.

  by DigDug on Wed Jan 09, 2008 at 03:54:21 PM PDT