2008 in Review - The Final Harvey Gantt review

From June to November, watching the Electoral polling maps was made more interesting by considering the Harvey Gantt Problem.  Only at the very last did it appear that Obama was going to win even if it was true.  This is the analysis I wrote on the last day before the election.

In this election, there has been substantial discussion of “the Bradley effect.”  Named for Mayor Tom Bradley of Los Angeles, the Bradley effect posits that polls are not accurate when it comes to electing African-Americans.  In the 1982 gubernatorial race, Bradley led in the polls and the exit polling suggested that he was going to be the winner, only to find out that the race went the other way.

The Bradley effect presumes something extremely cynical about people - that they will tell pollsters that they are going to vote for the African-American, but then they don’t.  My problem with all the speculation around the Bradley effect is that it is too cute.  It presumes that people will lie when there is no known reason for them to do so.  It presumes that the voter is going to think “well, I’m voting for McCain, but this pollster on the phone - who has interrupted America’s Got Talent to ask me who I’m voting for - might think I am a racist and that would be a bad thing, so I had better lie to him and tell him that I’m voting for Obama.”  It is too much to expect for people to go through that exercise.

One of the problems with someone positing a thesis like the Bradley effect is that some number of people jump on board without examining the original premise.  In this case, the problem is that Bradley was never the significant favorite - in the final weeks of the election.  He was favored in polls, but by small enough margins to be within the margin of error.  In my opinion, the Bradley effect has always been the wrong way to look at the question, if it remains a question, of why a black man cannot get elected in America.  The explanation to me is called the Harvey Gantt problem.

Using the current 538 composite polling for the map, there are three kinds of states: (1) where Obama leads by more than 50%, (2) where McCain leads, and (3) where Obama leads without reaching 50% (or neither candidate has a statistical lead).  The Harvey Gantt states - where Obama leads in the polls by less than 50% - are Florida, Missouri, North Carolina, Nevada, and Ohio.  The toss-ups, which under this theory all go to McCain, are Indiana, Montana, and North Dakota.  Interestingly, Pennsylvania - where McCain has invested a great deal of effort - and Virginia, which hasn’t voted Republican in a generation, are actually outside the polling for the Gantt Problem (which means that if McCain prevails in either one, the polls were way off and there is something to the Gantt Problem).

Under this calculator, Obama still wins the election, because he exceeds the 50% margin in states with a total of 286 electoral votes.  However, if that is the exact number, it suggests that there is validity to the Gantt Problem.

It isn’t clear that the Harvey Gantt Problem is real.  No one can hope that it is.  But if it exists, it says all too clearly that we still have a ways to go in overcoming racism.  An Obama victory doesn’t necessarily lay this particular thesis to rest, but it is definitely bright-line test to measure how far we have come.

So what was the real bottom-line?  Four of the five states where Obama led but had a less than 50% margin - Florida, North Carolina, Ohio and Nevada - all went for Obama; and the fifth, Missouri, was so close that a recount would have been ordered if those votes had been needed to decide the election.  With two of those states in the old segregated South, I think we can lay this hypothesis to rest.  I didn't say it was right, but I'm glad that I was wrong in thinking it might be.