Wednesday, October 10, 2012

The View From Topside

Nice view up here.  Rasmussen said it’s a tie as of Monday when I started writing this, but Gallup and Pew said Romney has the lead.  All kinds of state polls are showing movement in Romney’s favor, and the commentary since the first Presidential Debate (or, if you read the New Yorker magazine and have seen their latest cover, Romney versus the Empty Chair) has also reflected a drastic opinion change in Romney’s favor.   Not surprisingly, Obama supporters reject the idea that so many people have changed their mind so quickly.  The thing is, I tend to agree with them on that particular point.  While I do think that Governor Romney is now the front-runner and I do think he has some strong momentum, I also believe that what is happening is not a change of minds for the most part, but a matter of enthusiasm.

As an example, let’s look at the most recent Pew poll.  The Pew Research Center is not as established as Gallup (no one is), but they have been around since 2004, which – sadly –is longer than a lot of polling agencies.  They call themselves a ‘nonpartisan fact tank’, which is sort of a junk phrase, but they also do a lot of opinion polling.  Anyway, Pew released a poll on September 19 showing President Obama leading Governor Romney by 8 points, and more importantly, with Obama collecting 51 percent of the vote, a critical measure with the race entering the home stretch.

This is obviously great news for the Obama Camp, and very bad news for Team Romney. 

But Pew released another poll on October 8, this one showing Governor Romney leading President Obama by 4 points, 49 percent to 45 percent.  The poll indicates Obama lost six points of support while Romney gained six points, for a twelve-point swing in a single poll cycle.

This new poll is obviously great news for Team Romney, and very bad news for the Obama Camp.

But here’s the thing.  Are we really saying that one in every eight voters changed their mind because of the debate?  Sure, Romney won and looked good doing it, but the Romney supporters were already planning on voting for him, and I don’t believe huge numbers of Obama supporters decided to quit and go home, much less vote for Romney.  It could be that six percent of voters who are Independent swung from Obama to Romney, but since Independents usually make up no more than 30 percent of all voters at the highest participation level, that would mean one in every five Independent voter changed their mind because of the first debate.  Kind of hard to buy that.

I also reject the ‘conspiracy’ notion, that Pew wanted Obama to win and tweaked the results in the first poll … and apparently changed its mind after the debate in order to be on the winning side.  Polls are generally trying to be accurate, and to reflect the genuine opinion of the voters. 

So, to understand what happened, we go to the numbers.  Pew’s September 19-released poll contacted 3,019 adults nationally, of whom they determined 2,424 were Registered Voters, 2,192 were Likely Voters, and of the Registered Voters, 717 were Republicans (29.6%), 869 were Democrats (35.8%), and 757 were Independents (31.2%).   For comparison, the October 8-released poll contacted 1,511 adults nationally, of whom they determined 1,201 were Registered Voters, 1,112 were Likely Voters and of the Registered Voters, 403 were Republicans (33.6%), 396 were Democrats (33.0%), and 364 were Independents (30.3%).  That’s a +4 point change for Republicans, a -3 point change for Democrats, and a -1 point change for Independents in participation terms.  What else is interesting, is what happens if we look at support by party.   Republicans supported Romney 91-6 in the September poll, and 89-7 in the October poll, actually dropping a couple points.  Democrats supported Obama  92-6 in the September poll, and 93-5 in the October poll, gaining a point.  Independents supported Obama 44-42 in the September poll, but swung over to Romney in the October poll by the same 44-42 margin, a 4-point swing right there.  So, if the same weighting was used in both polls by party affiliation, Romney got a 1 or 2-point bounce in the polls, and that means the other 10-11 points of gain came from changes in the party affiliation weighting.   This demonstrates why the weighting of polls is so important in how results are presented.  A poll which over-weights or under-weights political participation can severely invalidate the results, in the same way that ignoring proper demographic weights for gender, age, or race can create unrealistic images. 

So what’s going on?  Is the party weighting due to bias?  I don’t think so, for that reason stated earlier that polls want to get the results right.  What happens is generally one of three things.  The Pew poll, it should be noted, apparently does not weight for party affiliation, but simply leaves the political affiliation un-weighted, so enthusiasm for a given party is reflected as well as stated support.  The problem with that point is the fact that Democrats are generally more eager to answer polls, possibly due to the fact that Republicans tend to be older than Democrats.  If this first answer is correct, the October 8 poll shows significant change because Republicans have been energized and Democrats demoralized by the debate results, but subsequent events could change that condition.

The second way polls handle party affiliation is to mirror the last Presidential election.  In 2008 Democrats represented about 39 percent of all voters, the highest level since 1996, while Republicans represented only about 32 percent of all voters (lowest since 1980).  Republicans may well argue that the 7-point disparity is unlikely to be repeated this year, but at least those polls can point to history for their basis.  An interesting variation on this is Rasmussen, which uses the 2010 election turnout results for its base.  Democrats argue that weighting is invalid because the mid-term election turnout may not be reflected in a Presidential election year, but Republicans can point to the 2010 election as the most recent historical model.  The actual election will show which of the two theories is correct this year, if either.

The third way polls weight for party affiliation is the most controversial.  Editors sometimes pick a number they like with little support beyond playing a hunch.  While this may seem to invalidate a poll, if the weighting is consistent through the year, the poll can still reflect trends and swings in opinion. 

With 27 days left in the race, some folks have already voted, many more already know how they plan to vote and won’t be swayed, and some have already decided not to vote, are too late to register, or plan to throw their vote away on the political equivalent of breaking wind just to make a bad odor in public.  Jay Cost wrote that no election is decided by October 1, which seems to run against a lot of what has been said and written this year. 

According to Cost, as many as twenty-two percent of voters won’t make their mind up until after October 1, with up to seventeen percent making up their mind in the last week before the election.  This tells us that a lot of people will be watching what’s going on and making up their minds on those last, final events. 

I am a bit skeptical on that last part, though.  I do agree that there will be undecideds up to Election day, and because polls push people to at least say who they ‘lean’ towards (which I still call an undecided) I think there are more undecideds than the four-to-seven percent announced in most poll results, but I also think that people know generally where they stand and how they feel about the incumbent and challenger once the general election gets underway; they are undecided mostly because they want their preferred candidate to close the deal.  Once they make that decision, they are pretty much locked-in, even if for the sake of appearing reasonable they say they are still deciding.  This happens a lot in certain states.  What I mean is, people in places like Texas or New York, California or Utah made up their minds pretty early in the race.  And the number of true undecideds is pretty low.  In Ohio or Florida, though, even if people have decided how they will vote, they may be telling pollsters they aren’t sure, because they recognize that their state might swing the election one way or the other, and so even if they know how they plan to vote, they may back off saying so in public for fear of looking like they decided too quickly. 

Today’s Gallup tracking shows a tie now among Likely Voters, and a 5-point lead for Obama among Registered Voters.  It’s just one poll, but then again it’s Gallup, which is as good as polls get for both transparency and historical record.  His Job Approval numbers are also up in Gallup, enough to give confidence to the Obama Camp if he can hold through October.  So, as good as last week for Romney, we’re still looking at a race where all sorts of possibilities remain viable.