Reading the Polls Wrong

I've always loved how the media reports on studies, especially scientific ones.  Usually they report the press release that comes with it, but sometimes they try to analyze the data on their own, usually with bad results, because they don't understand statistics and/or science.

The media also doesn't understand political polls, but this is mostly because they rely on partisans to help them translate it.  A couple weeks ago, the national polls were starting to show the race breaking toward Obama, and some people were declaring the race was over, while the Romney folks were protesting that the polls were oversampling Democrats.  Since the debate last week, the polls started moving toward Romney and now his partisans are saying that it's looking like he is going to win, while the Obama spinmeisters are trying to minimize the break.

So what's the truth?  It's still a tight race.

Now, I still think it's going to break one way or another and it won't be close, but reading the current polls it's not quite indicating that yet.   From what I can see, the numbers are coming down to the Party identification and the Independents preference.

When the polls showed Obama ahead, much of it was due to a higher number of Democrats being sampled over Republicans.  Now in 2008, the exit polls indicated a 39-32 break for Democrats.   It's unlikely that there will be that big a spread this time.  Historically about 90% of Democrats have voted for the Democratic candidate, and  90% of the Republicans have voted for the Republican Candidate.   Every % increase in the spread indicates a 1% increase in the margin before taking account of the independents.  

Independents on the other hand, have been pretty much split or leaning slightly for Romney in most polls, so the Democrat over Republican spread has been much of the margin.

Let's illustrate with the Pew poll.  On Sept 19th the released a poll that had Obama ahead by 8 51-43.  That poll had a 37-30% advantage for the Democrats and the Independents favored Obama 44-42%.   On Monday, they released another poll, this time showing Romney ahead by 4.  But in this poll the Republicans had a 3 point advantage in the sampling, and the Independents are going 46-42 towards Romney.

It's a trend but is it permanent?  Only the next data point will tell us.

The point is, that the media only reports the raw number most of the time without looking a the underlying data, or trend to give a better picture.  At this point it looks like the election will turn mostly on the net turnout of Democrats vs. Republicans in a 1:1 relationship to the difference.  i.e. if turnout is 36-33 Democrats, then Obama will have a 3 point advantage.  If the Independents stay close, the offset is about 3-4% diffence to a 1% gain.

In other words, if the Democrats turnout 3% better than Republicans and Independents favor Obama by 3% then Obama will win by 4%.  However if the Democrats turnout 3% better AND independents favor Romney by 12% (Say 52-40)  then it's a dead heat.  More realistically thought the Turnout and the Independents will likely go in the same direction.