A few sites I like to monitor for excellent insight into polling data are the oft-cited Electoral-Vote.com and FiveThirtyEight.com.
It’s useful for evaluating whether the news stories make any sense. MSN says the race is tightening; the polls show that the opposite is happening. I just saw a news story implying that Pennsylvania, with 21 electoral votes, is up for grabs and thus hotly contested. It’s actually almost 60-40 in favor of Obama, and FiveThirtyEight says Obama’s odds of winning the state are 99%, with McCain’s at 1%. Not such a hot contest.
Meanwhile, other sources talk about how it’s possible Obama will actually win McCain’s home state of Arizona. Indeed, Electoral Vote shows that it’s a weak GOP hold, with the most recent poll 50% McCain, 44% Obama. It’s not clear how current the numbers are on this one, but Five Thirty Eight puts Obama’s odds of stealing Arizona at 3%. So I guess I can’t contest calling it possible, but it’s really not worth your time to talk about it happening, because it almost certainly won’t.
Meanwhile, the two sites make it easy to see the real battleground states. Indiana is pretty much 50-50; Electoral Vote says it leans ever so slightly blue, and 538 says it leans ever so slightly red. Missouri’s really close, too, and also has 11 votes up for grabs; 59% odds of Obama, 41% odds of McCain, and a total tie according to EV.
For a real illustration of how probability versus polls give different numbers, take a look at North Carolina. EV says it leans very slightly blue, with Obama having 49% of the vote and McCain having 48% in the latest poll. 538 puts the odds of an Obama win there at 64%. Georgia leans a bit red, 46% Obama to 49% McCain, yet 538 says Obama doesn’t have much of a chance: 88% odds of a McCain victory. Florida’s another leaning-blue state, but 538 gives Obama a 79% likelihood of winning.
Of course they’re all just statistical predictions based on (quite) imperfect data. They can only look at the past. But both sites are run by people who are certainly much better with statistics than I am, and both sites seem to give much more insight than many of the braindead articles declaring that the race is tightening, and that the candidates are duking it out in Pennsylvania. If you’re going to report on the close races, you might at least consult some polls. (Another article discusses how it’s possible Obama will win the popular vote and yet lose the electoral college as happened to Gore in 2000; it’s possible just like McCain winning Washington, DC: theoretically possible, but the odds are very slim. Well under 1%, according to FiveThirtyEight.com.) In short, much of the news seems to be doing a pretty lousy job of incorporating actual statistics into their articles.
But then again, the reminder of Al Gore (and John Kerry) brings up another point: just like happened in New Hampshire’s January ’08 primary, the polls get it all wrong sometimes. (Or maybe, it’s the notoriously inaccurate electronic voting machines… Don’t get me started on those.) So trust the statistics over hot air news stories, but don’t trust statistics too much, either. November 4th is still a week away.