In the last two days, we have seen polls that say that Donald Trump is winning by 2% and polls that say that Hillary Clinton is winning by 12% as well as all sorts of numbers in between. The media has focused on the ABC poll that put Hillary up by 12% and basically ignored the IBD poll that put Trump up by 2%. That is not surprising since nearly all of the media supports the Democrat. But none of this really tells us where the race stands at the moment.
To try to make some sense of this, I looked at the internals of three polls: the ABC tracking poll, the IBD poll and the CNN poll that was released this afternoon which shows Hillary up by 5%. The most surprising finding of that review is that the results of the polls are nearly totally explained by the mix of the people polled. ABC polled 36% Democrats and only 27% Republicans. CNN polled 31% Democrats and 26% Republicans. (The remainder were independents.) If you adjust the ABC poll so that the percentages of Republicans, Democrats and independents are the same as that used by CNN, the outcome puts the results within 1% of the CNN poll results. If you adjust the IBD poll percentages of those polled to the ones used by CNN, the results again come within 1% of the actual CNN results. In other words, these three pollsters are all getting relatively the same answers from Democrats, Republicans and Independents regarding the election. The variation in the results -- which are quite substantial -- stem from the mix of voters used.
It is important to remember that pollsters usually adjust their results to fit the profile of the expected electorate. That being said, it is also worth pointing out that we have not seen an electorate with components like the one ABC used at any time in the last 20 years. The ABC poll is clearly out of synch with reality. The real question is which of CNN and IBD have the correct mix being used. There's no way ahead of time to tell; we may see a surge of Republican Trump voters like in the primaries and a fall off among Democrats who really don't like Hillary all that much. Such turnout figures would bode quite well for the GOP. Alternatively, we could see GOP turnout sag as some Republicans just don't want to vote for Trump. Only time will tell. The best thing that one can say is this: according to Nate Silver of 538 blog who is supposed to be a polling expert, the IBD poll has been the most accurate in predicting the results for the last three presidential elections. That is a great track record. If IBD is right again this year with its turnout model, then the election is now essentially tied. If the actual percentage turnout is more like the CNN model, then Hillary is ahead by 4 or 5 percent. It is not the landslide that the mainstream media keeps discussing unless the turnout changes to something completely different than anything we've seen in recent American elections.
To try to make some sense of this, I looked at the internals of three polls: the ABC tracking poll, the IBD poll and the CNN poll that was released this afternoon which shows Hillary up by 5%. The most surprising finding of that review is that the results of the polls are nearly totally explained by the mix of the people polled. ABC polled 36% Democrats and only 27% Republicans. CNN polled 31% Democrats and 26% Republicans. (The remainder were independents.) If you adjust the ABC poll so that the percentages of Republicans, Democrats and independents are the same as that used by CNN, the outcome puts the results within 1% of the CNN poll results. If you adjust the IBD poll percentages of those polled to the ones used by CNN, the results again come within 1% of the actual CNN results. In other words, these three pollsters are all getting relatively the same answers from Democrats, Republicans and Independents regarding the election. The variation in the results -- which are quite substantial -- stem from the mix of voters used.
It is important to remember that pollsters usually adjust their results to fit the profile of the expected electorate. That being said, it is also worth pointing out that we have not seen an electorate with components like the one ABC used at any time in the last 20 years. The ABC poll is clearly out of synch with reality. The real question is which of CNN and IBD have the correct mix being used. There's no way ahead of time to tell; we may see a surge of Republican Trump voters like in the primaries and a fall off among Democrats who really don't like Hillary all that much. Such turnout figures would bode quite well for the GOP. Alternatively, we could see GOP turnout sag as some Republicans just don't want to vote for Trump. Only time will tell. The best thing that one can say is this: according to Nate Silver of 538 blog who is supposed to be a polling expert, the IBD poll has been the most accurate in predicting the results for the last three presidential elections. That is a great track record. If IBD is right again this year with its turnout model, then the election is now essentially tied. If the actual percentage turnout is more like the CNN model, then Hillary is ahead by 4 or 5 percent. It is not the landslide that the mainstream media keeps discussing unless the turnout changes to something completely different than anything we've seen in recent American elections.
No comments:
Post a Comment