Tuesday, October 11, 2016

NBC/WSJ poll conducted after the Trump tape leak shows a mammoth Clinton gain

NBC/WSJ Poll conducted after the Trump tape leak shows a mammoth Clinton edge

One cannot place too much significance in a single poll, but this is the only scientific poll conducted so far in which the voters were ALL surveyed after the leak of the Billy Bush tape, and it indicates a potential landslide. But there's a huge fly in that ointment ...

Clinton has an 11-point lead among likely voters in a four-way race, and a whopping 14-point lead head-to-head.

The fly in the ointment is a sampling problem. Among those in the sample, the people who voted for Obama outnumbered those who voted for Romney 46-33, which means the group leans much farther left than a truly useful random sampling. Obama actually won the popular vote by only four points. That's not to say the pollsters cheated. While the odds are against finding 500 properly randomized voters who skew 46-33, long odds are not the same as absolute certainty. If a series of events is 52% probable (Obama's actual edge over Romney: 51-47), the probability of finding 58% (Obama over Romney in the sample 46-33) or more of those events occurring in a random sample of 500 is only 4/10 of one percent. This was probably that one time in 250. Those things do happen. The point is, however, that the freakish sampling skew probably makes this result indicative of absolute zilch.

So what is the real effect of the Trump tape? It's too early to say, but the 538 Blog takes a measured view.

Note this reader response to my comment, which I have brought up to the main page because it is perceptive. I assume he is right, because while I am proficient at statistics, I am not an expert in polling.

"The analysis of the 46-33% is wrong here. Pollsters have long found that many people forget who they voted for and overstate, frequently by a fairly large margin, that they voted for the winner, especially when the winner is popular (or in the case of Obama, reasonably popular.)

It's not quite the same effect, but it's largely based on the same thing that at one time at least 20 million Americans claimed to be among the 500,000 that attended Woodstock in 1969."

1 comment:

  1. The analysis of the 46-33% is wrong here. Pollsters have long found that many people forget who they voted for and overstate, frequently by a fairly large margin, that they voted for the winner, especially when the winner is popular (or in the case of Obama, reasonably popular.)

    It's not quite the same effect, but it's largely based on the same thing that at one time at least 20 million Americans claimed to be among the 500,000 that attended Woodstock in 1969.

    ReplyDelete