Time Finds New Ways to Distort OH Poll for Obama
It really must be panic time for the media. Time magazine is out with a new poll of Ohio, showing Obama with a 5-point lead over Romney, 49-44. That's about where polling showed the race a couple weeks ago, with the consensus now that the race is much closer. Still, with Obama still stuck under 50%, the poll isn't exactly horrible news for the Romney campaign. But, if you dig deeper, you start to see some problems. First off, the poll shows Romney winning Independents by 15 points.
Wait, what? Romney is crushing Obama among independents 53-38, yet is behind by 5 points? Regular readers will immediately be able to pinpoint the problem. Yes, Time dramatically oversamples Democrats, crafting together a poll that is D+9. This implies a bigger Democrat turnout than the wave years of 2006 and 2008. Your mileage may vary a bit, but that isn't going to happen.
But the Time poll goes one-step further than just the regular partisan oversampling. Basically, they split up the poll into two mini-samples. They separate those who plan to vote on election day and those who have already cast an early ballot.
Among those planning to vote on election day, the Presidential race is tied, with Obama and Romney each taking 45% of the vote. Amazingly, Romney enjoys a 23 point lead over Obama among Independents voting on election day.
Time then asks those who say they have already voted and finds Obama with an enormous lead of 60-30 among those saying they've already voted. Keep in mind, though, the sample size of this survey is just 145 people. To survey the whole state. This alone would carry a double-digit margin of error on the question.
Worse, the sample of those who say they have already voted is D+14. While this largely matches the early voting advantage Obama had in 2008, we already know their advantage isn't going to be that large this year. This year, Democrat requests for absentee ballots are running only 6 points ahead of Republican requests. This poll, though, assumes the Democrats will have more than twice the advantage than they actually have.
I should note that OH also has in-person early voting, but the overwhelming majority of early voting is done through absentee ballots.
Further, even the ridiculously small sample size for early voters is probably too large. The sample implies that about 20% of voters have already cast a ballot. But, only about 14% of voters have even requested an absentee ballot at this point. There is no way that 20% of voters have already cast a ballot if far less than that number have even asked for one.
But, there is even another problem with the handling of early voters in this poll. After arriving at results among those who have voted and those who will vote on election day, the poll simply merges these results together to produce the top-line Obama +5 headline.
This is tricky to do, because you are effectively smashing up two different polls with completely different margins of error. You can't produce anything like an accurate picture of the race in Ohio by doing this.
There is one more weird thing in this poll I want to note. 82% of voters between 18-39 report they are extremely likely to vote this year. 67% of seniors 65 and over say they are extremely likely to vote. There is no way young voters have a 15 point enthusiasm edge over seniors. It is a pretty good indication that there is something screwy with the sample.
I have long noted that this year the media learned how to weaponize polls to impact the election. But, the campaign has reached the point that even juicing the polls by oversampling Democrats is no longer enough to report that Obama has the lead. They are going to have to get more creative. Time certainly did with this junk poll.
Follow me on twitter