People shouldn't just dismiss the results of either this OR the polls showing Mandel leading. They simply demonstrate how differing electoral assumptions combined with some questionable polling methodology practices can lead to vastly different polling outcomes.
The first and most obvious difference is that the Luntz polling group is definitely a few points to the left of the Ohio - PPD poll, with the Luntz poll ignoring race and the PPD poll ignoring age as significant demographic indicators (even amateur pollsters know to include these as co-variates...), and Luntz having somewhat D-friendly age brackets (5% total more 18 - 29 year olds compared to 65+ is probably a Democratic pipe dream, particularly in Ohio). That being said, the two polled groups themselves are no where near 20 points apart from one another on a federal level.
The real difference between the polls lies in methodology. Notice how for every single Luntz survey question, the option listed first in the poll receives a massive boost. Is it really possible / remotely likely that Cordray would lose by 25+, Kasich would have +30 approvals, and Sherrod Brown would win by 20 points simultaneously? Of course not. The problem is that when you introduce large lists of options, people's attention spans dwindle and they tend to stop reading and just pick the best option they can instantly see. The NYT ran a good article years ago detailing how
being at the top of a 2-person ballot offers a 2-point margin lead on its' own, and the effect gets amplified for longer lists. Of course, we don't know how the other poll asked questions, so credit to the Luntz poll for releasing that information - maybe both surveys have opposite list biases! All that we definitively know is that Luntz has a bias towards the top line in every question, however.
The reality is that there was a significant shift against the item lines listed at the end across almost all questions, including Trump / Kasich's unfavorables, unfavorables for bipartisan redistricting commissions, Ohio legislature unfavorables, Cordray's numbers across the board against all candidates, and Mandel's numbers against Brown. From the looks of things and doing a significant amount of hand-waving, I would estimate that this bias probably amounts to roughly a 12 - 15 point swing across all questions, though of course it is very possible this survey simply polled more pro-Kasich moderate Republicans than other polls by a significant amount.
Don't dismiss polls even when the top lines seem ridiculous, because they contain valuable information regardless! Top lines rarely tell the most interesting stories - the assumptions they make that lead to that top line is the real question.
