There will be no shortage of polls as the presidential primary season rages on in 2016. They will report various results and some will be more accurate than others due to their methodology. Sorry to say one of the lamest polls we’ve seen is one conducted by a cable operation based right here in San Diego: One America News Network. Too bad, the home town players don’t get a pass.
It purports to show GOP candidate and apparent front runner Donald Trump won the last presidential debate of 2015. Don’t believe the numbers.
First, it’s an automated poll. This means no one was reached on a cellphone, because only real human beings are permitted to call cellphone numbers. So think about it: the only people who participated answered on a landline phone. According to the National Center for Health Statistics, 41 percent of U.S. households were wireless-only by the end of 2013, including nearly two-thirds of all adults 35 and younger.
Even more important, there was no pre-debate post-debate methodology to actually get at who won. All the poll did was ask who the person it reached was supporting. Was it the same person he or she supported before the debate? The poll doesn’t reflect anything beyond the fact there are currently more Trump supporters out there, not that Trump did something in the debate that caused him to “win.”
Let’s contrast this with a recent Greenberg (Democratic) poll which does account for cellphone users. This poll shows Hillary Clinton beating Donald Trump by eight percentage points. The breakout vividly demonstrates why it is so important to properly account for cellphones. Look at the difference in the results.
It’s important your survey sample includes calling people who only use cellphones. But you have to do more than just call cellphones. Getting it right is more about making sure cellphone users are weighted in your polling sample according to their proportion in the actual electorate. In other words, what proportion of those who will actually cast a ballot are cellphone-only users? We know more people use cellphones in younger age groups, but they also vote far less often than older adults who are still mainly using landline phones. It’s difficult to accurately account for all these things.
Greenberg’s poll probably overstates the cellphone group, but GOP polls typically understate them.
A lot of the variance between polls comes down to what different pollsters do (or don’t do) to include only those people who will actually vote AND not exclude people who will vote (like that 75 year old still without a cellphone). This is tricky stuff for sketchy research firms, those stuck in their methodological ways, or those who cut corners on cell phone data collection.