No, Polls Do Not Show Trump Beating Hillary
This is not an article excoriating Bernie Sanders supporters. This is an article about strongly encouraging those who use polls to prove their point to actually understand how polls work.
Making the rounds this morning on social media is the screenshot of this tweet:
Well that’s something to make you stop and think, isn’t it? All the talk about how Bernie should drop out because Clinton does better against Trump in the general election, well all that is really at odds with the simple fact shown in this tweet. I know some … passionate … Bernie supporters on social media. This tweet is being plastered all over the place, provided as evidence arguing that it’s Hillary supporters who need to give up the race despite the results of all those pesky elections. They can even provide you the link to the poll results, showing that exact two point lead by Trump over Clinton.
It’d be a decent argument too, if it was true in the slightest. But as Deep Throat once told Mulder, the best way to tell a lie is to hide it between two truths.
You can see every single major national poll taken for the possible match-ups between Clinton or Sanders and Trump at Pollster. Across the board, Clinton and Sanders both beat Trump by about ten points, with the difference between the two candidates being pretty well statistically insignificant. I say pretty well, because I’m not actually running the proper statistical tests on all those polls, I’m just eyeballing it with the intuition of someone who actually knows how to do the math in that phrase.
That two point lead for Trump? It’s on a single poll. Out of dozens that show a seven to fourteen point lead. It’s the only poll that shows a Trump lead. And who is that poll from?
If you heard a noise in the distance, that was just a political scientist sobbing, because one of us loses tenure every time someone takes a Rasmussen poll seriously.
Rasmussen does not use proper randomized sampling. They use robo-callers which are only allowed to call landlines, and so skew a solid ten to twenty points conservative on every issue. In response to those criticisms, which has led to their dismissal as a source of actual polling data by those who understand polling data, they tweaked their approach this year. My graduate student friend Kelsey Starr at GW, summed up the changes this way (and yes, the caps are hers): “that poll compensated for groups that are hard to reach by landline by using online polls that are targeted and WEIGHTED TO POLITICAL PARTY. WHO DOES THAT.”
This isn’t stats geek hairsplitting. Remember in 2012 when right up to the day of the election, every media outlet was showing Obama up by five, but Fox News insisted that the race was a dead heat with Romney ahead by a nose? The polls they kept citing were Rasmussen. These polls exist in order to insist that races are closer than they are. They exist to provide the nominal fictional facts to enforce whatever narrative the right wing needs at a given moment.
So, the cunning among you might be thinking, ah but what does the Rasmussen poll say about Sanders vs. Trump, so we can compare apples to oranges? Nothing, because there isn’t one. They only did Clinton vs. Trump. Because this poll was never intended to find the truth, it was intended to provide a single quotable data point that Clinton loses to Trump, and then have everyone cite the existing pile of polls showing Sanders beating Trump. It’s a lie sandwiched between two truths so that those who already want to believe it swallow it down as juicy gospel pastrami.
When the precise people who benefit from a protracted primary hand you evidence that you should protract that primary, maybe take a second look at that evidence. The devil doesn’t even need to quote scripture when he can just write his own and paste it in.