Okay, so a lot of people seem to be implying that the polls favoring Clinton are rigged somehow and that Trump is actually winning. Look, I've seen this argument in the past and it's often wrong. In 2012 I saw republicans claim mainstream polls were rigged and that Romney was winning. That was bogus obviously. I also saw people saying polling doesn't tell you everything in the democratic primary and that Bernie would actually win. Now, in the latter case, we saw some truth in this because polls underestimated independent and young voters and also there was a huge push by Bernie in the days before the primary that made the results skew more in his favor than they should be, but this is not normal or typical, and should not be treated as such.
What do polls try to do?
Polls are there to try to accurately reflect what the results of the election will be. It's a science to try to get them just right. We have had cases in the past (the 1936 election comes to mind) where polling has been inaccurate for various reasons. However, social scientists are smart and educated on the matter these days and are trying to ensure that the results are accurate. This means they choose sampling methods that get to the truth, try to ask questions in nonbiased ways, etc. There is little incentive to lie in these polls, because while you can cheer on your candidate from an echo chamber for some time, eventually reality will hit you in the face and you'll be proven to be wrong on election day. This is what happened to Romney's campaign in 2012. They ignored the polls because "liberal bias" and then were shellshocked when they lost.
A few common claims of rigged polls and my attempts to explain them in context
--"The polls are rigged because they sample more democrats than republicans!"
You don't want to necessarily sample an equal number of democrats and republicans. You want a sample that represents the general voting population. If there are, say, 60% democrats and 40% republicans, you don't want to sample both groups 50/50. That will skew the results. You want a random sample that represents the population. If there are more democrats than republicans, that's fine, as long as it is representative of the population.
--"The polls are rigged because they removed the neither option!"
I think I saw this accusation with a Reuters poll recently. Apparently they had three options: Trump, Clinton, and "Neither/other." They removed "neither" and just called it "other" because they felt that having a "neither" option was skewing the results. Essentially, people who are pissed off at both candidates, but who would support those candidates anyway, were saying they supported "neither", when if they went to the polls, they would choose one or the other. So they tweaked the polls to say, Clinton, Trump, or "other" (aka, third parties, Johnson, Stein, etc.). This apparently changed the results to Clinton's favor and some people are freaking out over it. However, from what I understand they didn't skew the poll intentionally to make it favor Clinton, they're trying to UNSKEW it in order to make it more representative of how the population WILL ACTUALLY VOTE.
--"They're not including Millennials in their sample."
This one seems a bit more troublesome, but then I remembered they did something similar in the primaries. The reason they're oversampling older folks and undersampling the youth vote is because older voters are more likely to consistently vote and are thus their opinions are more representative of how the election will actually play out. On the other hand, Millennials are seen as less reliable voters and sampling them as they would older people would supposedly skew the results. Again, you want to make sure the polls represent what will actually happen come election day. As such, you sometimes need to shake up your methodology to ensure that you get rid of inherent sampling biases.
I do think that undersampling the youth vote may be a mistake though. I think that's a huge reason Sanders did much better than expected. When the youth vote turned out heavily in Sanders' favor, the end results would often play out much differently than the polls would suggest. So this attempt to clean up the data may backfire. Maybe not though. I don't see many Millennials supporting Trump and while I would suspect they would lean Clinton they're not particularly enthusiastic about her like they were toward Sanders, so who knows what will happen.
Conclusion
I hope this puts some fears of rigged polls to rest. While the media may love to focus on polls that fit their narrative in order to make candidates look better or worse than they actually are, there's not really a lot of incentive to actually rig the polls. People who do this do so at their own peril, because history tends to vindicate the people who try to make their methodology as active as possible. Sure, you can make some phony polls showing Clinton or Trump way ahead, but come election day, will they really matter? Probably not, and those who skew polls will look like fools for doing so.
Pollsters base their methodology around what gets accurate results. Their primary goal is not necessarily to push a political narrative, it's to try to estimate what is really happening in the real world. Samples are chosen not to fit narratives, but to be representative of the voting population as accurately as possible. Same with the polling questions. They're supposed to be framed in ways that don't skew the results, and most changes that occur are there to ensure more accurate results, not more biased results toward specific candidates.
No comments:
Post a Comment