Throughout the 2016 campaign, political analysts and pollsters were all but certain that Hillary Clinton would become the next POTUS, however, they were almost all wrong. The polls and predictions were off almost across the board.
Who got it right?!?!?!?!
According to Real Clear Politics, only two pollsters the day before the election had a Trump victory, that’s right only TWO out of 21 polls had just the winner correctly called. One of the polls that had a Trump victory was a tracking poll from ISB/TIPP. Looking further into their data, they thrived where the other polls has faltered. They correctly predicted a resounding Trump victory in the mid west and southern regions, while also showing that Hillary Clinton underperformed with black/hispanic voters. These accuracies are what allowed ISB/TIPP to correctly predict a Trump victory. Another poll that had a Trump victory was the USC Dornsife/ LA Times. They actually had an overshot prediction of Trump winning by 5%. There are two reasons as to why their poll had predicted a Trump victory. The first reason is that they had accurately shown that uneducated voters who generally break democrat went to Trump. The second reason is the same as why ISB/TIPP was accurate, it showed that Hispanics weren't flocking to the democrats as they had in the past.
What went wrong?!?!?!
The big question remaining, is why were so many pollsters off? The answer lies in two major mistakes. The first mistake is that there was an error in weighting methods. The pollsters miss weighted the Hispanic vote and the uneducated vote. These are two groups that aren't very likely to respond to polls, so there is a big opportunity to weight them and the majority of pollsters mis-weighted them. The second mistake is more of an error on the respondents part. It is an error that has to do with social desirability. Social desirability is the desire to be seen as a certain way by society so many people different feel comfortable responding as Trump supporters so they either lied or didn't respond. This same phenomena can be seen with black candidates. Respondents are more likely to respond in a poll in favor of a black candidate so as to been seen as a socially conscience individual. This is called the “Wilder” effect, Hillary Clinton experienced this effect. That is a large reason why I believe much of the polling data was off.
What They Could Have Done Different
I don’t think that the polls could have done much about the “Wilder” effect. Maybe ask questions less directly so respondents can’t decide which candidate which views most directly supports. This may be unethical, but i don’t see another way to get around the “Wilder” effect. The incorrect weighting is an issue that I believe the pollsters would have to look beyond the data to who Trump is appealing to. It seems almost a little too obvious to me that uneducated white voters would be all aboard the Trump train. As for the miscalculation of the Hispanic vote, I think that could have been predicted as well if the pollsters looked outside of the numbers. Trump was staunchly against illegal immigration which is something I know from my experience with family friends who went the legal process are really against as well. Looking outside the numbers isn't something that pollsters do very much but I think that it could have really helped them in the 2016 election.
Importance and Limitations of Polls
Despite the occasional “inaccuracies” of polls, they provide a way to look into the mind of the average voter and figure out what strategies would work the best in a fairly cost/time efficient way. With that being said, they have some major limitations. Firstly, they don’t give a definite number, instead they operate on a 95% confidence interval. This means that the pollster is 95% confident that the true number is between a range of numbers, so the pollster isn't even 95% confident in the number that you see on TV or the internet. Secondly, pollsters have to rely
on a sample of the population and then play a guessing game trying to predict the rest of the population. They are pretty good but it isn't a 100% yet. Thirdly, they can’t take undecided voters into account. If polls were 100% accurate and predictive then there would be no need for the election, but unfortunately we are not there yet.
Sources:
http://www.realclearpolitics.com/epolls/2012/president/us/general_election_romney_vs_obama-1171.html
http://cesrusc.org/election/
http://www.investors.com/politics/ibd-tipp-presidential-election-poll/
http://www.realclearpolitics.com/epolls/latest_polls/president/
You bring up some good points as to why many polls were off. What’s interesting is that the polls most likely did neglect to factor in certain demographics. Even though a lot of people, no matter the demographic, don’t wish to participate in polling, it doesn’t mean that they shouldn’t be factored into at least an unresponsive category/subgroup.
ReplyDeleteFew people thought that after all of Trump’s unprecedented remarks throughout the campaign though that he would get that many votes, especially from the hispanic community. In all fairness, using systematic polling or not, I don’t think anyone would have guessed the outcome of the election from a subjective point of view.
I agree absolutely! I think that the weighting of the polls was what was the biggest error in their methods. I also agree that guessing the demographics that Trump got was a complete toss up.
ReplyDeleteDan I really enjoy the fact that you taught me a new term, the "wilder effect." I also in my blog brought up the same idea that the wilder effect has to offer, however didn't know there was a term to describe it! Really good analysis of the polls Dan, solid job as always!
ReplyDelete