Monday, November 7, 2016

Why You Shouldn’t Trust the Presidential Polls

Why You Shouldn’t Trust the Presidential PollsThe 2016 election will forever be remembered as one of the most bizarre in U.S. history. We’ve seen “October surprises,” sex scandals, outrageous accusations, and unmitigated lying before, but nothing has ever come close to what voters have witnessed in this year’s race.

The latest issue to emerge is the FBI’s shocking Oct. 28 announcement to reopen an investigation into Hillary Clinton’s use of an unsanctioned private e-mail server after finding a trove of e-mails on a computer belonging to former Rep. Anthony Weiner, the estranged husband of top Clinton staffer Huma Abedin. Weiner is under investigation for sending sexually explicit messages to a 15-year-old girl.

The news of the reinitiated investigation shocked the political world, and polls are now showing a significant tightening with less than a week left before Election Day. An ABC News/Washington Post Tracking Poll released on Nov. 1 now shows Donald Trump ahead of Mrs. Clinton by one percentage point. Trump had been down by more than 10 percentage points in the same poll just one week ago. Rasmussen Reports also shows the two candidates in a dead heat, with each earning 44 percent of the vote.

Although most recent polls now show the race essentially tied, there are numerous reasons to believe the polls are not to be trusted.

First, scientific polls rely on assumptions to accurately predict future outcomes, which means polls are only as accurate as the models that serve as their foundation. A good example of this is the male-to-female ratio of voters. Female voters are far more likely to vote for Democrats than male voters. If a poll predicts a much higher or lower ratio of women voters than what occurs on Election Day, then the poll will likely be skewed and thus inaccurate. The same is true with other demographics, such as the percentage of minority voters, Millennials voters, voter turnout in general, and other similar factors.

Because polls are designed to project the behavior of more than 120 million people, an incorrect voter model in a survey of several hundred people can easily throw election results off by many percentage points. If that’s the case, you might be wondering, why have polls historically been relatively accurate?

Scientific polls have been able to project accurate results by utilizing historic data about voter groups from previous elections. Voter demographics often correlate with very predictable trends based on a variety of factors, such as the state of the economy, changing demographics of the nation, generational characteristics, information about the current administration, and other considerations. Pollsters also rely on the information they collect while polling. If their surveys show high levels of interest amongst certain voter groups, it’s often reasonable to predict those groups will come out to vote in larger numbers than usual.

The problem with this year’s election, however, is many issues prevent pollsters from accurately building their models. Perhaps the most important of these is the favorability of Mrs. Clinton and Mr. Trump. For the first time in modern history, both the Democratic and Republican candidates are incredibly unlikable. Some 60 percent of people say they dislike Mr. Trump, and after the FBI made its announcement, about the same number say they dislike Mrs. Clinton. If a majority of voters dislike both candidates, which is unprecedented for a presidential election, how can pollsters be sure voter turnout will mimic previous years?

Second, both candidates have significant segments of their own parties that dislike them, which means neither base is motivated to vote in favor of its own party’s candidate. Rather, the motivation to vote is mostly based on a desire to stop the opposing party. Because this is also unprecedented in modern presidential elections, it’s hard to predict how the normally reliable voting bases will behave.

Third, many models are relying on voter data from the 2008 and 2012 elections, both of which involved an incredibly unique candidate, Barack Obama, the first African-American president. Will minority voters come out in similar numbers to support two older, white candidates? Will Millennials, who seem to be the group least impressed by both Clinton and Trump, come out to vote in similar numbers as they did in 2008 and 2012? My guess is they won’t, but many voter models assume the opposite to be true.

Fourth, I believe there is a significant amount of voter shame in the 2016 election. Because of the scandals linked to Clinton and Trump, neither is a candidate most voters are proud to support, which means they may be less likely to tell pollsters the truth about who they are truly voting for.

The truth is no one has any clue what’s going to happen on Election Day because it’s virtually impossible to build accurate polling models given the problems previously mentioned. Either candidate could end up winning in a landslide or it could be an incredibly close contest, making it even more important for voters to come out in support of the candidate they support most, regardless of what pollsters tell them will happen.

No comments:

Post a Comment