The Danger of Following Polls
Monday, May 18, 2015on
This month’s British parliamentary elections were another nail in the coffin of political polling. Nate Silver fared terribly, after making his name by using amazing statistical analysis to accurately predict the 2008 and 2010 elections. Silver predicted the Conservative Party would grab 272 seats, but when the dust settled, Conservatives wound up with 329. This caused Silver to take to his widely-read blog to write The World May Have A Polling Problem: “it’s become harder to find an election in which the polls did all that well.” I’m not jumping on Silver. He has a difficult job that is getting tougher every year. I wrote a post on this blog back in October warning people that polls were becoming increasingly unreliable. In the decade that I ran political campaigns across this state, polling accurately told the story until 2008. After that, the accuracy fell off dramatically. One 2012 State House race I worked on has polling showing the race was within the margin of error only five days before. On election night, it was a 13-point landslide. This is important today not because of the British elections, but because the 2016 Presidential horse race has already begun. Every week, there’s a new poll about not only who is leading the Republican race, but how those people will fare against Hillary Clinton. They’re mostly bogus. Don’t put any stock in them. Definitely don’t pick a candidate because of them. Most of the candidates are complete unknowns at this point to most people. Perhaps they know who Bobby Jindal is, but most voters couldn’t cite one of his positions. They only know Jeb Bush because he’s the brother and son of another President. Scott Walker was a guy who faced down public unions in some Midwestern state. The same goes with most of the other 13 presumptive GOP candidates. Silver notes that Presidential Primary polls are notoriously awful. Since 2000, they have missed the correct prediction by 7.7 percentage points. Primaries are full of swing voters who change their minds often. They’re also generally small groups, especially compared to a state-wide election or national general election. (In South Carolina 603,000 people voted in the 2012 GOP Presidential Primary and nearly 2 million people voted in the general election that fall.) The General Elections are easier since you can use statistical sampling to determine that X number will vote Democratic, X number will vote GOP, and you only have to predict the small number in the middle. As I wrote last October, polling is a reflection of a moment in a churning sea. “People lie to pollsters. Sampling a poll is increasingly difficult. Cell phones, the Do Not Call List, and other generational changes mean getting valid phone numbers is hard. They lie about things that are not socially acceptable (“Do you plan to vote this November?”) and about things that could be embarrassing. There is also a distrust of pollsters by some conservatives, who believe it’s the liberal media doing the poll, and older-generation minorities in the South, who vividly remember a day when you just didn’t give someone your opinion over the phone.” The polls are flawed, but the polling right now doesn’t matter anyway. At this point in 2007, Senator Hillary Clinton had a 22-point lead on a certain Senator from Illinois. She built a 30-plus point lead in the fall. New York Governor Rudy Giuliani had a 14-point lead on Senator John McCain. In October, Giuliani was tied with Arkansas Governor Mike Huckabee and McCain was tied for fourth with former Senator Fred Thompson. Just a few months later, McCain would cruise to victory. Of course, that doesn’t mean candidates don’t get excited when they’re in first place, even if it is a virtual 10-way tie. Listen to the candidates. Go to their Upstate events. Ask them questions yourself. Don’t worry about the polls. They’re probably not giving you the entire story.