Polls - Often More Accurate Than Those Reporting

About the polls, they're often more accurate than those reporting on them. Here's a quick excerpt from the Sun Sentinel and my take on it. Can you trust the polls? The answer is more complicated than you might think. 

Excerpt: Political polls strive only to capture a single moment in time. That makes them relatively poor indicators of the future, leading to surprise results like Andrew Gillum winning a Democratic primary when he was polling third or even fourth place. 

In this year’s Democratic primary, most polls showed Gwen Graham leading by several points with Philip Levine in second place. On primary day, Gillum pulled off a huge upset. And let’s not forget the biggest recent shocker: Donald Trump over Hillary Clinton. 

So what do polls tell voters? Should they trust them? And if they are essentially a guess based on what a small number of voters say, how they can be more accurate?  

There are some red flags about a poll’s accuracy: a disproportionate number of undecided voters, unusual ways to contact them and large margins of error. 

Bottom Line: Of all topics in political cycles, especially in the wake of the 2016 election, none is more commonly misunderstood or misexplained than this one. I’ve said this on many occasions since Election Day 2016. Let me start with the 2016 election cycle for a moment. When people suggest “the polls” were wrong in the 2016 Presidential Election they’re wrong. The national polls showed Hillary Clinton winning and she did. For what those polls were sampling, which was national averages. 

An average of national polling showed Hillary Clinton winning by 3 points. She won by just over 2 points on the national popular vote. The problem was when media and politicos took that information to suggest she’d win the election. On that very day I laid out the case as to why she’d win the popular vote and Donald Trump would win the election. The approach I’ve used and always use regardless of the specific election is this. 

  • What type of election cycle is it?  

For this cycle, it’s a midterm election cycle with a first-term Republican President. 

  • Are there any historical trends in the cycle? 

Yes, 92% of the time the President’s party loses seats Congressionally. 

  • If there are any breaks in the trend what were they and why?  

Yes, 3 out of 32 times with major macro catalysts.  

  • How do undecided voters typically break in a given election in a given cycle? 

Slight advantage to the opposition party to the President, often in state races as well.

Then I get down to business using accredited polling of likely voters averaged from sampling derived only from the most recent week and project out undecided voters and 3rd party voters along historical lines. In other words, it’s not easy but good polling is also highly useful if you know how to use it. I’ll give you an example of useless polling from yesterday cited by most news sources. Yesterday Ipsos produced polling showing Nelson with a 5-point lead among likely voters and Gillum with a 6-point lead with likely voters. But the polling was utter crap and useless the moment it came out. Why? The polling was conducted from the 17th through the 25th. Meaning the most recent information is already a week old and the earliest information is over two weeks old. 

There are literally 7 polls on those races with more current data than that poll. Yet, because it was the most recently released the media reported it as if it was the latest. That’s what all of this really gets down to. You have a bunch of people reporting stuff who don’t really know what the heck they’re reporting on in the first place magnified by some with agendas in media compounded by the complexities in the cycles. And when they all blow it, they turn around and blame the polls. 

Photo by: Mark Wallheiser/Getty Images


View Full Site