We are now a week removed from the gubernatorial contest and it’s time to assess the accuracy of the polls leading up to election day.
As I’m sure you’ve heard, over the last few years there have been a lot of questions about polls and whether with the advent of cell phones they are as reliable. Others have concerns about automated polling that relies on a computer and not a live person. All of these are reasonable questions and have provided some issues in other races in other states.
But, it was not the case in Virginia in 2009. According to an analysis by politico.com, SurveyUSA was the most accurate, missing McDonnell’s 58% mark by just one point and perfectly predicting the 18 point spread. Roanoke College’s poll was the second most accurate.
Many people thought the 18 point spread was unrealistic in purple Virginia. I have to admit that I was one of them. Given the totality of all the polls out there, it was obvious that Republican Bob McDonnell was going to win. But, I would have bet the margin to be closer to 10 points. That was just a gut feeling, nothing scientific.
Let’s be clear, polling can be a gamble. It’s up to each individual pollster to try and guess the make-up and size of the electorate. That includes how many Republicans, Democrats, independents, women, and men. They also guess at the racial and age make-up of those they think will vote. So you can see why we sometimes get wildly varying numbers from different polling outfits.
But Virginia’s contest was not one of them. All of the polls predicted McDonnell would win easily and two were almost spot on. Well done.
Posted by Jay Warren at 10:48 PM. Filed under: main •
(0) Comments • Permalink
