On Election Night 2012, there was a scene that many of us watching will never forget. Worried that he might have just lost millions of dollars on investing in the presidential race, Republican strategist Karl Rove ran around the Fox News studios like a chicken with his head cut off claiming that there was no way his man Mitt Romney had lost the election. Megyn Kelly (you remember, the “Santa is White” gal) even told Rove that the game was up.
But almost every Republican thought that they were going to win. Why?
Of course, the polls were skewed! Democrats were being over-sampled in every survey conduct! There is no way that Romney could be losing because those who have been conducting polls for decades were doing it wrong. This led to the comical Dean Chambers “Unskewed polls” website which claimed that Romney was going to win in an electoral landslide.
To counter this so-called “Democratic advantage” in the polls, Republicans started a new tactic where they would try to figure out what the voter turnout would be, and then only sample within those margins. Basically, they were predicting races, not polling races. As Doug Schwartz of Quinnipiac Polling said last year of the “skewed poll” controversy, “If a pollster weights by party ID, they are substituting their own judgment as to what the electorate is going to look like. It’s not scientific”.
In the last few days, we have seen two polls released regarding the CD-13 race between Democrat Alex Sink and Republican David Jolly. Both of these polls are showing a tighter than expected race. But guess what? Yep, they are using the “unskewed poll” method. Both pollsters state that they are predetermining what voter turnout will look like. In the case of the latest McLaughlin poll, it states that “interview selection was random within predetermined election units.”
So what does this exactly mean? Well, this is how it works: The polling firm first determines what they feel the turnout will be by party ID on Election Day. Then, they sample people they need from each political affiliation. Let’s say, for example, that the pollster predicts that Democrats will only make up 35% of the Election Day turnout. But when they do the polling, 45% of respondents say that they are Democrats. There are two ways that they can deal with this. First, they have their predetermined margin of error, which determines how many voters they need to call, and then only call that certain amount. Any residual Democrats will not be included into the survey. The second way is that they can systematically remove respondents from the survey. Either way, both are unscientific and not honest.
So why is this bad? Survey research is supposed to be done by conducting surveys at random. This means that I have just as much of a chance of being called for a survey as you do. This is the reason why online polling, such as Zogby Interactive, has been considered not reliable, because people are not chosen at random but are willing and knowing participants. So, when a pollster determines how many of each party is going to be polled, the “random” part is thrown out of the window.
But what about “weighting”? We hear this term thrown around all over the place. This process is usually used when there is an underrepresented group in a sample. For example, if Hispanics are 10% of the population and only 5% of the survey, then the pollster conducts weighting in order to truly reflect Hispanics in the district.
But why isn’t this done with party affiliation? Well The Atlantic said it best, so I am going to put their quote here…
Unlike race, gender, or age, all demographic traits for which pollsters weight their samples, party identification is considered an attitude that pollsters say they should be measuring. When party identification numbers change, it’s an indication of deeper political change that a poll can spot.
Now we know the story behind the two recent polls, one from St. Pete Polls and one from McLaughlin. The method they used is the same method that the “unskewed polls” people used in the 2012 Presidential election which predicted a Romney landslide. Who knows, on the day of this special election, maybe Peter Schorsch will be the new Karl Rove.