The Mathematics of Poll Skewing:
The vast differences in polls that are being reported both nationwide and in toss-up states have drawn significant attention from both sides of the political fence. In past elections polls have seldom been accused of being skewed to change people’s votes but this year such allegations have been tossed from both sides.
Today’s Real Clear Politics average uses polls that show a 7% difference in the national presidential election with Rasmussen calling the race a tie and the National Journal having President Obama leading by 7%. It is clearly impossible for both of these polls to be correct and the margin by which they vary makes it clear that their sampling patterns or their voter turnout models are dramatically different.
The term voter turnout models for those new to poll skewing is the assumption of what the turnout of voters will be by party on election day. The purpose of this article is to take a real world example and show how the same data with different assumptions leads to quite different results. All of the data chosen for this article comes from the highly respected polling firm Gallup.
Current Gallup data from the trailing 3 weeks indicates that 93% of Democrats will be voting for President Obama and 4% will be voting for Mitt Romney leaving only 4% undecided. Similarly with Republicans 92% express support for Mitt Romney, 5% support the President, and 3% are undecided. Among those describing themselves as Independent 44% support Romney, 43% are in the President’s camp, leaving 13% who have not yet decided.
Gallup is still using registered voters in the polls they report but they do keep a separate accounting of likely voters and typically change their reporting as the election gets closer. Just prior to the 2012 election the registered voter turnout model for Gallup was 32% Democrats, 31% Republicans, and 33% Independents. When looking at the average for likely voters on the very same day according to Gallup the numbers were 30% Democrats, 37% Republicans, and 30.5% Independents.
If one does the math using the registered voter turnout model for Gallup in 2010, President Obama leads Mitt Romney by a margin of 45.5% to 44.5%. When the average of likely voter turnout model is used the tables are turned and Mitt Romney has a lead of 48.7% to 42.9%. Amazingly the change in the poll just by going from registered voters to likely voters with Gallup’s turnout models is a full 7%.
Of course the undecided are still at 6.2% and the overwhelming majority of those are Independents.
The importance of this analysis is simply that by making small variations in the assumptions used in the voter turnout model pollsters can essentially get the answer they are looking for. The Rasmussen poll always uses likely voters and bases its assumptions on previous turnouts and voter indications. Gallup uses registered voters until Election Day approaches and then shifts to likely voters. Other polls use registered voters which tend to oversample Democrats by roughly 4%. Still others make up their own models seemingly from the thin air with some recognized names oversampling Democrats by as much as 11% which is 4% higher than even the 2008 Presidential election when then candidate Obama turned out more Democrats than any election in history.
The only real conclusions that one can draw right now is that pollsters have agendas and are hiding behind the legacy of legitimacy that their organizations claim to have. With the variance in assumptions of the turnout models and the over 6% that are undecided the current election is not in the bag for either candidate.
Gallup Voter Turnout Model: October 4, 2010
http://www.gallup.com/poll/143363/gop-positioned-among-likely-midterm-voters.aspx
The University of Southern Indiana and the University of Evansville men’s basketball teams will match-up in the Vectren Community Classic October 27 at the Ford Center in Evansville, Indiana. The exhibition game is the fourth meeting between the two programs since 2004. Tip-off time is set for 7:05 p.m. Tickets go on sale to the general public on October 1.
The University of Evansville is proud to announce that it has received a $30,000 grant from the Alcoa Foundation in support of OPTIONS, a summer program that introduces middle and high school students to the vast career possibilities in engineering and computer science.
