Election Polling… the pitfalls and possibilities

It’s no secret that pre-race polls can be anywhere from completely misleading to spot-on accurate. Hence, it makes it difficult to know what current pre-race polls indicate. One thing is certain, Obama is favored to win this one. However, if recent elections have taught us anything, it’s that we should be cautious in believing premature election announcements. Even with a solid lead in the polls, there are vocal folks out there warning that an Obama victory is certainly not a foregone conclusion. Obama included. While the legacy of hanging chads in 2000 may be the freshest in our minds, incorrect assumptions of final results are hardly a recent phenomenon - witness President-elect Truman holding a newspaper incorrectly announcing his defeat to presidential candidate Dewey in the picture above…

One thing unique to this presidential election is the concern that the ‘Bradley Effect’ (also known as the ‘Wildner Effect’) will rear its ugly head. In 1982, Tom Bradley, an African-American and long-time mayor of LA, ran for governor of California. Opposing him was white Republican George Deukmejian. Polls leading up to the election pointed towards a substantial lead for Bradley. Exit polls confirmed this once votes were cast, and several media outlets announced Bradley’s projected victory as something of a foregone conclusion. Despite winning the majority of votes cast on election day, the final count was not in Bradley’s favor. Absentee ballot counts swung the election in favor of his opponent. Post-election research revealed that a smaller percentage of white voters had voted for Bradley than was indicated by polling. Additionally, a statistically anomalous percentage of ‘undecided’ white voters also opted for Bradley’s opponent. Other political race outcomes have also seen what many to believe to be hallmarks of the Bradley Effect.

The good news is that the Bradley Effect is by no means a given. There have been plenty of political races that showed no signs of such. Further, some analysts surmise that the current presidential election will see a ‘Reverse Bradley Effect.’ Rationale for this is listed as under-polling of African-Americans and young voters and reluctance on the part of some Republicans to openly say that they will vote for a black Democrat. If the Democratic primary can be viewed as a indicator, it’s worth noting that Obama saw 3% points more support in the primaries than he’d been estimated to have by polling. Again, the most common explanation given to this particular statistic is that no polling is done to people’s cell phones (only to home landlines). I, for one, as well as countless other young voters don’t have a phone line other than my cell phone. Hence, there’s no chance I’ll be polled (and there was no chance I would have been polled for the primary).

In 2004, exit polls were conducted by Edison/Mitofsky at ~1500 locations and via ~500 phone interviews in 13 states with high proportions of absentee and early voters. In the 1460 exit poll precincts where the group collected both exit poll data and actual final vote returns, the exit poll overstated the actual difference between the Democratic presidential nominee (Kerry) and the Republican one (Bush) by 6.5 points in Kerry’s favor. This discrepancy was later attributed to the fact that voters for Kerry participated in the exit polls at a higher rate (voters who are solicited for an exit poll do have the option to decline to participate). Safeguards are in place to ensure that some of the problems of the past aren’t revisited on Election Day 2008, but I think the lesson is that pre-election polling and event exit-polling are not 100% reliable. Be wary of results that you hear. Every news network wants to call the race as soon as possible, but it’s obvious that the data they have at their disposal to do just that isn’t without problems. Networks announced (prematurely) that Kerry would win in 2004, and we all remember the Florida debacle of 2000 where exit pollsters awarded Florida to Gore before the polls closed. Later, the networks (not the pollsters) retracted that call and awarded it to Bush, and that call was also retracted in short order.

Here’s a sampling of polling results from the 2004 election - when things are as close as they were in that election, it’s easy for a couple points of error to skew the end result to indicate the wrong winner.

As with the Presidential election, The Baby Formula contest is a close one. Daily voting has begun, so make sure to cast your vote to help decide who takes home part of $1500!

Get a Trackback link

1 Comments

1. SypePypeInema, November 18, 2008:

   All that glitters is not gold ;)