Wednesday, February 3, 2016

“Poll-ease”… Why were the Iowa Caucus predictions so off?

February 1st marked the official start of the 2016 presidential election voting season, as Iowan voters determined the number of delegates each candidate would receive. The Iowa Caucus was met with much enthusiasm, as the Republican party recorded approximately 180,000 caucusgoers, which beats their previous record by around 60,000 individuals. 

The Iowa Caucus has become a hallmark event, and some news anchors went as far as to dub January 31st as “Iowa Caucus Eve.” Prior to the actual caucus, many news media sources published predictions regarding the outcome of the caucus. Many sources like the Huffington Post predicted that Donald Trump would win with a fairly solid margin of 5-15% depending on the source.

It’s fair to say that taking a well-designed and highly accurate poll is difficult given the constraints of reality and the variability of human behavior; however, the results of the Iowa Caucus, which put Ted Cruz on the top of the other GOP candidates and have Clinton and Sanders in an approximate tie, present a couple of questions, which fall under the overarching question “Why were the Iowa Caucus predictions so off?”

Carol Bialik from FiveThirtyEight published an article about some of the lessons from this times polls, one of which is to poll to the end given that voter turn out is everything. Do you agree with this? What other factors could have contributed to the technically inaccurate predictions? Is the Iowa Caucus just to darn hard to accurately poll and predict?  

Moving away from the Iowa Caucus, do you think that the news media and the free media should still report the predictions prior to the caucus, even if the predictions may not be completely accurate? What effect do you think these predictions have on voters and voters of states with impending primaries or caucuses? What standards should the news media be held to when reporting statistics (Should they have to report polling methods? The margin of error? The sample size? The source?) Finally, how do you think the use of polls has evolved with the free media?



Carolyn Ku said...

It is impossible to make polls absolutely correct, and it does not surprise me that the polling before the Iowa Caucus proved to be inaccurate. It is possible that the polls were so inaccurate because they were weighting the responses based on previous turnouts, and given that this year's turnout was so much higher than previous years, it is possible that this resulted in a different proportion of voters from different demographics. For example, it is probable that more Evangelicals showed up to vote due to Cruz's intensive groundwork courting them. It is also possible that Donald Trump supporters were over represented in the polls because far-wing voters are more likely to respond to pollsters in order to bolster their candidate. Although the Republican polling was inaccurate, it seems like the Democratic polling pretty accurately predicted a near tie between Clinton and Sanders. This suggests that pollsters were better at weighting the response of young voters, who were largely in support of Sanders, and the lack of support for O'Malley.

Scott Chow said...

I honestly don't believe, despite the poll's inaccuracies, that they do much to influence voter's decisions on WHO to vote for. Last time I checked the selective hearing bias is alive and well among the population of America, myself included: no mom, I don't remember you telling me to do the dishes. That being said, I believe the harmful effects of polling in general come by reporting who is NOT going to win. In reporting in on the likely losers of an election, polls further discourage potential voters coming out to vote, given that their candidate is unlikely to win. That being said, their is a strong counter to that point, which is the idea of perfect (or at least better) voter information. If voters are able to know ahead of time that their extremist candidate is unlikely to win, then they're better able to represent their self interests by voting with one of the winning candidates that they agree with most. Both the ideas of increased voter turnout and a better informed public seem to be valuable to the democratic system as a whole but as it stands at the moment, I don't believe the current media coverage of elections in general will allow for the American public to enjoy both. And given that voter turnout seems like the primary issue (how well informed the electorate is doesn't matter if they don't vote) I would argue to curb the practice of polling before elections.