Didn’t See That Coming: Election Polls, Public, Information Cascades
This year’s election produced a wide range of reactions including outrage and disbelief. Why the disbelief? We can look to multiple sources that possibly fed people the idea that Donald Trump would not win. One of these sources is the polls of which the majority predicted Hillary Clinton to prevail. The reference articles in this blog contribute to the conversation of how these polls turned out to be badly mistaken. Some points to consider, as mentioned in the articles, would be the demographic of survey participants, changes in voters’ decisions after submission, and the potentially misleading voter history.
On the surface level, it is obvious that each poll is a representation of its participants and since input is voluntary, these depictions do not perfectly reflect the whole population. Another flaw stems from the fraction of participants that abstain from voting or change their vote after their data is collected. But these arguments can be applied to any election; why would they impact the accuracy so drastically now? Both articles touch upon yet another pitfall of polls, this time focusing on those that take previous elections’ voter turnout as a predictive factor. In other words, they weighed Democratic or Republican likely-voters based on previous election trends. This data is analogous to that of information cascades. The more an occurrence arises, the higher probability of that occurrence dominating. Unfortunately, as many have realized, this election was unprecedented. Looking at past information or even current opinions is not the same as knowing the actual ballots individuals casted and votes electorates made.
Even knowing others’ decisions can lead to false assumptions, as in the case of localized information cascades. Though the people in an individual’s immediate circle can be trusted to speak truthfully about their ballots, the demographic of such a smaller network can prove unrepresentative of the greater United States. If one lives in a majority-liberal state, the individual would likely only hear from those who share the same beliefs and consequently make the same decisions. Then the term “pluralistic ignorance” comes into play. Pluralistic ignorance is when many people believe one thing to be true, but the converse is actually reality. In a purely conceptual information cascade, it would make sense that by knowing similar and reoccurring truthful votes and predictions that the highest probable outcome would be of the same vein. However, due to the bias nature of the common knowledge in one’s network, the results of information cascades can lead to a possibly unpleasant awakening.
http://www.theatlantic.com/politics/archive/2016/11/what-went-wrong-polling-clinton-trump/507188/