With less than two weeks until Election Day, there’s been a barrage of polls attempting to capture the sentiment of the American voter.
While Clinton has had a decisive lead in many, and Donald Trump has called them “phony,” the truth is that on their own, each individual poll doesn’t really tell us much. Each varies in its data, and every day seems to tell a new story.
How, then, are we to understand these polls if they all seem to be saying something different?
Every four years, the race to the White House is narrated by polling data. This powerful information is used by the media to tell a story and the campaigns to inform their strategy. But like any other dataset, its interpretation is contingent on statistical validity.
That’s why aggregated polls, the ones that average the collection of polls released every single day, are the ones to pay attention to.
Take for example, the latest release of polling data. ABC News has Trump at 47% and Clinton at 49%. CNN has Trump at 45% and Clinton at 51%, Politico/Morning Consult has Trump at 43% and Clinton at 46%, and UPI/CVOTER has Trump at 48% and Clinton at 49%. Each poll is from a reputable source, and was conducted over a similar timeframe, and yet all vary in overall percentage and difference between the candidates.
Individual polls vary, and sometimes significantly, for several reasons.
First, polls capture a specific snapshot in time. This is perhaps most apparent with post-debate polls. Sentiment measured on the night of a debate will undoubtedly shift over the course of that week and up until Election Day.
As Pew points out, there are a number of factors limiting statistical significance in this instance. Political pundits, for one, have yet to voice their opinions, which undoubtedly help to sway public opinion.
Also, quick reactions to events are not always representative of how people truly feel about something. Attitudes can shift once someone has had the time to process a situation. These forces influence the findings of a poll, and thus skew the depiction of voter sentiment.
Second, polling sample sizes and sample compositions vary wildly and one single poll rarely represents the country as a whole. One poll should be seen as one data point among many. Even if the sample is the same, once pollsters have raw data they make weighting decisions that influence the outcome of a poll.
How sample weighting impacts polling results can be learned in Nate Cohn’s article of The New York Times Upshot “We Gave Four Good Pollsters The Same Raw Data. They Had Four Different Results.”
For people who are not statisticians and survey specialists, it’s hard to factor this in and comprehend the whole context of a poll. In addition, not all polls are transparent enough to publish their survey design.
Today’s polls are released with rapid speed. Planning is critical to the survey process, with the survey sample ranked among the highest concerns. Polling a truly representative sample on such short notice should raise questions.
Who are these pollsters calling and how do they choose these people? At random, or do they poll the same group of random people over and over, to really measure a trend? While a single poll can certainly still tell us something, on its own, it’s not saying much.
Instead, aggregated polling data presents the most accurate picture of voter sentiment. These polls take the average of the numerous individual polls, and some even reference outside factors such as economic performance and consumer sentiment. It is the mix of all this data that gauges the sentiment of the electorate.
While sources like Nate Silver from FiveThirtyEight and RealClearPolitics are promoting statistically-sound data by publishing aggregated polls, the majority of news sources continue to report on individual poll movements.
Voters simply cannot rely on one news report to present a statistically accurate picture and to put disparate polling data into context.
To anyone other than a data scientist, this inundation of daily polls can be quite confusing. As the countdown to Election Day ticks away, take the daily movements of the polls with a grain of salt. It is the cumulative that has the power to cut through the clutter and spotlight what is really going on.