In the 2016 U.S. presidential election, 216 million people were of voting age, and 55% of them voted for president (
www.electproject.org/2016g). As we all know, Donald Trump was elected (based on the Electoral College), but he lost the popular vote to Hillary Clinton by 1.9%. Of the major polls, only the LA Times/USC Poll picked Trump to win the popular vote. An unweighted average of the major 2016 polls computed by Real Clear Politics (
www.realclearpolitics.com/epolls/2016/president/us/general_election_trump_vs_clinton-5491.html) had Clinton winning by an average of 3.2%, well within the margin of error (a concept that we will discuss later in the chapter). Where the polls fell down in 2016 was predicting the presidential election outcome in states, particularly Midwestern states (
www.washingtonpost.com/politics/were-the-polls-way-off-in-2016-a-new-report-offers-a-mixed-answer/2017/05/04/a80440a0-30d6-11e7-9534-00e4656c22aa_story.html). In this chapter, we attempt to give the reader a basic understanding of the issues involved in polling, and review some of polling's greatest failures.
Why Are 1,112 People Enough to Represent U.S. Voters?
For now, let's assume that every person of voting age votes, and in all elections, a Democrat runs against a Republican. A pollster wants to predict the fraction of the Republican vote in three different elections:
- A mayoral election in a city with 20,000 people of voting age
- A state gubernatorial election ...