N.B.: This post is a hypothesis about “what happened” in the US presidential election on Tuesday. It deliberately makes no statement about what ought to have happened, nor should it be read searching for hints regarding my political views.
Numbers, like dead men, tell no tales. As late as the afternoon of election day, mainstream media outlets confidently predicted, on the basis of dozens of polls, a relatively easy victory for Hillary Clinton over Donald Trump. That is not what happened. Many people were surprised, whether positively or negatively, by Trump’s decisive victory over Clinton, and many are trying to figure out why it happened the way it did. Yet the bare numbers, the responses to opinion polls and the voting tallies, are silent about what chain of events led to these results. All explanations of what happened are necessarily speculative, but this post proposes one mechanism. In particular, it explores the possibility that the mainstream media might have unintentionally discouraged people from voting for Clinton.
That was not the point, of course. Most mainstream news outlets had explicitly endorsed Clinton’s candidacy. Even Fox News, typically a bastion of conservative politics, clashed horns with Trump’s supporters and seemed too liberal for the Republican candidate. The message presented by the New York Times and other liberal news sources seemed to be, “We’re winning!”
There are various possible explanations for the discrepancy between the newspaper poll results and the voting results. Some journalists are suggesting it was a failure of data analysis. Perhaps the newspaper polls, despite their intentions, were not sufficiently representative. Perhaps they did not find access to the actual bastions of support for Trump. Or perhaps Trump supporters lied to the newspaper polls, claiming they would vote for Clinton. Or perhaps many Clinton supporters changed their mind; Clinton herself allegedly blames her defeat on the late announcements by FBI director James Comey regarding the investigation into her email scandal. Yet, as one friend pointed out to me, the polls had worked well in 2012, as well as in this cycle’s primary elections, when Trump’s supporters had not feared declaring their support.
Perhaps a clue may be found in the trend of popular votes in the most recent four elections, 2004, 2008, 2012, and 2016. (I have seen variations of this graphic online, but I generated this from Wikipedia data.)
The Republican totals here are relatively constant, varying less than 3.5% from maximum to minimum. By contrast, the total number of Democratic votes varies almost 18% from the tepid 2004 vote for John Kerry to the enthusiastic 2008 vote for Barack Obama. Clinton’s turnout was scarcely 3% better than Kerry’s loss, and more than 8% below Barack Obama’s (lower) 2012 turnout. (In fairness, Trump’s popular vote was almost 3% below George W. Bush’s 2004 victory, and marginally less than Mitt Romney’s 2012 defeat.) The biggest variability is in the Democratic popular vote, and perhaps we should look for factors affecting that aspect in particular.
One (sinister) possibility is that the fear of voter intimidation discouraged voting among demographics who might expect to be profiled as Clinton supporters. Donald Trump had warned his supporters that the system was rigged and “not to allow” Democrats to steal the election. Some of his supporters mentioned plans to conduct surveillance on polling places. The mainstream media enthusiastically reported these remarks to make their case against Trump’s fitness for office, thus spreading the fear of voter intimidation. It is possible that some visible minorities, especially in swing voting districts, feared the immediate personal danger more than a Trump presidency, especially since the latter was declared unlikely (see below), and so they decided to avoid the polls. Although in the event the only violence reported near a polling place that day was unrelated to the election, this was not necessarily obvious ahead of time. The media’s amplification of Trump’s remarks may have scared away potential Clinton voters.
On the other hand, it may be relevant to remember that this was largely a negative election, in which many voters were voting against a candidate rather than for any candidate. The mainstream media reported extensively on various groups of people who were supporting Trump, not because they liked him, but to prevent another Democratic presidency. What some journalists may have forgotten is that, despite their own enthusiasm for the Democratic candidate, there were also many who declared they would vote for Clinton, not because they liked her, but to prevent a Trump presidency. But when the news media reported that a Trump victory was very unlikely, perhaps many of these people decided their vote was superfluous, and decided not to vote. In a negative election, declaring one candidate the likely winner may well cut away the reluctant votes in their favor, while energizing the opposing side. A premature declaration of victory is likely to backfire in an election season like 2016.
Physicists tell us that in very sensitive measurements, such as at very small scales, one cannot observe a phenomenon without altering it. This election season has been very sensitive, and I am sure I am not the only voter who paid attention to the media coverage of election opinion polls while deciding for whom to vote. Some journalists are blaming their failed predictions on their poll numbers, or on their own bias toward even-handedness (!), yet the decisive shift from forecasts to reality may have taken place in the popular reception of media coverage, which trumpeted both a potentially violent voting experience and one candidate’s almost certain defeat. A pro-Clinton media may have unintentionally driven down the popular vote for their preferred winner.