As is well known, all major forecasters confidently predicted that Hillary Clinton would win the presidential elections of 2016. On the morning of the election day (November 8th 2016) at 10:41 am, when a number of American voters had already cast their votes, FiveThirtyEight’s Nate Silver still predicted that Hillary Clinton had a 71% chance to win. Others had predicted Clinton to win with even higher probability, between 85% to 99%.

A new article, just published by Cliodynamics: The Journal of Quantitative History and Cultural Evolution, brings up a little known fact that should have caused the statisticians take a second look at their predictive models.

There are many reasons why people lie to pollsters. In fact, people regularly lie to themselves. What we need is a more reliable predictor of likes and dislikes, than directly asking people. We need a good “proxy.”

In the article Trends in first names foreshadowed Hillary Clinton’s electoral defeat Stefano Ghirlanda argues that one such proxy is provided by the naming patterns. He writes, “naming decisions are not entirely rational, and can be influenced by seemingly extraneous factors of which parents are not aware.” He then proceeds to show “that trends in first names foreshadowed, in the USA, the defeat of Democratic Party candidate Hillary Clinton in the 2016 presidential election.”

The idea is quite simple. Parents tend to name their children after people they admire. They certainly avoid names associated with those they dislike. When was the last time you met anybody born after 1945 whose first name was “Adolf”?

Ghirlanda analyzed the data on first names of babies issued a Social Security number (which is, essentially, all the babies born in the US). Here’s the pattern that he saw when he looked at the dynamics of the name “Hillary”:

“Hillary” (and its variant “Hilary”) has been trending up in popularity for several decades, but after 1992, when Bill Clinton was elected president and Hillary Clinton was thrust into the limelight, the popularity of this name collapsed. There was a slight recovery to 2009, but then Hillary Clinton became the Secretary of State (from 2009 to 2013), and the popularity of her name started trending down again.

Such naming dynamics “suggest an exceptional negative reaction to Hillary Clinton’s public image.” Returning to the post-1992 collapse, Ghirlanda writes:

I show that such a sudden reversal is unique among naming trends, and is also unique to the 1992 election. In other elections between 1884 and 2012, a First Lady’s name had little impact on naming trends. These considerations, and others detailed below, suggest an exceptional negative reaction to Hillary Clinton’s public image, which may have ultimately affected the 2016 election. These results show that identifiable, if subtle cultural trends can foreshadow major social events.

Readers of my blog will, no doubt, point out that explaining past patterns is much easier than predicting future dynamics. As Yogi Berra once said, prediction is very hard, especially if it’s about the future.

Fair enough. So let’s test Ghirlanda’s insight. The next election will be held in 2020. Let’s then take a reading in 2019 on how the name “Donald” is faring, and predict whether our current President is going to stay on for the second term.