Don’t Be Astonished if Trump Wins

The conventional wisdom is that Hillary Clinton will win tomorrow’s presidential election in the US. The New York Times, the Washington Post, and most other media outlets are predicting a Clinton victory with varying degrees of certainty — the Huffington Post goes so far as to declare, with dubious precision, that Hillary is 98.2% likely to triumph.

The problem is that most of this certainty derives from the results of polls, and polling is, by definition, an inexact science: It involves interrogating a small sample of people and using their responses to guess the intentions of the entire population of voters.

Statisticians can readily calculate the margin of error which results from this sampling, but that figure is generally disregarded. Headlines proclaim the results of a poll as (say) a 48–46 lead for Clinton, and rarely mention the margin of error of (say) 3%. Clinton really might be leading by 5 percentage points among that population, or she might be trailing by 1, but this uncertainty is routinely downplayed.

In addition, many polls use statistical models to estimate the eventual composition of the electorate and to predict who is a “likely voter,” and these models rely on assumptions which may or may not turn out to have been accurate once the real votes are counted. Nobody yet knows what percentage of presidential ballots will be cast by black or Latino voters, or men, or senior citizens, or people with a college degree, but many pollsters have made guesses at those figures, and published results based on them. Some of those guesses will prove accurate, and some will not.

Poll-driven conventional wisdom already has been dramatically overturned by the actual outcomes of two major votes in the last five months. In October, Colombian voters rejected a peace deal with that country’s FARC rebels, in what was described as a “shock” result. In the United Kingdom, opinion polls indicated a consistent preference to remain in the European Union from around April 2015 to June of this year, when the surprise vote in favor of “Brexit” upended politics and devastated markets in Britain and Europe. But the conventional wisdom is essentially saying that “it can’t happen here.”

A victory for Donald Trump is fully plausible, though still unlikely. The result of any indivdual poll may be inaccurate or lie within the margin for error, but the preponderance of results over time favors Mrs. Clinton. The electoral map shows more swing states leaning her way than his, and early voting figures indicate high turnout among Latinos and other key Democratic constituencies. Nevertheless, Clinton supporters should keep the champagne on ice until the votes are counted.