STRENGTH IN NUMBERS

A data journalist for the Economist explains his much-derided and now much-distrusted profession.

In the 2020 presidential election, a poll taken by ABC News and the Washington Post projected that Joe Biden would defeat Donald Trump in Wisconsin by a towering 17 percentage points. Ultimately, Biden’s lead was less than 1 point. How did the pollsters get it so wrong? In 2016, how did everyone who called the election for Hillary Clinton misread the signs? Morris looks deep inside the often flawed assumptions of the pollsters and efforts to overcome the mathematical flaws inherent in their surveys, from sample bias to margins of error. Before doing so, however, he defends the use of polls as an important mechanism to give voice to voters in a representative democracy. “We must understand that both the concept and the significance of public opinion took root gradually, and their development continues to this day,” he writes. When properly conducted, he adds, a poll can have the force of a referendum, given that a key assumption of democracies is that the voice of the collective is stronger than that of the individual. But how to assemble that collective to give meaningful results? As Morris notes, some of the problem lies on the side of the pollsters, who must attain samples sufficiently large and diverse to represent as many demographics as possible. Some, though, lies on the side of those being polled, who, it seems, tend not to answer truthfully, especially when they suspect that the poll is biased toward one end or another. In the 2020 election, right-leaning voters tended not to respond to polls at all, again leading to projections of a Biden landslide. Those readers with a bent for statistics will take interest in the author’s descriptions of such matters as sampling errors, the law of large numbers, and the corrective tools of smoothing and aggregation.

Morris makes a solid case for polls as tools to give voice to the people while allowing that improvements are needed.