Saturday, February 10, 2024

Trump’s lead over Biden may be smaller than it looks


Consider only the highest-quality polls, and the Republican’s advantage melts away

image: The Economist/Getty Images
IF AMERICA were to hold its presidential election tomorrow, Donald Trump would be picking out curtains for the Oval Office. The Economist’s polling average puts him up by 2.3 points over Joe Biden nationwide. And across the six swing states expected to decide the election—Arizona, Georgia, Michigan, Nevada, Pennsylvania and Wisconsin—he leads by an average of 3.8 points. Betting markets list Mr Trump as a clear favourite. Never in his past two campaigns were his general-election polls this strong. Is it time for the world to brace itself for a second Trump presidency?
The election is still nine months away. Historically, polls taken before the summer of an election year have been poor predictors of results. But no former president has sought to return to office since the advent of modern polling. Opinions about the omnipresent Mr Trump are much firmer than they are about typical challenger candidates, who at this stage of the race are usually still fighting to secure their party’s nomination. As a result, even though Mr Trump is not yet the presumptive Republican nominee, current head-to-head polls between him and Mr Biden may be unusually informative.
Nationwide surveys over the past month have varied widely, ranging from an eight-percentage-point lead for Mr Trump to a six-point edge for Mr Biden. Polling averages, which blunt the effect of such outliers, suggest that Mr Trump holds a clear lead. But the polls that comprise such averages differ in their methods and degree of rigour. Democrats hunting for a silver lining can take solace in one clear pattern: pollsters with the best records of accuracy show better results for Mr Biden. In contrast, their lower-quality counterparts give Mr Trump the edge.
image: The Economist
Public trust in polling has weakened following the industry’s high-profile underestimates of Mr Trump’s support in 2016 and 2020 (although polling before the 2018 and 2022 midterm elections was accurate). Reliably estimating pollsters’ accuracy—measured by the size of their historical errors and whether they consistently exaggerate support for a particular party—requires a large sample of surveys across many elections. FiveThirtyEight, a data-journalism outfit, recently updated its ratings of American pollsters. It assesses them on a combination of their records and their methodological transparency.
Some pollsters are consistently more accurate than the field. But there are lots of ways to judge quality. The Economist’s general-election polling average weights polls solely by sample size and recency, so that larger and newer polls contribute a greater share to the overall average. Using this methodology, Mr Trump leads Mr Biden in national polls by 2.3 points. That compares with a 0.2-point lead for Mr Biden in an unweighted average that gives polls from six months ago the same weight as those from this past week.
image: The Economist
The size of Mr Trump’s lead varies widely by the quality of pollster, as assessed by FiveThirtyEight (see chart). This early in the election cycle, the pollsters in its highest tier of quality have conducted polls only sporadically. (One exception is a weekly survey conducted by YouGov, an online pollster, for The Economist.) However, in total, 13 polls have been conducted in 2024 by firms in this group. On average, they show a virtual tie between Mr Trump and Mr Biden.
By contrast, most polls released in January 2024 have come from pollsters’ middle class: firms with good but not exceptional records. Polls in these (“good” and “decent”) tiers show Mr Trump with a 2.4-point and 1.7-point lead respectively. Meanwhile, pollsters with a poor record or no prior published results show Mr Trump with an average margin over Mr Biden of around six points.
National polls reflect the general mood, and correspond to the popular vote. But thanks to the electoral-college system, winning the popular vote is no guarantee of electoral victory. In 2000 and 2016, for example, Republican nominees won the presidency despite losing the popular vote.
Still, in recent decades the electoral college has typically benefited Republican candidates. If Mr Trump were to win the popular vote by a six-point margin, he would almost certainly win at least 358 electoral-college votes, giving him the largest Republican victory since George H.W. Bush took 426 in 1988. This would bring into play even states that Mr Biden won comfortably in 2020, such as Maine, Minnesota, New Hampshire, New Mexico and Virginia.
To survey-watchers who think that all polls are created equal, Mr Trump has opened a modest but growing lead nationwide. But to those who maintain that pollsters’ historical accuracy predicts future accuracy, Messrs Trump and Biden are in a dead heat. 
Subscriber only | The US in brief

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.