The danger of using small sample sizes to reach big conclusions was demonstrated once again when a poll of 800 voters indicated that former Vice President Joe Biden had squandered his lead in the Democratic presidential primary, only to have a larger survey released two days later that showed him with a still-comfortable edge over his closest pursuers.
A Monmouth University poll released Aug. 26 showed Mr. Biden had slipped 13 points since it last took the voters’ temperature in June, dropping from 32 to 19 percent. His loss had been the gain of Bernie Sanders and Elizabeth Warren, as the Senators from Vermont and Massachusetts each hit 20 percent, with Ms. Warren picking up the most support since the earlier survey.
The caution flag that came with the poll was that because it had just 800 respondents, the margin of sampling error was plus or minus 5.7 percent. That meant it might be off by as much as 11.4 percent in gauging voter sentiment.
Many pollsters and political consultants will argue that any poll of fewer than 1,000 people is more likely to be an outlier. We’re not sure why that’s the magic number, except that the larger the group being examined, the greater the possibility it’s on target. To put it in sports terms, a baseball player with no hits in three at-bats against a particular pitcher hasn’t faced him often enough to justify a conclusion that he’s overmatched, but one who could muster only five hits in 50 at-bats could fairly be said to be “owned” by that pitcher.
Quinnipiac University on Aug. 28 released a survey of 1,422 voters taken between Aug. 21 and 26 (the Monmouth survey was conducted over five days rather than six, Aug. 16 through 20), a large-enough sample to reduce the margin of error to 3.1 percent, which meant it could be off either way by 6.2 percent. It also had a larger contingent of voters who identified as either Democrats or leaning Democratic: about 46 percent compared to 37 percent of those polled by Monmouth.
And its number for Mr. Biden—32 percent—was identical to what he had gotten in Quinnipiac’s previous survey, released 22 days earlier. The poll showed Ms. Warren, who had surged to 21 percent in early August after being widely regarded as the best performer in two nights of candidate debates on the final days of July, dropping to 19 percent, while Mr. Sanders continued to regain support he had lost in early summer, with 15 percent preferring him.
Kamala Harris remained at 7 percent after falling 5 points on the Quinnipiac scale in early August after a tougher time in her second debate than she had when her strong showing at Mr. Biden’s expense during a late-June faceoff briefly propelled her into second place with 20 percent. Pete Buttigieg completed the Top 5 with 5 percent, unchanged since earlier in the month.
All five of those candidates had comfortable margins when respondents were asked by Quinnipiac to state a preference if each went head to head with President Trump next November. Mr. Biden had the largest edge, at 16 points, 54 percent to 38 percent, but Mr. Sanders also led the incumbent by 14 points, Ms. Warren was up 12 points on Mr. Trump, Ms. Harris by 11 and Mr. Buttigieg by 9.
Quinnipiac polling analyst Mary Snow said that what should be particularly worrisome to the President was that in none of those match-ups did he exceed 40 percent of the vote. White women, who stunningly chose Mr. Trump over Hillary Clinton in the 2016 election, prefer his five most-likely Democratic opponents this time by double-digit margins.
In case you were wondering, the poll showed Mayor de Blasio finally getting to the elusive 1 percent mark, although that wasn’t good enough to make him eligible for the next Democratic debate Sept. 12. And while he was the choice of 3 percent of voters ages 18-34, no other age group gave him as much as 1 percent. His overall standing put him even with Sen. Kirsten Gillibrand, who early that evening announced she was dropping out of the primary contest.
We depend on the support of readers like you to help keep our publication strong and independent. Join us.