A USA Today/Suffolk University poll conducted after last week’s presidential debate between Democratic nominee Kamala Harris and former President Donald Trump shows Harris leading 49%-46% in Pennsylvania, considered a key battleground state for the election on November 5th. The poll of 500 likely voters across three counties in the state also showed a greater number of voters viewing Harris favorably and less viewing her unfavorably than her opponent (49%-43% and 47%-53%, respectively). In addition, Harris holds an advantage in the gender gap, as she leads with female voters 56%-39%, while Trump leads by a smaller 53%-41% margin with men.
While these numbers are promising for the current Vice President, the overall picture stills point to a tight race. The RealClearPolitics average of the latest nationwide polls shows Harris leading by 1.8%, well below the 5% popular vote advantage considered necessary to overcome the hurdle of the Electoral college. The current RCP averages also show Trump with slight leads in Georgia (+0.2%), Arizona (+1.3%), and even Pennsylvania (+1.3%), despite the recent data from USA Today. RCP shows Harris leading by similarly slim margins in Wisconsin (+1.2%), Michigan (+0.7%), and Nevada (+1.2%). RCP’s national electoral map currently lists all of these as “toss-up” states.
The small advantages for each candidate according to the averages also raise the question of the margin of error and polling accuracy more broadly. Trump’s 2016 victory famously flew in the face of available polling that showed Clinton winning, and in 2020, numbers that predicted a comfortable win for Biden did not reflect the fine margins on which the election was ultimately decided. Josh Clinton, political science Chair at Vanderbilt University, has recommended doubling the margin of error indicated in a poll, and to remain skeptical of the result in the difference between the candidates is still within it. The 3% advantage held by Harris in USA Today’s Pennsylvania poll is already well within the stated 4.4% margin of error.
The number of respondents also plays a key role, as polls reaching larger numbers of voters are logically viewed as more credible. Senior politics reporter for Vox Christian Paz considers a minimum of 1,000 respondents for national polls to be a good baseline. With that scale in mind, USA Today’s 500 respondents for a single state appears proportionally appropriate. That said, another poll conducted by InsiderAdvantage over the same period in Pennsylvania, with 800 respondents, found a 2% advantage for Trump.
