All too often, media attention is applied to national polls in presidential elections. However, we don't really have a national presidential election in the United States—we have 50 state elections. While the polling industry was criticized for missing the Trump victory in 2016, most national polls were actually pretty much on target (Clinton by 3.3 points). In fact, the former Secretary of State won the national popular vote by a margin of roughly three million (a spread of 2.1 points). It was the various state polls that were mostly off-base—giving Clinton slim margins in Michigan, Pennsylvania and Wisconsin. Of course, when the votes were counted, a slight shift of a few thousand votes resulted in all three going to Donald Trump—providing his Electoral College victory. Could those state polls have detected that small shift in sentiment at the end of the campaign? Maybe. Could the state samples have provided a better balance of key demographics like age, party ID and gender? Probably.
In any event, it's the state polls that should be paid attention to in these last three months before November 3rd. And, tellingly, some of those state trends are moving toward the incumbent. According to Real Clear Politics, former Vice-President Biden's lead in six key "battleground" states (Wisconsin, Pennsylvania, Michigan, Florida, North Carolina, Arizona) is down to an average of 5.6 points. In Pennsylvania, one poll (CNBC) has Biden's lead down to just 2 points.
A whole range of issues will impact the ebb and flow of these numbers—including the perceived response to COVID-19, the state of the economy, and on-going protest demonstrations. But it is these state-based dynamics that will ultimately decide this race—not national trends.