Since many will not be subscribers to the ‘Real Clear Politics’ bulletin I reproduce below their interesting article on polling and that you can tell which way the election will go by 9:30 p.m. on election day. Another way would be to compare the early states results as they come in with the same states and the same percentage of voters tabulated with that of 2016
“We May Know How Good (or bad) The Polls Are Before The Election Is Called
What is the deal with these polls? Polling missed Trump’s share in 2016 and 2020, but after post-mortem studies, adjustments within the industry, new trackers, and a plethora of more Republican-friendly firms entering the scene, polling averages now find former President Donald Trump at a stronger showing nationally and in key states like Pennsylvania than he was four years ago. Vice President Kamala Harris is performing stronger in national polling than President Joe Biden, upending what had been a small but steady lead nationally for Trump. There’s extensive debate about whether we should “trust” this year’s polls, with arguments against ranging from differential partisan nonresponse to zone flooding. Crosstabs took a particular amount of heat on social media earlier this cycle, and I’ll make a quick shout-out to Adam Carlson, who took the task of aggregating crosstabs seriously enough to build a regularly-updated average of them, visible here.
Personally, the concept of “trusting” polls seems to miss the entire point of them. Polling, when done properly, can give us insights on the voting populations’ thoughts and motivations at a specific point in time. But when obsessed over, folks exhibit full blown anxiety- or euphoria- over daily or even hourly rises and drops of tenths of a point. Instead of treating these as the fascinating tools they could be, people have held onto forecasts and the poll of the day as flips of the electoral tarot cards yielding good tidings or doom. And that’s before considering the dreaded “Margin of Error”, which is often overlooked when considering the state of a race and the accuracy of a poll.
Well, this is not fortune-telling. Polls (or better yet polling averages) can be a valuable tool but they aren’t iron clad guarantees of a specific outcome.
We are not here to tell you what Saturn in Aquarius will mean for voter turnout in Philadelphia.
Tunneling back down from the models to the very polls that feed them, and which have caused such consternation for so many, we will not know if they were “off” again for quite some time.
That time being 9:30pm Eastern Time, November 5th, 2024.
What is so significant about this oddly specific moment? Based on recent cycle election return patterns, this is around the moment where we can expect to have a majority of the early/absentee vote reported out of Pennsylvania and North Carolina, most of Florida already counted, and the election day in-person tallies rolling in from Metro Atlanta. By this point in the night, we will know if polling in Florida, a non-swing-but-once-was-state, missed the mark again like in 2020, or landed near-spot-on. A host of non-swing states will have released significant shares of their unofficial results, too.
With the early vote out of Ohio, North Carolina, and Pennsylvania’s largest counties *mostly* accounted for by 9:30pm, we can begin to see how Democratic it was (Democrats vote earlier more consistently than Republicans in every state mentioned save for Florida, where it’s a bit of a push now), and how it compared to what was surveyed. By the time we enter late October, state-level polling will often have voters broken down into the “already voted” and “planning to vote on election day” categories. We can compare what those who claimed they had already voted told pollsters, versus what we are seeing in that category half an hour past nine.
It is important to keep in mind that not all of the absentee vote will be counted by this point in the Keystone State. But compared to November 2020, recent election returns have run like lightning. In April of this year, *half* of the Democratic Presidential primary vote had reported by 9:15 pm, and most of those tallies corresponded with the absentee vote drops from reporting counties. While the in-person vote tallies will take hours to complete, the absentees will not take as long as they did four years ago.
By 9:30pm, the vote tallying in Kentucky and Indiana will have mostly concluded. These are the first two states to close in the country, and only a small share of each will remain outstanding. By no means complete, if you have polling in each before, and compare it to the results at about this time, you may be able to tell if any pattern of national poll “miss” is forming. Much of the early (in-person) vote in Texas will have been reported by this time, too. A red-leaning state, it is currently polling tighter than earlier this year. In 2020, many Democrats were excited by polling that pointed to the possibility of Blexas, a possibility that came up over 621,000 votes short.
By 9:30pm, Florida is winding down with its reporting. State regulations mandate that all Supervisors of Elections must release their early vote tallies by 7:30pm local time. Even given the split time zones, which means the early vote is fully reported by 8:30pm, and in that interim between zones, much of the election day vote is in from Duval to Miami-Dade, Monroe to Alachua. Indeed, by 8:30pm on August 20th, 90% of the statewide primary vote had been reported by the Florida Division of Elections.
By 9:30pm, most of Greater Atlanta will have reported its in-person early vote and a sizable chunk of its absentee vote. Election day in-person votes will have been reported from Hall, Clayton, Forsyth and Cherokee, and a scattering of such votes in Cobb, Fulton, Gwinnett and DeKalb will have been reported as well. Like with the states that drop their early votes first (Ohio, Pennsylvania, North Carolina) we can begin to compare early totals to poll crosstabs broken down by “time” (early or E-Day) of vote.
By 9:30pm, election day in-person votes will be rolling in from Maine, New Hampshire, Vermont, Connecticut, Massachusetts, Rhode Island, Delaware, Maryland, Virginia, West Virginia, New York, South Carolina, Tennessee, Illinois, Iowa, Missouri, Arkansas, Nebraska, Kansas, South Dakota, New Mexico, Oklahoma, Mississippi, and Alabama. Minnesota and Louisiana will be reporting a solid percentage of their early vote. We will have an explosion of separate data points we can begin, carefully, to compare against our polling pre-election. It will by no means be perfect real-time comparisons. But with a sizable portion of America’s counties reporting, the polls will start looking (the industry hopes) pretty spot on, or (once again) several clicks off.
This doesn’t mean we will know, with certainty, who is winning by this time: current polling shows negligible separation in Pennsylvania, North Carolina, Georgia, Arizona, Nevada, and Michigan; and a slight Harris lead in Wisconsin (see, “Margin of Error” above). If polling remains close, and the results we are seeing by 9:30pm mirror what the polls currently show, neither candidate will be feeling overly confident. Within-the-margin-of-error bars means waiting for several days of updates in Arizona and Nevada, even a week’s worth; and if there is a late surge in mail ballots returned in Pennsylvania, thousands of ballots may not be counted for another day or two. But at least we will know if the polls this time were more accurate.
Because after all is said and done, the only polls that matter are the ones on election day.
Excellent article and a lot of interesting information