ANALYSIS — Even though Election Day is still seven weeks away, ballots are being sent and votes are being cast.
But please don’t try to count them before Nov. 3.
After months of polling, punditry, prognostication and projections, reporters and analysts are looking for anything tangible to report on and analyze. It will be extremely tempting to extrapolate results based on the reported numbers of early votes cast in person and by mail. But early vote totals only tell us the registration of the voter, not the contents of their ballot.
Why will it be so tempting to interpret (and over-interpret) these votes?
First of all, they are actual votes. Instead of polling numbers featuring registered or likely voters, these are real ballots, and thus they will be weighted more heavily in the minds of some reporters.
Second, there will be a lot of them. More than 136 million people voted in the 2016 presidential election. According to an early September poll conducted for CNN, 31 percent of registered voters plan to vote by mail (by sending in an absentee ballot before Election Day) and 25 percent planned to vote early (by going to a polling place and casting an early ballot). With turnout likely to be greater than four years ago, that means at least 70 million people are likely to cast a ballot before Nov. 3. That’s a lot of votes to ignore.
Third, it will be a steady stream of votes. The media loves things that can be tabulated (votes, delegates, money). And the hard work by the U.S. Elections Project, tracking and updating early voting statistics state by state, might be a blessing for keeping track of such a large amount of data but a curse when trying to not overreact to the same data.
Fourth, the votes will have a partisan affiliation. States with voter registration can track if the ballot cast is from a registered Republican, Democrat or some other affiliation. That’s the icing on the cake when it comes to tempting analysis.
So why not indulge in extrapolating the final results from early vote numbers?
On the most basic level, we don’t know what those ballots say. Sure, most Republicans will vote for a Republican and most Democrats will vote for a Democrat. But we don’t know that for sure until the ballots are actually counted. In 2016, 9 percent of self-described Democrats in Michigan voted for Donald Trump, as did 8 percent in Florida and 7 percent in Wisconsin, according to the exit polls. Faulty assumptions about crossover voters by a percentage point or two in those states could have easily led to faulty analysis about the eventual winner.
There’s also a danger in over-analyzing voter registration numbers since they can be a lagging indicator on partisan preference, particularly for federal elections. For example, in Kentucky, Democrats outnumbered Republicans in voter registration, 1.7 million to 1.5 million as of July. Yet the Bluegrass State hasn’t voted for a Democrat for president since Bill Clinton received 46 percent in 1996. And Kentucky hasn’t voted for a Democrat for the Senate since Wendell H. Ford was reelected in 1992.
It’s also unclear whether the early votes are proportional to the total number of votes cast. In fact, they almost certainly won’t be. According to the same CNN survey, 79 percent of registered Democrats said they planned to vote early compared to just 28 percent of Republicans. That means the early vote tabulations will almost certainly favor Democrats, but that doesn’t guarantee the overall electorate will be skewed disproportionately toward Democrats after Republicans vote in person on Election Day.
Analysis based on early vote totals is worse than projecting off a poll since mail-in ballots or in-person early votes will come from a self-selecting group of people who make up a slice of the total electorate, while a scientific survey is designed to capture a snapshot of the entire pie.
The call for caution when it comes to early voting statistics is probably futile. But we’ve got to try.
Nathan L. Gonzales is an election analyst for CQ Roll Call.