Thanks to a decade-plus effort to extend opportunities to vote, we now have plenty of early voting data on which to chew. We don’t have results from early voting, as the ballots have not yet been tabulated at all, but in several states we do know the breakdown in party affiliation for the ballots that were requested, and were returned. Some of that data suggests certain outcomes, and some less so, but the early vote data generally works best when compared to earlier cycles for indications of turnout rather than ideas of who’s leading in the races.
Exit polling is even more problematic, and we’re likely to get a lot of confusion from them as news organizations comment on early releases. This came to a head in 2004 when early reports from exit polling suggested that John Kerry had moved ahead of George W. Bush on Election Day, only to have Bush win the popular vote and the Electoral College. That caused allegations of voter fraud — maybe the last time Democrats admitted to the possibility — and a lame attempt to block Ohio from filing its Electoral College returns.
That entire exercise ignored the fact that exit polling doesn’t do a good job of predicting outcomes; it exists to explain them. Exit polling data gets collected all day long to find the eventual turnout model for elections, especially in demographics such as age, gender, ethnicity, affiliation, etc. That data only becomes valid when it is fully compiled. Partial data sets for exit polling do not provide predictive outcomes because the turnout models can change significantly during the day, perhaps especially because of early voting. That is exactly what happened in 2004, when media outlets used non-predictive data in predictive ways, and while the data sets were still being compiled.
That isn’t to say that completed exit polls are meaningless. The networks will use the data in part to plug into their election models in order to call races — but that takes place while the results of actual voting are being published, after the polls have closed. We’re not likely to see much of that data, though, except perhaps in a few states where Senate and gubernatorial races are particularly close. Even then, though, the problem of early voting then begins to have an impact. Exit polling takes place by having data collectors stand outside of the precincts and conducting in-person surveys. That will give us a look at the results of Election Day voting, but in some states, half or more of the total vote will have been conducted well before Election Day, and a substantial chunk of that by mail. That may mean that even the complete data sets won’t actually be all that reliable, unless and until follow-up surveys get conducted among early voters after the election.
In the end, though, the completed exit polling only gives us a short-lived sneak peek at potential results. Within a couple of hours in most states, we’ll get the actual results anyway. The best course is to be patient, but in any case, don’t put any value in exit polling until those data sets are complete.