There have been a lot of efforts thus far to read the tea leaves of early voting and declare the 2016 election over already before Election Day, thus discouraging voters from coming out on Tuesday. This is, in fact, one of the anti-democratic features of early voting, and a likely contributor to why at least one peer-reviewed study has concluded that early voting reduces voter turnout. But in fact, experience has shown us that early voting numbers may not be all that useful in predicting results, in part because early votes just cannibalize the most-enthusiastic part of a party’s voting base:
[P]redictions from early voting have a decidedly mixed track record. As University of Denver political scientist Seth Masket has suggested, the relationship between early voting results and Election Day results is pretty weak. That isn’t to say it is non-existent…A contradictory report has been published at the Monkey Cage blog, but that finds a relationship utilizing proprietary Catalist data, not the publicly available data from which everyone is presently extrapolating.
But if predictions from 2012 were iffy, predictions from 2010 and 2014 were awful. In 2010, analysts saw huge Democratic advantages in turnout in places such as Ohio and Iowa and thought that perhaps there was no enthusiasm gap in the election. In 2014, it was widely assumed that early vote totals were good news for Democrats in states including North Carolina and Iowa; Thom Tillis ended up winning in North Carolina on the back of strong Election Day turnout, while the 2014 Iowa Senate race was decidedly not close (as early vote analysts had suggested); Joni Ernst won by almost 10 points.
That’s not to say that Trump supporters shouldn’t be worried about the early-voting reports showing, say, a lot of Hispanic voter turnout in Nevada and Florida – a result that suggests Hispanic-heavy states like Arizona and Georgia could also be at risk. On the other hand, Democrats should probably be worried about generally lower African-American turnout across the board. But we simply don’t know what the total Election Day turnout will be just yet.
I’m on record for over a year now saying I don’t believe Trump can win a general election, and I still don’t think he will. But as election analysts like Nate Silver and Sean Trende and Nate Cohn have been warning, just because the likely outcome from the polling and early voting data shows a likely Trump loss does not mean that’s what we’re going to see on Tuesday. Intensively-polled races are rarely way, way off – but an error of a few points off the poll averages is fairly common. Polls underestimated Republicans rather systematically in 2002, 2014 and 2015, and Democrats in 2012 and 1998. As Trende noted a few days ago, if the polling averages underestimated Trump by just as much as they underestimated Obama at the same point in 2012, Trump would win.
Even knowing which polls to look at, of course, is a hotly contested subject, as the different poll aggregators at RealClearPolitics, 538, Huffington Post Pollster, and the NY Times Upshot all include and exclude different polls they regard as unreliable because they are partisan or too internet-based or have poor track records. I tend to use RCP in part just because it’s the most user-friendly of the sites and has the longest track record for historical comparisons. We face today something of a polling “tragedy of the commons,” in that people pay so much more attention to the poll aggregators than the polls themselves that there’s less incentive for pollsters to produce them – which in turn affects the reliability of the aggregators.
RCP, at this writing, has Hillary winning the “no tossups map” 297-241, with Florida costing Trump the election, but Hillary winning Florida by 1.2 points, Pennsylvania by 2.5, Colorado by 2.9, Michigan by 4.
538 has Trump losing by an average of 291-246, although its map of where states lean has Trump losing just 272-268 (due to losing New Hampshire, Colorado and Pennsylvania), and a 35.2% chance of winning.
HuffPo has Hillary up by 1.4 in Nevada, 2.1 in North Carolina, 2.7 in Florida, 2.9 in New Hampshire, 4.5 in Colorado, 5.5 in Pennsylvania, 5.8 in Wisconsin, 6.5 in Michigan, 6.6 in Minnesota.
Upshot has Trump trailing by 0.6 in Nevada, 1.5 in Florida, 2 in New Hampshire, 3.5 in Wisconsin, 3.8 each in Colorado and Michigan, 3.9 in North Carolina, 4.7 in Pennsylvania.
Axiom Strategies, which polls “battleground counties” in eight states, has Trump up in the battlegrounds in Ohio, Florida, North Carolina and Nevada, Clinton +1 in Colorado, +2 in Pennsylvania, +4 in Virginia and Wisconsin.
If you’re wagering money on this race, the safe bet right now is that all of these poll-based averages and forecasts – which all point in the same direction across multiple states and nationally – are right, and Hillary wins. Moreover, Hillary has a formidable ground operation, and Trump has virtually nothing. But given the history of polling in recent years, it’s not at all implausible that polls this close could be wrong across the board by just enough to flip the map. And if that happens, after Hillary leading the polls all the way to the end, expect a very ugly aftermath to Tuesday.