Predicting who wins a presidential election is probably not a wise thing to do, especially for a political science professor. Get it wrong, and you have years to live it down. I won’t offer a prediction; I’ll just offer some unhelpful hints to explain why we really don’t have a good guess about who will win, primarily because we aren’t sure who is actually going to vote. That usually makes all the difference.
Right now, September 2020, there are national polls, state polls and polls of polls. Everywhere you look there are predictions. But what do we really know? First, national polls are interesting because they can give us a sense of the overall popular vote. Remember, though, that we elect a president based on the Electoral College vote. Winning the popular vote and 10 dollars gets you a bucket of popcorn at a movie. Most theaters are closed, so you don’t even get the popcorn. What matters are the state-by-state races.
Of course, there are polls for every state and averages of polls for every state. A useful way to track those is to look at the Real Clear Politics Battleground State averages. We know, however, that turnout is a big factor. The people who are polled aren’t necessarily representative of the people who will vote; they are representative of the people who pollsters think will vote.
We do, however, have 2018 as a possible guide. The 2018 midterm election is the last time people actually got up off their couches and voted. Perhaps it is a good indication of what the 2020 vote might look like. Neither Trump nor Biden were on the ballot, so it is an imperfect analog for 2020. But it does give us a way to compare support for candidates belonging to each major party.
Using data from House of Representatives races, we can see which party carried the overall statewide vote in all 50 states and D.C. (taxation with only partial representation!). If we add up all the votes for the House candidates from each party, we can get a total party vote for each state. Then we can give the electoral votes for that state to the party that received the majority of votes. From that, we can get a picture of what the Electoral College might have looked like if 2018 had been a presidential year (sort of). That assumes 2020 will reflect the trends of 2018. Hold that thought because it might be a bad assumption.
I did this same exercise for the 2016 election back in June of 2016: What if the 2014 vote for House candidates was a preview of the 2016 election? That calculation gave the 2016 victory to Trump: 321-217, not too far off from his actual margin of victory, 304-227 (seven electoral votes went to other candidates). While this predicted the Trump victory, the details don’t quite fit: Republicans won the 2014 vote in the following swing states that Obama won in 2012—Colorado, Florida, Iowa, Minnesota, Ohio, Pennsylvania, Wisconsin and Virginia. In the 2016 election, Trump actually won these swing states that Obama won in 2012—Florida, Iowa, Michigan, Ohio, Pennsylvania and Wisconsin. The 2014 “prediction” got five of eight right. Michigan was a surprise. Colorado, Minnesota and Virginia leaned back to the Democrats in 2016.
"Remember, though, that we elect a president based on the Electoral College vote. Winning the popular vote and 10 dollars gets you a bucket of popcorn at a movie. Most theaters are closed, so you don’t even get the popcorn. What matters are the state-by-state races."
What does the House vote in 2018 predict for 2020? Biden gets the victory, 296 to 242 electoral votes. Of the key swing states, in 2018 Democrats captured Arizona, Iowa, Michigan, Minnesota, Pennsylvania, Virginia and Wisconsin. Of those states, only Arizona and Iowa were Democratic victories by under 5% (one of the definitions of a swing state). That could be bad news for Republicans. Perhaps even more disheartening for Republicans is that support for Republican House candidates’ in some states was shaky—Florida by 5%, Georgia by 4.5%, Montana by 4.6% (wow!), North Carolina by 2%, Ohio by 4.7% and Texas by 3.4% (double, triple, quadruple wow; that’s 38 electoral votes). All of these suggest declines in Republican support from 2016 except Florida (an increase) and Georgia (about the same).
Should Trump be worried? Should Biden be excited? Honestly, they are all worried. That’s part of the process.
Now, here’s a reason to ignore everything I’ve written so far.
Take the exercise above and do it for the 1994 midterm elections. What did they predict for the 1996 presidential election? The 1994 Republican Revolution would have translated into a Republican presidential rout, 380 to 158 in the Electoral College. Instead, Democrat Bill Clinton won reelection in 1996 with an electoral total of 379 to 159, a nearly perfect reversal of the 1994 numbers. In 2010, Republicans would have cruised to a 345-193 margin of victory, but Obama won in 2012 by 332-206.
What does it all mean? The 2018 midterms may foreshadow 2020 and they may not. (I’m so helpful that way.) It depends on who comes out to vote. The electorate for midterms and presidential elections can be different animals. Obviously, personality matters. In 2020, an outsized personality such as Donald Trump is likely to be the difference in who wins or loses. If Trump supporters sat out the 2018 election, but are ready for 2020, Trump might do what Clinton and Obama did—win in spite of crushing midterms. If those who oppose Trump repeat their large turnout from 2018, Trump may be a one-term president. We’ll see who actually turns out to vote. In any case, whatever you believe, whomever you support, VOTE.
About the Author
William Newmann, Ph.D., earned his doctorate in public policy from the Graduate School of Public and International Affairs at the University of Pittsburgh in 1999. He earned a master's degree in political science from Drew University in 1985 and a bachelor's degree in political science from the University of Pennsylvania in 1983. He joined the VCU Department of Political Science in 1992.
Professor Newmann’s research focuses on the way presidents use their advisers when making decisions in national security affairs.