Saturday, May 14, 2016

A Close Clinton-Trump Race? Sometimes the polls are correct

What do presidential polls today tell us about the race in the fall and the final results come November 8?  This is a matter of academic and of course media debate, especially with the latter spending incessant time parsing the latest polls.  The simple answer is that there is a lot of confusion surrounding polling but that when done correctly they do give us some insights into the fall race. Having said that, a probable Trump-Clinton contest looks closer than many think.
First, it is true that surveys or polls are merely snapshots in time.  Depending on the wording they tell us what a sample thinks about some issue (such as their presidential preference) at a point in time and they are not always predictive of the future, especially in the future is distant and when we can assume that voters are undecided or are uninformed now and are likely to gather new information and change their minds more in the future.
At one time one could assume that presidential voting preferences were like a funnel.  By that, the further out from an election the more undecided voters there were and as they became more informed there were fewer and fewer undecided.  Thus the funnel shape.  Such a model also assumed voters were less well informed about candidates the further the election was away, that candidates were not as well known to voters the further an election was away, and that partisan preferences were not as fixed or that there were many undecided voters who could actually swing in their preferences.
So many of these assumption many no longer be true.  Political science literature points to the reality that partisan preferences have hardened and that there are fewer and fewer swing voters, if in fact many really do swing at all (besides swinging in or out from voting).  In 2016 it also appears that the penetration of the social media may be changing the knowledge that voters have about candidates such that they are better informed or at least now more about the candidates than may have been true in the past.  Finally, assuming a Clinton-Trump race, these are two candidates who are perhaps better known than any other two candidates in recent American presidential politics.
The point is that a lot of polling regarding these two candidates may be more accurate than we think.  Most if not all voters know who these two candidates are and they have already arrived at their views regarding what they think about them.  The only issue may be among a small handful of voters–perhaps no more than 10% of the electorate in a few states–is how they view Trump versus Clinton and who is the lesser of two evils given that both have high disapprovals.  Clinton’s collapse in polls vis-a-vis Sanders, conversely, may actually represent somewhat a more traditional model where her poll numbers have changed as voters acquire information about the relatively unknown candidate Sanders, or that the polls simply miss likely Sanders’ voters because they are not among the traditional group of people likely to vote in primaries or attend caucuses.
What this all suggests is that current polls that test Clinton against Trump may have more accuracy than one thinks and that they might be good predictors or what might happen in November. What does that mean?  More clarification of the polls is in order.
First, most national polls suggest Clinton has a really large lead over Trump in aggregate  public opinion polls and therefore Democrats are salivating over the prospects of a Clinton rout.  Think again.  Remember that presidents are not elected by direct public opinion or national popular vote but by the Electoral College.  Remember Gore winning the popular vote to Bush in 2000 but losing the electoral vote.  Clinton probably does have a huge popular vote advantage, no doubt reflected by larger Democratic majorities in places such as New York, California, and other states where she will do well.  But remember that the presidential election is fought in 50 separate states (plus the District of Columbia) and in many ways it is down to about a dozen or so swing states where the battle will be won.  Here the recent Quinnipiac poll suggesting closer races in Florida, Ohio, and Pennsylvania are important.  Clinton and Trump are essentially tied here.  Other polls suggest some non-swing states may be close, perhaps suggesting good news for Clinton.  That may be the case but the point is that one should ignore national aggregate polls and look instead to state by state polling for something more accurate.
The one issue where the polls are perhaps not accurate at this point is in terms of party support for Trump and Clinton.   The conventions have not occurred and neither of the candidates have firmly consolidated support among their party bases.  We hear more about that with Trump and the GOP but Clinton faces a similar problem.  But there are signs this week that Trump is beginning to consolidate support.  Contrary to news reports, Republicans may still prefer Trump to Clinton and will vote against her or for him.  The same may be true for Clinton and Democrats.  The upshot is that it is still possible for this election to turn into one where Trump and Clinton consolidate partisan base support and fight over a few swing voters in a few swing states.  Yes, this is a unique election in many respects but it is still more than two months before the convention and there are many reasons to think that many political trends will stabilize such that the current polling in the swing states will represent an accurate picture of what might happen this fall.
The moral of the story is that there are many reasons to think that the closeness of state polling in critical swing states might actually portend a very close election where turnout is key and wooing the few swing voters in those states is determinative of who wins.

Final Note: Since last November 2015 I have given several talks arguing that the winning presidential candidate this year will need to raise $1.5 billion.  This week the NY Times ran a piece were Trump estimated that he needed to raise $1.5 billion for his campaign.


  1. I really liked this article and other articles you've authored.

    Would you be OK if I shared one of your articles with the community? I can provide more information about Writer Beat or answer any question, but better than anything I can say in words, please take a look at the site.

    If yes, just give me an "OK" and I'll handle the rest (there is no fee).


  2. If Clinton enjoys a "huge" popular vote advantage, Trump eking out an EC win is next to impossible. Gore's lead was a piddling 0.5%, and at the moment, D's arguably have a structural advantage in the EC. In 2012, Obama's national popular vote margin over Romney was 3.86%. In CO, the tipping-point state, Obama's margin was 5.36%. In other words, assuming a uniform national swing, Obama could've lost the popular vote by ~1.5%, or 1.9 million votes, and still prevailed in the EC.