Monday, September 19, 2016

Polls Mislead Now As In 2012?


Setting aside all the fireworks ginned up by commentators, there's a question more fundamental than the strange appearance of this year's election campaign: At this point in an election, how good are the polls in predicting the actual outcome?

This week everybody started to jump up and down because the polls showed Trump "closing the gap" in the national popular vote. Worse, it also appears he's catching up in some of the key swing states that will decide the electoral college, the vote that actually hands over the keys to the White House.

Yet I still persist in saying Trump will lose in spite of the new round of polls. See: Ignore the "Tightening". Trump Still Can't Win.

Today I've done some more checking and found more support for my conclusion. I remembered 2012 as a nail-biter right up until election night. That memory is corroborated by this roller coaster bit of a polling chart from 2012, covering the last 8 weeks of the campaign.  2012 - General Election: Romney vs. Obama | RealClearPolitics



As the above chart shows, right about now (mid-September) in 2012, Romney took a great leap upward in the polls and caught Obama. Then he nose-dived again, then subsequently lunged upwards to catch and pass Obama, then down again, then a bit later bounced up again briefly so as to catch him one more time near the end. Then finally Romney chased Obama in in vain in one final spike upward by the two just as the campaign ended, Obama finally leaving him in the dust.

In other words, it's possible that what we have witnessed this week is just the first of several yet-to-come up-and-down spikes by Trump. In 2012 the first set of spikes was attributed to the two nominating conventions, which were late that year, the GOP being in the last week of August and the Democrats in the first week of September. This year the conventions were much earlier. Maybe the early September spikes in both years are more due to the fact that a lot of people just don't pay attention until the first whiff of autumn.

Whatever the reason for the spiking, the homestretch polls of 2012 show that perhaps we are in for several more "closing the gap" episodes by the GOP candidate.

What about the last gap? (Or should I call it "the last gasp"?) The one on election night?

In 2012 Romney didn't even come close to closing that gap. Instead it widened into a chasm! Romney got only 206 electoral college votes to Obama's 352. Not even close!

Were the last 8 weeks of the 2012 campaign an aberration, a one-time only failure of projections based on polling?  It doesn't seem to be. Let's look at the scorecard for the projections in 2008. Keep in mind that Obama beat John McCain in 2008 by a whopping 365 to 173 in the electoral college vote. Nevertheless, at this point in September 2008, the electoral college projection was woefully wrong compared to the eventual outcome. Based on the averaging of polls in mid-September 2008, RealClear Politics gave John McCain 216 electoral college votes to Obama's projected 202. That's a miss that was a mile.

These wildly wrong projections aren't due to bad polling, although RealClearPolitics includes some second-rate pollsters in its averaging. It's more likely just very hard to project an election outcome even just six weeks or a month ahead of Election Day. The would-be voters who pay the least attention until these closing weeks are also less commited to their initial choice. They may swing and sway quite a bit during the last weeks before the election.

That's especially likely this year with a mercurial personality like Trump on the loose.  And he is definitely on the loose. For a couple of weeks he's been reined in by new campaign management, but this weekend he has suggested Hillary Clinton's Secret Service protectors be disarmed. Is he suggesting someone shoot Hilary Clinton?

One would think a lot of Republican suburban housewives would flinch away from this latest  hooligan remark.  If they do, that might reverse the recent gains Trump's made, much of which has been attributed to disaffected GOP voters getting more comfortable with him during the recent weeks he's toned down the raucous rhetoric.

Supposedly we'll know more in another couple of weeks as to whether this recent Trump spike means anything. Analyst Nate Silver has a piece in "538" on why it's too soon to tell:
Election Update: Democrats Should Panic … If The Polls Still Look Like This In A Week

Nate Silver may be right and that a week or two will give us the clear picture, but I'm not sure. The little chart from 2012 shows the GOP candidate Romney still dancing up and down the walls after that first big jump, with the up and down repeating until Obama finally pulls away at the very last moment, leaving Romney to continue counting his cars and tying his dogs to the car roofs.

Next time, I'll take a look at what may be the actual state of the states. i.e. what's swinging in the swing states.

Meantime, relax. We have a ways to go. Probably a number of ups and downs.

So give the fingernails a rest.




  









ALSO Chris Cillizza of W. Post   Don't look now: Donald Trump has all the momentum in the 2016 ...  (but buried near the end) "The electoral map still heavily favours Clinton unless Trump can find a way to make Pennsylvania competitive, a task that has so far proved elusive. Trump still must win states like Ohio, Florida and North Carolina, which, even with his recent surge factored in, remain no better than toss ups today."


1 comment:

  1. I knew it in 2012. I know it now. It's going to be a landslide. I have no expertise, but I was right once. Just in case, I'll be out urging people to vote in Colorado. Go Hillary!

    ReplyDelete