A Pollster’s New Year’s Resolutions and What We Learned from Mark Twain

January 11, 2017

Well, it’s a new year and two months have elapsed since the November elections, two months in which we have had time to assess what happened and take in much of the post-election analysis as well.  Frankly, the miscues and misguided analysis that plagued the pre-election coverage of the Presidential race seem alive and well in the ensuing post-election coverage.  We thought it might be helpful to break down a few aspects of the election, what hindsight offers and provide some advice on avoiding the pitfalls of the instant punditry that continues to get it wrong.  Of course, we realize that as pollsters our credibility may be suspect given the conventional post-election wisdom.  But here are some New Year’s resolutions you might find worth considering:


1:       Don’t get your opinions from the pundits who provide them.

If you want to know what is likely to happen or what actually happened you have to dig deeper than the ratings-hungry, political talk shows that have devolved into partisan point- making intended to deflect bad news or overstate positive news.  The conventional wisdom that comes out of this process was wrong before the election and things don’t appear to have changed very much since then. Let’s start with the instant analysis of what went wrong with predictions. Immediately following the election, the story was about polling missing the mark and praise was heaped on two polling firms which had Donald Trump winning the national vote, one having him ahead by 3% and another having him ahead by 2%.  In fact, these two organizations were described in an article in one newspaper as being “the gold standard going forward.” Apparently they “got it right” by being wrong. They had Donald Trump winning the national vote by 2 or 3%. He actually lost the national vote by a little over 2%.  On the other hand, the Real Clear Politics polling average had Hillary Clinton winning by 3.2%, only missing the actual result by a percent, well within any poll’s margin of error, and a lot better than the 4-5 point difference from polls now dubbed the “new gold standard.”


2:       Respect Margin of Error and the views of the undecided voter, especially in “Change” elections, which tend to break very late.

Polls have a margin of error and it is unwise to ascribe more specificity to poll data than the science dictates. Moreover, don’t assume you know how undecideds will break.  Being ahead 48-44% does not mean a candidate will win 52-48%. They may, and if so their pollster looks like a genius, but there are a lot of factors involved in getting from 48% to 52%, or even to 50.1%.  Heading into the 2016 election, voters were quite unhappy and the “right direction/wrong track” numbers were widely negative.  Faced with an unhappy electorate, late deciders generally break for the candidate they think is most likely to bring about change, especially in areas where significant economic dislocation has taken place.  Exit polling bears this out, where a plurality of voters said the most important quality they were looking for in a candidate was one who “can bring change.”  Donald Trump won these “change” voters, 82-14%.

In elections with insurgent candidacies and a disaffected electorate, those of us in the campaign consulting world are all too familiar with the expression, “an incumbent gets what an incumbent gets.”  That is, if you are an incumbent heading into Election Day and you are not over 50% in your polling in a two-way race, you cannot assume you will get there.  This is especially true in hotly contested primaries but also in general elections as well.  (Note: although she wasn’t technically the incumbent, perception is reality in politics and Hillary Clinton was, after 25 years in the limelight, as close as anyone could ever get to incumbency without actually holding the office.)

We saw this scenario play out in the three states that decided this election: Michigan, Pennsylvania and Wisconsin. Without Trump’s razor thin victories in these states the collective gasp and exasperation from the media about how all the predictions were so wrong would have never come to pass.  Every other state broke in a way that forecasters presumed they could, although not necessarily as they would have liked.  So let’s look at these three states.  Were the polls “wrong?”  No, they were not, not if you read the results correctly.  In every single one of these states, Hillary Clinton met or exceeded what she was getting on the final Real Clear Politics polling average, not by much, but the RCP average did not overstate her support.  Furthermore, many of these public surveys in these states were taken 1-2 weeks before Election Day, some, in Wisconsin for example, before the Comey email announcement (Oct 28th) and before the Wikileaks drip, drip, drip had reached full crescendo.


Real Clear Politics Polling Averages vs. Actual Election Results


And one final point on this: The race broke for Donald Trump very late, despite the prevailing sentiment that the race was already decided for Hillary Clinton. Yet the same article in the newspaper that bestowed dubious “gold standard” awards, also took modelers to task for not putting Michigan and Pennsylvania in the Trump column “even as the Clinton campaign rushed to furiously defend those states in the final days of the election.” Well first, putting these states in the Trump column would have been wrong at that point, even in hindsight. Those states ended up too close for any poll to predict the outcome with certainty. More importantly, if the Clinton campaign was “rushing furiously to defend those states” and victory in either of them could have swung the election to Trump, why is the result such a surprise?  Donald Trump had to thread the needle, but there was enough of a possibility that the Clinton campaign was scrambling at the end.


3:       Always remember: People vote their pocketbook.

Perhaps the biggest irony of this race is that the Clinton campaign never offered a compelling economic message to appeal to displaced and disaffected voters in the Rust Belt area of America where the race was lost.  Ironic because the mantra of the 1992 Bill Clinton campaign was, “it’s the economy stupid.”  It worked in 1992 and it worked in 2016.  It’s just that in 2016 a Clinton wasn’t driving that message. Instead, she ran an onslaught of negative ads trying to reinforce and increase Donald Trump’s already high unfavorable ratings. That proved a misguided approach as Trump led handily among voters who had an unfavorable opinion of both candidates (47-30%). 

It is said that politicians run negative ads “because they work.”  And they do.  But at some point – oh, let’s say, maybe when both candidates have unfavorable ratings well above 50% – swing voters are looking for a reason to vote FOR someone.  Donald Trump provided a simple, compelling message to voters unhappy with both choices. They may have been leery, but he provided action items and change. Basically, he offered a message, however simplistic.

To be fair, the Clinton team had to walk a tightrope here with the need to get out the African American vote and keep the Sanders voters in her camp, while also voicing concerns about the current state of the economy and how she would change things.  Incumbency has a lot of benefits – and perceptually Hillary was the ultimate incumbent – but unfettered messaging is not one of them.

Exit polling is pretty telling in this respect.  Like all polling, exit polling is not exact and is subject to the same challenges as pre-election polling.  But sometimes the data are pretty clear, with or without margin of error. When we look at the swing states that decided the election and the ones which Trump won, we see that in all but one of them Donald Trump was seen as being better able to handle the economy, and by significant amounts. In Ohio (+12) and Wisconsin (+11) he held net double-digit leads on being better able to handle the economy.  In Iowa it was +9, in Florida +6 and in North Carolina +6.  In Michigan the exit polling gave a slight edge to Hillary Clinton on this question but that exit survey also had Hillary Clinton ahead in their results. (Note: Exit pollsters did not ask this economic question in Pennsylvania)


Who would better handle the economy?
(Net Trump)

But, one might say, it goes to reason that states that went for Trump would naturally give him the edge on questions like this and states that went for Hillary would give her the edge.  And on the economic question that is most certainly true. But when these same voters were asked who would better handle foreign policy, respondents chose Hillary in every single swing state where Donald Trump won.  And the net Hillary number on this foreign affairs question in those Trump swing states ranged from +4% to +14%…for Hillary.


Who would better handle foreign policy?
(Net Clinton)

Hillary Clinton won the debate on foreign policy, but she lost the election on economics.


4:       Don’t ignore the ethnic changes taking place in America.

If you are a Hillary Clinton supporter finally able to sleep a little but still convinced the world is coming to an end you probably have nightmares in which a TV chyron scrolls across the screen over and over and over with the words “the white vote, the white vote, the white vote.”  And if you hear the term “older white men” one more time you’ll undoubtedly think about doing something rash and out of character.  But before you do, it would be prudent to take a look at the exit polling and get a little more insight into this “white” vote meme.

To be sure, Donald Trump did better with white voters in Wisconsin and Michigan and that seems to be what carried the day in those states.  But in Pennsylvania, Donald Trump’s net support among white voters was only 1 point better than Mitt Romney’s, certainly not enough to flip the state to the GOP column. What appears to have happened in Pennsylvania was lower turnout among African American voters. This group voted almost identically for Hillary Clinton as they did for Barack Obama. But in 2016 they only represented 10% of the overall electorate.  In 2012 they represented 13% of the electorate.  Had they represented 13% of the vote last November and cast their ballots accordingly Hillary Clinton would have received an additional 150,000+ votes in Pennsylvania.  She lost by 68,000.

In Florida it wasn’t just that Donald Trump improved his support with whites over Mitt Romney’s. Trump’s net support among white voters was 8 points higher than Mitt Romney’s. That helped, but his net support improved by double digits among African Americans, with whom he enjoyed a 15 point net improvement over Mitt Romney.

And nationally when we look at ethnicity, Donald Trump and Mitt Romney look identical among white voters.  Romney and Trump both won white voters by a net 20 points. But Trump’s net support numbers improved with African American voters (+6), Hispanics (+6) and Asians (+9) over Romney’s net support with those groups.

So what did happen with the white vote? Well, it depends. In some states the white vote behaved pretty much as it has, in some it provided the margin of victory, and in others Donald Trump saw significant dropoff in his white support. In the western states where Hispanics make up 10% or more of the population exit polling shows Donald Trump doing significantly worse among white voters than Mitt Romney did.  In California, Trump’s percentage of the white vote was 8 points lower than Mitt Romney’s.  In Arizona it was 12 points lower than Romney among whites. In Colorado it was 7 points lower and in New Mexico it was 9 points lower.  In Nevada it was a wash.  In Texas there was no exit polling done, but overall Donald Trump only won Texas by 10 percentage points while Mitt Romney won Texas by 16 points, so some falloff in the white vote there most likely took place as well.  Conventional wisdom about the “white vote” does little to provide insight into how this election actually played out.

Perhaps the biggest danger for Republicans is to presume that Trump‘s victory proves the point that Republicans can ignore the Hispanic vote.  They can’t.  The country is changing inexorably and the makeup of the electorate – especially the makeup of new voters – does not favor Republicans.  In 2012 whites made up 74% of those actually casting ballots in the Presidential election, down from 76% in 2008 and 79% in 2004.  In 2016 they made up 71% of the electorate.


Decline in White Voter Turnout


 Source: US Census Bureau, Current Population Reports 2012
* 2016 Source: Network News Exit Polls

And let’s not forget, Hillary Clinton won the popular by 2.8 million votes. Of course, the Electoral College tells a different story and if you exclude California, Donald Trump did win the popular vote.  Nonetheless the country is changing and not just California. And that will increasingly impact the Electoral College as well.

From 2012 to 2016 there was an increase of 10.7 million new voters in the United States.  Seventy percent of those voters were non-white. Many of these new voters are Hispanic.  These new Hispanic voters tend to be younger and thus, less inclined to vote. For Republicans that may provide a false sense of security. The full force of the increasing Hispanic voting population has not yet had its full impact.  They are quickly increasing their percentage of the electorate to be sure, but as they age their voting participation percentages will increase as well.

Today African Americans and whites have similar voting participation percentages (around the mid-60s for both groups). But less than half of eligible Hispanics actually show up to vote in presidential elections. Given the fact that they now represent the same percentage of the American electorate as African Americans – and will soon be a larger percentage of it – higher turnout percentages among Hispanics could at some point in the fairly near future have the profound effect on American elections Democrats were hoping for in 2016.


5:       Listen to Mark Twain

Finally, what has to be the hardest lesson of this campaign for the Clinton team is the one of lost opportunity. In retrospect there were really only three must-win states to secure a victory over Donald Trump: Michigan, Pennsylvania and Wisconsin. Those were states where public polling and focus groups showed the Trump message could appeal to some Obama/Kerry/Gore/Bill Clinton voters. Trump only needed one of these states to win or tie the Electoral College and he wound up winning all three, but each by razor thin margins.  Reportedly, Hillary Clinton never campaigned in Wisconsin. True, she campaigned in Pennsylvania and Michigan down the stretch but these states were part of, or should have been part of, a firewall that ensured her election. It appears they were no more important than anywhere else, in some cases, less so. That miscalculation will likely haunt Hillary Clinton and her team for a very long time.

Which leaves us with the timeless wisdom of Mark Twain:

“Put all your eggs in one basket…and watch that basket.”

Enjoy 2017 and get ready for 2018, it promises to be a barnburner of an off year.



<- Go Back