About two weeks ago I was asked the following question, “What is with these polls, they’re all over the place?” A follow-up question was equally interesting, “Why can they never quite get to a full 100%?” For those wondering if the polls are believable and/or wondering why we are seeing such extreme variance, the following piece is for you.
My advice: do not believe the polls you are reading. Why? Because pollsters are using a potentially outdated methodology.
I use the term “potentially outdated” because we are simply unsure. What we know is that (A) we are in a change election [i.e., the party at power is seeking to keep the White House for someone other than the incumbent] and (B) Donald Trump is disrupting the political narrative and balance of power from within his party. Consequently, unlike Ronald Reagan who faced a deeply unpopular incumbent, Trump is facing a distrusted successor. Conversely, like Ronald Reagan, Trump represents a Republican Party movement that began about twenty-years earlier within the GOP and has now emerged as the possible future face of the Republican Party.
Change from within
In 1980, Ronald Reagan was the torch-bearer of a decidedly anti-Washington, Western-states libertarian movement that began with Barry Goldwater. It is hard to remember in 2016, but until 1960, the Republican Party was the party of big government – just not the party of big government social programs. That libertarian-esque Goldwater-Reagan wing was marginalized until the revolt from within the party in 1976 and again in 1980. For Trump, the paleo-conservatives are now offering a similar transformative opportunity.
Begun in 1992, Pat Buchanan’s paleo-conservative message of blending Nationalism with Libertarian-esque personal independence was a bit too much for Republicans to handle. But the message resonated with a key constituency of voters. Over time, the failures and betrayals of so-called conservatives and the inability of libertarian leaning Republicans to galvanize a major bloc from within the party led to a paleo-conservative resurgence. It just needed the right spokesman. In 2016, that became Donald Trump.
Both with Reagan and Trump, the disdain for Washington by the voter was an overwhelming factor in the growth of their popularity. In 1980, Washington, DC was incompetent. In 2016, Washington, DC is deeply corrupt. Reminders of Washington, DC’s failures – tangible and moral – abound on Main Street today, just as they did in 1980. That makes it very possible for a leader like Reagan, and possibly Trump, to gain voters from the other party because he is offering a new option heretofore unseen outside of the most intimate Republican circles.
Correspondingly, it is hard to predict through traditional political polling whether or not an accurate measure of the voting constituency can be taken because assumptions of voting behavior – Republican, Democrat, Independent or otherwise – are impossible to assign with certainty. Pollsters use those past voting assumptions to complete their analyses. In so doing, they could be tracking the wrong trend data because there is a good chance that Donald Trump will take a greater proportion of Labor Democrats than previous Republican candidates and/or Clinton may take a greater number of so-called “moderate” Republicans.
Let me give you an example: the latest NBC News/Wall Street Journal Poll.
Over the Columbus Day weekend (October 8th – 9th) the news media reported that Clinton now led Trump by eleven percentage points (46% to 35%) in a four-way race and fourteen percentage points in a two-way race. The Left of course was giddy with itself. News organizations raced to display the results of approximately five-hundred people surveyed.
Interestingly, however, the usual third party contributor to NBC and WSJ is Marist University. Marist either refused to play a part in the unscientific, political hit job of NBC/WSJ or simply did not play a role. Either way, the methodology used by not only NBC and WSJ, but other pollsters betrays an overwhelming lack of appreciation for the nature of this historic campaign.
Delving deeper into the NBC/WSJ poll – as with others – we see lopsided Democrat participation in these polls. Whether or not that is a function of the use of landlines versus cellular polling is hard to discern. However, registered Democrats and Republicans are nearly statistically even (roughly D 31% and R 29%). Thus, over-sampling Democrats will likely skew the results, as they did in the NBC/WSJ poll cited above.
It’s a Weight Problem
But over-sampling Democrats is NOT the problem with this year’s polling. Sure, it could lead to a two-point difference in the results, but it may actually FAVOR Trump. Therein lies the real problem with the polls. The real issue is basing statistical variance on past voting assumptions used to weigh the responses of the voters.
Marist, for example, states the following within its methodology statements and similar methods are used by nearly every other polling organization:
“… After the interviews were completed, the two samples were combined and balanced to reflect the 2013 American Community Survey 1-year estimates for age, gender, income, race, and region. Results are statistically significant within ±2.7 percentage points… The results for this subset are statistically significant within ±3.0 percentage points…likely voters defined by a probability turnout model which determines the likelihood respondents will participate in the 2016 Presidential Election based upon their chance of vote, interest in the election, and past election participation…”
(Bold was added by me for emphasis)
Note that Marist is basing its assumptions on 2012 voting patterns. So is everyone else. THAT is the real problem with the polls in 2016.
Consider the following facts about 2012 and then ask yourself if 2016 sounds remotely similar.
In 2012, White voters as a group voted as a smaller percentage of their historic turn-out than ever before. That is not a function of their smaller percentage of the electorate than the past. In fact, in 2012 there were 1.1 million MORE White voters than 2008. Instead, they seemed to just disappear from the voting booths.
By contrast, in 2012, Black voters turned out to vote in historically higher numbers than their traditional turn-out.
Additionally, according to the Bipartisan Policy Center, overall turn-out in the 2012 election seemed to drop, except in states within which Senate races were contested. In those states within which an active Senate race was contested, the incumbent Senator’s political party turned out in higher numbers everywhere except in Massachusetts (Warren v Brown).
Keeping all of that in mind, using the 2012 election results as a basis to gage the 2016 election possibilities seems downright false.
You would have to assume that (1) Trump’s crowds are LESS enthusiastic than Clinton’s (e.g., Romney v Obama), (2) White voters will continue to sit out the Presidential election, (3) Black voters will vote for Clinton with the same level of enthusiasm as they had for Obama, (4) there are less Republican Senators guarding their seats than Democrats, and (5) Senate Republicans will actively denounce Trump during their own re-election “get-out-the-vote” efforts.
Of the five assumptions, only the fifth is possible given the latest controversy over leaked audio of Trump from 2005. Of course, that comes with an extremely dangerous down side for the Republican Senators. As studied in the 2012 Warren v Brown race in Massachusetts, Brown’s attempts to keep the focus on himself versus joining the top of the ticket in 2012 cost the Senator potential Republican and split-ticket voters. Otherwise, the other four scenarios in 2012 are highly unlikely in 2016.
Trump’s enthusiasm gap over Clinton is enormous.
White voters, especially White Labor, seem genuinely interested in this year’s election.
Black voters are unlikely to see this election as a referendum on Obama as they did in 2012.
There are simply more Republican Senators defending their current seats.
Cumulatively, using 2012 statistical voting patterns as a basis of weighing the voting tallies of polling participants makes very little sense. But that is what they are doing. Thus, when Marist says that it is adjusting its assumptions based on 2013 data, and assuming a statistically significant variance of 2.7% within those same assumptions, they are likely deeply underestimating the allure of Trump to some Democrat constituencies and conversely, the disdain felt by many toward Hillary Clinton in the Republican Party.
What’s with the gap?
It will also explain the gap that is often manifested between Trump and Clinton in two-way polling. Notice that the two never really meet to a full 100%. Usually it is Clinton 45% and Trump about 40%. That would indicate that 15% of the electorate remains undecided. Ignoring for a moment that traditionally, three-quarters of undecided voters in a change election year vote for the challenger. The greater question is “WHY” there remains such a gap so late in the election cycle?
Again, it goes back to statistical weight. Groups like Marist are undervaluing Trump support based on Romney support. That absurd connection is making it impossible for the pollsters to find a 50/50 balance or something closer to 100% than that which we should be experiencing in 2016. In essence, they are undervaluing Trump by the same statistical variance they highlight in their own methodology statement. Trump probably has about three percentage points more than is being reflected in these polls – but NOT necessarily at Clinton’s expense.
Or, in other words, Clinton may have 45% after all, but Trump likely has 43% – or maybe more – which is not reflected in polling due to possibly obsolete models.
Are there ANY trustworthy polls?
So, is there any believable polling outfit still out there. Yes! The Leftist Los Angeles Times and the University of Southern California (USC).
No one can accuse either of pro-Republican or pro-Trump bias.
Since the beginning of the national election, the LA Times has kept the race about 2 percentage points apart, with Trump typically in the lead. Why would that be the case? Because unlike the phenomenon I just described, the LA Times and USC have chosen NOT to assume that 2016 will mirror 2012.
Rather, they have embraced the crazy concept of actually taking the voters at their word when they describe their voting history and their enthusiasm.
Going back to that last Republican change election I described in 1980, Reagan was down nine percentage points at this same time of the election in a three-way race (Anderson and Carter). Why? Because polling firms in 1980 used the same assumptions on voting patterns as those being used today. They used 1976 election modeling.
I suggest they stop doing that in 2016.