Traditionally, the real presidential election season every four years in America begins with the Labor Day holiday. This election year Labor Day fell on Monday September 5. The slower summer vacation time is past, and voter interest in serious political matters intensifies.
This election season has been unusual in so many respects, it is hard to see how interest could get much more intense, but there it is. In any case, American voters had 69 days from Labor day to make their final decisions. As of today, October 5, there are 34 days until Election Day.
Following polls in the popular (including social) media has become almost as much a part of the process as voting on the day. In the end, of course, there are no prizes for winning polls on any day except the final grand prize on November 8th. Nonetheless, Americans are now quite used to keeping tabs on score estimates as the time runs out.
We have said before there is no one poll, or polling organization, or sampling methodology which is uniquely accurate to predict the final outcome. Truly, it ain’t over til it’s over.
However, looking at a collection of public polls, which individually are conducted using a defensible set of statistical methods for sample selection, can provide a reasonable picture of the overall state of the race.
Both parties do their own private polling, using sophisticated modeling projections to get even finer tuned results, for determining where to spend ad dollars or directing paid staff time for example. These methods and efforts are very closely guarded party secrets and kept close at least until after election day, lest the other side gain some advantage from reverse engineering. They are, in one sense, the Crown Jewels of election data management. They control allocation of resources for all federal election contests, not just for President. So, your average citizen doesn’t get to see these secret results ahead of time. The model results can be distorted simply because the parties are trying so hard to produce extremely detailed forecasts from only fragmentary data, and significant error can be introduced by pushing too far ahead with fancy statistical methods. This is the ultimate electoral inside baseball.
So, in fact, the general trends in the aggregate of well conducted public polling may thus be the best indicator for predicting election results. A Poll of Polls, if you will.
That is the kind of data we all can see, thanks to the Herculean efforts of a number of public service minded organizations. As a bonus, they usually offer their results to the general public free of charge. They are nerd heroes of 21st century American democracy.
One of the respected sources of near real time summary polling data is provided by RealClearPolitics. Their website, with lots of other stuff is found at
The parent organization is considered to have a somewhat conservative lean in general, so it should be a pretty fair indicator of voter support for Trump and the Republicans this election cycle, as measured by pollsters.
Real Clear Poll of Polls (Adapted)
There is a wealth of data available from the RCP website. Too much, in fact, for a mere mortal to take in with a simple visual sweep. To provide a summary perspective guide, we have adapted their reported data in the following way.
On a weekly basis, starting with Labor Day weekend, we have included every “General Election” Poll reported by RCP, and whether it was ‘won’ by either Trump or Clinton, or a tie. All General Election polls, whether they were simple two-way contests, or included third party candidates (Johnson and Stein) are added. Results are combined and listed day by day, along with the Daily Net Margin for the highest ranking candidate that day. For some days, no General Election Polls are reported, and those results are left blank. Each week, beginning on Saturday and ending on Friday, the polls for the week are summed, with wins for each candidate, and the Average Weekly Net Margin is computed by taking the total net margin for the leading candidate, divided by the total number of polls reported that week.
So, if Trump were ahead by 10 margin points in a weekly collection of 5 polls he would have a Weekly Net Margin of 2.0 points. So far, he hasn’t led in a Weekly Summary since Labor Day. There are still five weeks left in the campaign, but time is running short. Two debates have already occurred, and the post-convention bounces for both parties have come and gone.
Daily and Weekly Poll of Poll Results
Where are we, polling wise? Here goes:
There are only preliminary results for the fifth week, since the Thursday and Friday results are not in yet. For Week 5 (10/01-10/07), Trump has won or tied 3 out of 14 polls (21.4%). Clinton has a Weekly Net Margin of 2.8 points over Trump in 14 polls.
For Week 4 (09/24-09/30), Trump won or tied 4 out of 23 polls (17.4%). Clinton had a Weekly Net Margin of 2.6 points over Trump in 23 polls.
For Week 3 (09/17-09/23), Trump won or tied 4 out of 14 polls (28.6%). Clinton had a Weekly Net Margin of 2.7 points over Trump in 14 polls.
For Week 2 (09/10-09/16), Trump won or tied 5 out of 17 polls (29.4%). Clinton had a Weekly Net Margin of 2.0 points over Trump in 17 polls.
For Week 1 (09/03-09/09), Trump won or tied 2 out of 10 polls (20.0%). Clinton had a Weekly Net Margin of 2.1 points over Trump in 10 polls.
In the first five weeks of the 2016 Presidential Election Super Bowl (one-half of the contest), there have been 78 polls scored, of which Trump has won of tied18 contests (23.1%). His winning-tied percentages, week by week, range from 17%-29%, a fairly narrow and consistent range. He has not won any week outright, and Clinton’s Average Weekly Net Margins range from 2.0-2.8 points per week, also a fairly narrow and consistent range.
Trump has won or tied the combined poll collection on 4 days out of 21 days for which any General Election Polls were reported. On those 4 days with Trump’s 4 poll wins, his average winning net margin was 0.75 points.
No polls were reported on 12 days out of the 33 days (36.3%) elapsed since the counts began. This means there is at least one new General Election poll for two days out of every three expected, including weekends.
We have seen 78 polls reported over 33 days, or an average of 2.4 polls per day. We should therefore expect about another 85 polls until the election (not including election day itself).
Summary
Based on this RCP Poll of Polls analysis, the good news for Republicans is that the polling results trend would predict a fairly close election overall. There has not been any real significant change over the last five weeks since Labor Day, except perhaps a small (about 0.5 point decline) in their fortunes after the first Presidential debate.
The less welcome news is that Trump has been behind for all five weeks, and has not shown the earnestly hoped for and welcome surge of improvement prayed about during the last month.
There is still time left, but the election sands are running out far too fast for comfort. Based on the best daily public polling available to date, if Trump is to prevail, it will be by only the thinnest of margins, in a breath-stopping squeaker.
Trump’s relatively lackluster fundraising (compared to Romney in 2012 and Clinton in 2016), and the remaining questions about the Republican party’s financial resources to staff and coordinate volunteer offices and juice their own final get out the vote ground game, may both come back to haunt responsible officials and Trump’s campaign managers, despite Trump’s ever sunny rally portrayal that everything is well in hand.
Maybe he has some secret sauce no one else has ever thought of.
Maybe, but I think I’ll stick to McDonalds.
Thanks again to all the men and women who produce RealClearPolitics polling data. Yu are American patriots. We salute you.