Contrary to popular belief, the purpose of polling is to explain political events not predict them. Polls are snapshots not motion pictures. Like everything else, some polls are good and some are bad. It takes a discriminating analyst to tell the difference.
On the plus side, the national polls were remarkably accurate measuring the outcome of the national popular vote during the 2016 presidential election. The Real Clear Politics average of the last round of national polls had Hillary Clinton leading Donald Trump by +2.1 percent. The former first lady actually won the popular vote by 2.1 percent. Not bad for polls of 1,500 voters with an actual turnout of more than 130 million Americans.
{mosads}The problem is that victory in the presidential race is determined by the Electoral College, not the popular vote. Networks and voters would be better served with more polling in battleground states, and less nationally, in 2020.
Hillary Clinton’s edge in the popular vote obviously didn’t translate into an Electoral College majority. The reality is, the national media puts more emphasis on and money into national polls than they do battleground state polling. For this reason, late trends in swing states went unnoticed. Donald Trump’s late surge in the industrial Midwest states punched his ticket to the White House.
One factor to consider in polling is margin of error. NBC News recently did a poll of voters for the Tennessee Senate race: Democratic candidate former Gov. Phil Bredesen led his GOP opponent, Rep. Marsha Blackburn, 46 percent to 44 percent. The survey has a margin of error of plus or minus 5 percent. The 2 percent Bredensen edge does not mean he is ahead. His support only means there is a 95 percent probability that the Democrat’s vote is somewhere between 51 percent and 41 percent, while Blackburn’s support is somewhere in the range of 49 percent to 36 percent. Who’s really ahead is anybody’s guess.
The big problem is, it is much more difficult to poll in primaries than in general elections. Turnout is much lower, so it’s difficult for pollsters to figure out who will vote and who won’t.
Furthermore, many primary voters decide how to vote in the last few days before the primary after the phone calls have already been made for the last preprimary polls. Primary voters wait until the last minute to make a decision because it’s easier to choose between a Democratic candidate and a GOP candidate than it is for voters to choose between two Democrats or two Republicans.
Two of the biggest congressional primary upsets this year were the victories of Alexandria Ocasio-Cortez in New York and Ayanna Pressley in Massachusetts over heavily favored veteran Democratic incumbents.
The last public poll I saw in the Pressley race in the 7th Congressional District of Massachusetts was conducted a month before the Sept. 4 primary by WBUR, a public radio station in Boston. The WBUR survey gave her primary opponent, Rep. Michael Capuano (D-Mass.), a 44 percent to 31 percent edge.
A 13 percent lead may look solid to some observers. But Pressley won the primary 59 percent to 41 percent. What happened? If they had looked into the details (“internals”) of the poll, political junkies may have come to a different conclusion.
First, a quarter of the voters were still undecided. Experience tells me that many of those people waited until the last minute to make their decision. Undecided voters usually gravitate away from the incumbent towards the challenger.
The context may be more important than the head-to-head matchup. Most primary voters (61 percent) did not think Democrats in Congress had been aggressive enough challenging Trump. This argument was a key part of Pressley’s message against the incumbent.
Then there’s the question of name recognition. Capuano had a 13 percent lead but he also had a 15 percent edge in name recognition over his challenger. This differential suggested that Pressley could close the gap as voters came to know her better when they focused more on the candidates during the last two weeks of the primary.
Finally, there’s the issue of gender. In Massachusetts, as in most places, women are more likely to be Democrats than men. Accordingly, the radio station’s voter screen produced a sample that was overwhelmingly female (57 percent female and 43 percent male). The presence of such a large feminist voter bloc in a year when female Democratic candidates are doing so well indicated that there was less to Capuano’s lead than met the eye.
You may see hundreds of polls over the next two months before the midterms, and thousands before the next presidential election. Keep these factors in mind.
Pay attention to the screen for the survey. Did the pollster survey anybody 18 years of age or over, registered voters or likely voters? The screen makes a big difference in the analysis of the results.
All polls and pollsters are not equal. Be aware of the pedigree of the survey before you evaluate the results.
Finally, look under the hood when you see a poll. The head-to-head matchups are often less meaningful than the rest of the data in the poll which provides the context for the bottom line.
Brad Bannon is a Democratic pollster and CEO of Bannon Communications Research. He is also a senior adviser to, and editor of, the blog at MyTiller.com, a social media network for politics.