I would have titled this piece “Donald about to lose,” on account of Trump’s outrageous antics, had I not seen the ABC News report on its 10/22/16 poll, that 56 percent of Hillary Clinton’s backers say they are voting mainly to support her rather than to oppose Trump. In contrast, 54 percent of Trump’s voters are mainly motivated by opposition to Clinton, not support for him.
The 56 percent are the affirmatives among those saying they would vote for Clinton “if the election were held today.” Trump’s affirmatives, on the other hand, are a minority 41 percent of those planning to vote for him.
From last June, Clinton’s affirmatives are up by 8, while Trump’s affirmatives are down by 3. Donald’s negative campaigning is losing to Hillary’s positive campaigning—as she said, “when they go low, we go high.”
After the August conventions, Clinton’s chance of winning the electoral college vote—which is what counts—rose to 80 percent. It dropped back to below 60 in September (“Hillary holds on,” Opinion, 9/24/16), rose again to the 80s after the presidential debates, and slipped only a little recently.
Last Wednesday’s odds in Nate Silver’s fivethirtyeight.com (“538” for short; “polls-only” forecast) were 85 percent for Clinton and 15 percent for Trump; yesterday’s odds were 82 to 18, or still quite strong for Clinton.
The way to predict a US presidential election is to use the statewide polls, not the nationwide ones. I used to go to the subscription-only state poll section of PollingReport.com. Knowing now that PollingReport.com is one of 538’s sources saves me its fee of $70 for four months.
The merit of the 538 system is that it uses ALL the polls, not just a selection of them—even the “best” poll goes wrong sometimes. 538 has rated over 300 pollsters, on the basis of races called correctly, simple average error, and other indicators. It takes sample size, length of time before the election, and number of competing polls into consideration. The grades are A-B-C-D-F, for good-satisfactory-passable-conditional-failure (“FiveThirtyEight’s Pollster Ratings,” and “The State of the Polls, 2016”). The bulk of the grades are Bs—i.e., satisfactory for the polling industry as a whole. 538 excludes F-rated polls from its forecast.
Because only half of American households have land lines now, 538 indicates which pollsters include cellphones in the sampling base. It identifies those that poll by internet—their forecasts are competent but not as accurate as those based on voice calls. It shows that professional affiliations (with the National Council for Public Polling, the American Association for Public Opinion Research, and the Roper Center polling archive) matter for polling quality.
Among the major media pollsters, 538 rates ABC/Washington Post at A-plus. CBS/New York Times, the Los Angeles Times, CNN/Opinion Research, and NBC/Wall Street Journal are at A-minus. Survey USA is at A. Fox News is also at A now, but was at B before 2011.
Several academic pollsters do well: Monmouth University (New Jersey) is at A-plus; Marist College (New York), Siena College (New York), and University of Cincinnati are at A; Quinnipiac University (Connecticut) is at A-minus. They are small but competent. (Large survey doers like the University of Chicago and the University of Michigan aren’t rated because they don’t forecast elections.)
Trump’s charge that opinion polls are rigged is just a sign that he won’t be a good sport when he loses.
Contact mahar.mangahas@sws.org.ph.