No one is more perplexed by Donald Trump than the leaders of the Republican Party. On May 5, after Trump effectively clinched the Republican nomination, Paul Ryan famously asked:
Does he share our values and our principles on limited government, the proper role of the executive, adherence to the Constitution?
As the speaker of the Republican-dominated House, Ryan could have posed a harder question: Do Republican voters “share our values and our principles”?
The answer to this question, based at least on the 10.7 million votes cast for Trump in Republican primaries and caucuses so far, is “no.”
But that’s not all. There is also strong evidence that most traditional public opinion surveys inadvertently hide a segment of Trump’s supporters. Many voters are reluctant to admit to a live interviewer that they back a candidate who has adopted such divisive positions.
In matchups between Trump and Hillary Clinton, Trump does much better in polls conducted online, in which respondents click their answers on a computer screen, rather than in person-to-person landline and cellphone surveys.
An aggregation by RealClearPolitics of 10 recent telephone polls gives Clinton a nine-point lead over Trump. In contrast, the combined results for the YouGov and Morning Consult polls, which rely on online surveys, place Clinton’s lead at four points.
Why is this important? Because an online survey, whatever other flaws it might have, resembles an anonymous voting booth far more than what you tell a pollster does.
In a May 2015 report, Pew Research analyzed the differences between results derived from telephone polling and those from online Internet polling. Pew determined that the biggest differences in answers elicited via these two survey modes were on questions on which social desirability bias — that is, “the desire of respondents to avoid embarrassment and project a favorable image to others” — played a role.
In a detailed analysis of phone versus online polling in Republican primaries, Kyle A. Dropp, the executive director of polling and data science at Morning Consult, writes:
Trump’s advantage in online polls compared with live telephone polling is eight or nine percentage points among likely voters.
This difference, Dropp notes, is driven largely by more educated voters — those who would be most concerned with “social desirability.”
These findings suggest that Trump will head into the general election with support from voters who are reluctant to admit their preferences to a live person in a phone survey, but who may well be inclined to cast a ballot for Trump on Election Day.
Conflicting online and phone poll findings in response to Trump’s call on Dec. 7, 2015 — five days after two terrorists killed 14 people in San Bernardino, Calif. — for a “total and complete shutdown of Muslims entering the United States until our country’s representatives can figure out what is going on” demonstrate the difficulty gauging Trump’s strength.
At the same time, YouGov, operating online, found substantial and growing support for Trump’s proposal, with a plurality, 45-41, in support. When YouGov repeated the question on March 24-25 — just after the terrorist attacks in Brussels — support had grown to 51-40.
This December-to-March shift was strongest among independent voters, who increased their support from 42-37 in favor of the ban to 62-37 in favor. Similarly, a March 29 Morning Consult online poll found majority support for the ban, 50-38, with voters who identified themselves as independents favoring Trump’s plan 49-36.
I asked a number of experts about the disparity between online and phone polls. All of them — Alan Abramowitz, John Sides, Michael Tesler and Lynn Vavreck, political scientists who specialize in the analysis of poll data — agreed that in the case of highly contentious issues, respondents can be more willing to express their real views anonymously, to a computer rather than to a human.
Kathy Francovic, the former CBS polling director who now works for YouGov, told me that “it’s easier to express potentially ‘unacceptable’ responses on a screen than it is to give them to a person.” Douglas Rivers, a political scientist at Stanford and the chief scientist for YouGov, agreed, noting in an email that stronger support in online polls for a ban on Muslims
may be due to social desirability bias — people are more willing to express support for this privately than when asked by someone else.
Needless to say, Trump has expressed confrontational views on a number of fronts. He claims that as president he will impose harsh tariffs on imports from China, suspend Muslim immigration, deport 11 million immigrants and build an $8 billion wall that Mexico will pay for.
Taken together, these positions have provided a foundation for the strong correlation between support for Trump and white ethnocentrism and white racial resentment.
One method of ranking whites on ethnocentrism is to measure the degree to which they believe Caucasians are more trustworthy, intelligent, industrious and less violent than African-Americans, Hispanics and other minorities. These are the kinds of questions that prompt certain respondents in phone surveys to mask their views and provide socially acceptable answers instead.
The accompanying chart, which uses data provided to The Times by Marc Hetherington and Drew Engelhardt, political scientists at Vanderbilt, shows that white Republicans are the most ethnocentric of all voters, but also that there are substantial numbers of ethnocentric white Democrats and white independents.
This suggests that Trump could potentially find significant levels of support not only among Republican voters, but also among white Democrats and independents.
Now that Trump appears to have the Republican nomination in hand, the question becomes: Can he capitalize on racial resentment among Democrats and independents in the general election?
Perhaps not surprisingly, Hetherington and Engelhardt found that racial resentment follows a similar pattern to the expression of white ethnocentrism. It is highest among Republicans, but it is also present among Democrats and independents. The second chart derived from their data shows that in rankings of racial resentment, more than half of white Republicans, 58 percent, fall into the top four most resentful categories.
What should prove worrisome for Democrats is that 42 percent of white independents also fall into the four most resentful categories, as do 22 percent of white Democrats.
Even polls using traditional phone survey methods find notable support for issues high on Trump’s agenda. You can see this, for example, in attitudes toward the Chinese, Muslims and Mexicans — all of whom Trump has demonized.
Anger toward China appears to offer fertile ground for Trump in the general election. In its 2015 American Values Survey, the Public Religion Research Institute asked how responsible China is for American “economic problems.” Solid majorities of Democrats (70 percent), independents (72 percent) and Republicans (80 percent) said China is “very” or “somewhat” responsible.
Or take another Trump theme: Islam. The P.R.R.I. values survey asked whether they agreed or disagreed with the statement “the values of Islam are at odds with American values and way of life.” Among all voters, 56 percent said that they agreed. Republican were strongest at 76 percent, but independents came in at 57 percent, with Democrats trailing at a still robust 43 percent.
The Polling Report, an aggregation of public opinion surveys, presents data on immigration from multiple sources. On a basic question — what should happen to the 11 million undocumented men, women and children now living within the borders of the United States — most traditional surveys show strong support for finding ways to legalize the status of those who have not committed crimes and have paid taxes.
A March 2016 Pew study found, for example, that voters preferred allowing undocumented immigrants to stay in the United States over attempting to deport them by 74-25. It also found that a majority said immigrants strengthen the country (as opposed to adding a burden), 57-35. These are not good numbers for Trump.
But poll results (irrespective of whether questions are posed online or by phone) can change quite a bit depending on their exact wording, the specific issues addressed and even the placement of a query in a series of questions.
For example, a September 2015 Pew survey asked a related but different set of questions about immigrants that produced results more favorable to Trump’s prospects. Voters reported (50-28) that they believe that immigrants damage the economy (as opposed to making it better), with a fifth saying that immigrants don’t have much effect. Voters also reported that they think that immigrants make crime worse rather than better (50-7), with 41 percent saying that they don’t have much effect.
There are a few conclusions to be drawn.
First, the way Trump has positioned himself outside of the traditional boundaries of politics will make it unusually difficult to gauge public support for him and for many of his positions.
Second, the allegiance of many white Democrats and independents is difficult to predict — cross-pressured as they are by the conflict between unsavory Trump positions they are drawn to and conscience or compunction. The ambivalence of many Republicans toward Trump as their party’s brazenly defiant nominee will further compound the volatility of the electorate.
Finally, the simple fact that Trump has beaten the odds so far means that it is not beyond the realm of possibility that he could beat them again. If he does take the White House, much, if not all, of his margin of victory will come from voters too ashamed to acknowledge publicly how they intend to cast their vote.