One Reason Trump Does Worse in Live-Interview Polls? Embarrassment

Don­ald Trump isn’t shy about his poll res­ults. It took less than five seconds of scrolling through his Twit­ter feed last week—I timed it—to find the latest tweet cel­eb­rat­ing good num­bers (“New Re­u­ters Poll - thank you!).

But these days, his suc­cess of­ten de­pends on the type of poll that’s cited. Trump rates very well in In­ter­net- or rob­ocall-powered polls, in which he leads the Re­pub­lic­an field in Iowa by at least 4 points, ac­cord­ing to Huff­Post Poll­ster. But if you con­sider only live-in­ter­view polls—in which ac­tu­al people speak with re­spond­ents over the phone—he’s cur­rently trail­ing Ted Cruz.

It’s a vari­ation of what poll­sters call the “mode ef­fect,” in which people give dif­fer­ent an­swers de­pend­ing on how they’re polled. People have writ­ten about this in re­la­tion to Trump. But an ex­plan­a­tion hasn’t come eas­ily. The most com­mon the­ory: A siz­able per­cent­age of poll re­spond­ents, though will­ing to punch a phone key to say they sup­port Trump, are still too em­bar­rassed to ac­tu­ally tell an­oth­er hu­man be­ing.

Last week, re­search firm Morn­ing Con­sult put this hy­po­thes­is to the test. Re­cruit­ing 2,397 re­gistered Re­pub­lic­ans and Re­pub­lic­an-lean­ing voters on­line, the com­pany split the sample in­to thirds—send­ing one group to an­swer elec­tion ques­tions on a web­site, an­oth­er to an auto­mated in­ter­act­ive voice-re­sponse phone line, and the rest to a call cen­ter staffed by live in­ter­view­ers.

Crit­ic­ally, this sur­vey drew its re­spond­ents from the same gen­er­al pool, which had all answered demo­graph­ic ques­tions be­fore­hand. The only vari­ation was the mode of in­ter­view.

The find­ings, re­leased Monday: The Trump mode ef­fect is def­in­itely real. Just over 38 per­cent of people who answered via a Web form said they sup­por­ted Trump, com­pared to 32 per­cent of their peers who spoke to a call-cen­ter em­ploy­ee—a 6-per­cent­age-point gap. But that gap among col­lege-edu­cated re­spond­ents widened to 9 per­cent­age points.

Morn­ing Con­sult ac­know­ledges its ana­lys­is has flaws. Re­search­ers re­cruited poll re­spond­ents on­line, po­ten­tially skew­ing the sample young­er, and they com­pensated them. And some par­ti­cipants may have been turned off by the second step of call­ing an in­ter­view­er or filling out an­oth­er sur­vey, drop­ping out of the pro­cess.

But the mode ef­fect has been stud­ied be­fore. Earli­er this year, Pew Re­search Cen­ter found that people polled over the phone by live in­ter­view­ers were far more likely to say they were sat­is­fied with their fam­ily life, or to agree that gays and les­bi­ans face dis­crim­in­a­tion. Con­versely, In­ter­net re­spond­ents were more likely to voice an un­fa­vor­able opin­ion of Michelle Obama or Sarah Pal­in.

From the re­port:

The so­cial in­ter­ac­tion in­her­ent in a tele­phone or in-per­son in­ter­view may also ex­ert subtle pres­sures on re­spond­ents that af­fect how they an­swer ques­tions. Re­spond­ents may feel a need to present them­selves in a more pos­it­ive light to an in­ter­view­er, lead­ing to an over­state­ment of so­cially de­sir­able be­ha­vi­ors and at­ti­tudes and an un­der­state­ment of opin­ions and be­ha­vi­ors they fear would eli­cit dis­ap­prov­al from an­oth­er per­son.

Pew found the largest mode dif­fer­ences on ques­tions in which “so­cial de­sirab­il­ity” was most likely to play a role, such as opin­ions on pub­lic fig­ures or per­son­al is­sues. (They found al­most no dif­fer­ence with fact-based ques­tions, such as wheth­er the re­spond­ent has a driver’s li­cense or a pass­port.) This sup­ports Dropp’s em­bar­rass­ment the­ory—no one wants to tell a stranger that they’re miser­able, or that they sup­port a man many con­sider a buf­foon.

What does this mean for Trump? It’s hard to say. His suc­cess on­line could in­dic­ate deep­er sup­port for his can­did­acy than polls cur­rently sug­gest, as Henry Olsen re­cently wrote. But con­sider the first real test for Trump:  the Iowa caucuses, where res­id­ents meet at pub­lic build­ings and plead their can­did­ates’ cases be­fore cast­ing a secret bal­lot. If some Iow­ans are already too em­bar­rassed to share their true thoughts to a call-cen­ter op­er­at­or, ima­gine how they’ll feel telling their neigh­bors.