Poll­sters miss mark again in as­sess­ment of Trump vot­ers

Lag­gard nearly won Vir­ginia GOP pri­mary


Poll­sters had ma­jor trou­ble spot­ting the surge of sup­port for Corey Ste­wart in Vir­ginia’s Repub­li­can gover­nor’s pri­mary this week, suggest­ing they still can’t fig­ure out how to suc­cess­fully sur­vey Trump sup­port­ers.

Mr. Ste­wart, chair­man of the Prince William Board of County Su­per­vi­sors and one­time Trump cam­paign chair­man in Vir­ginia, ran as the Trump fig­ure in the race, pur­su­ing an anti-es­tab­lish­ment mes­sage that cen­tered around a tough ap­proach to il­le­gal im­mi­gra­tion.

Polling — which was strik­ingly in­fre­quent — showed Mr. Ste­wart fail­ing to crack 20 per­cent sup­port, trail­ing Ed Gille­spie, the pu­ta­tive front-runner, by as much as 27 per­cent. In the end, Mr. Gille­spie nipped Mr. Ste­wart by just 1 point, 44 per­cent to 43 per­cent. That left a lot of Trump-style Ste­wart sup­port­ers hid­den in ear­lier polls.

Michael McKenna, a Vir­ginia-based poll­ster and strate­gist, said the results should be an­other wake-up call.

“Re­searchers have to do a much bet­ter job of mak­ing sure that they are get­ting re­sponses from all kinds of rel­e­vant pop­u­la­tions,” Mr. McKenna told The Wash­ing­ton Times.

The sen­ti­ment swept through the polling com­mu­nity last year when sur­veys, while cor­rectly catch­ing Hil­lary Clin­ton’s pop­u­lar vote lead, missed Don­ald Trump’s per­for­mance in a num­ber of bat­tle­ground states.

Me­dia polls were par­tic­u­larly bad at find­ing

Trump sup­port­ers. They un­der­sold Mr. Trump’s sup­port in bat­tle­ground states he eas­ily won, such as Iowa and Ohio, and out­right whiffed on Wis­con­sin, Michi­gan and Penn­syl­va­nia by pre­dict­ing a Clin­ton vic­tory.

The poor show­ing sparked soulsearch­ing, with poll­sters de­bat­ing their short­com­ings.

Some spec­u­lated that Trump vot­ers were shy about telling poll­sters their lean­ings, while oth­ers said they botched the ex­pected turnout. That meant poll­sters were count­ing soft Clin­ton sup­port­ers who, in the end, didn’t show up.

Mr. Ste­wart said poll­sters also don’t know how to prop­erly mea­sure how en­er­gized vot­ers are on­line.

“There are a lot of peo­ple who were ac­ti­vated, who were awak­ened dur­ing the Trump cam­paign, and we still don’t know who is go­ing to show up and vote in Repub­li­can pri­maries go­ing for­ward,” he said as he sur­veyed the polls that un­der­shot his sup­port.

He said polls dur­ing most of the pri­mary sea­son were based on “bad sam­ples” and tools such as so­cial me­dia are help­ing cam­paigns tap vot­ers whom poll­sters are miss­ing.

“There was a very con­ven­tional way of de­ter­min­ing the vot­ing uni­verse, which is old-fash­ioned and out of date,” Mr. Ste­wart said. “Peo­ple are ac­ti­vated now more by so­cial me­dia, and they can be ac­ti­vated on a dime, whereas be­fore it was pre­dictable. You had the me­dia and pretty much the same peo­ple show­ing up on the left and the right. Now, so­cial me­dia is trans­form­ing ev­ery­thing.”

One of the misses this year was a Wash­ing­ton Post/Ge­orge Ma­son Univer­sity sur­vey in May that pre­dicted a blowout. It showed Mr. Gille­spie hold­ing 38 per­cent sup­port and Mr. Ste­wart well be­hind at 18 per­cent, barely ahead of Mr. Wag­ner’s 15 per­cent.

Mark Rozell, dean of the Schar School of Pol­icy and Gov­ern­ment, which teamed up with The Post on the poll, said a num­ber of fac­tors played into Mr. Ste­wart’s sur­prise show­ing.

“Low turnout, the ide­o­log­i­cal can­di­date with more in­tense sup­port over­per­forms. This hap­pens in low turnout, off-cy­cle cam­paigns,” Mr. Rozell said.

He said the pri­mary vote sug­gests Mr. Gille­spie will have a tough time in the gen­eral elec­tion.

One poll got the Repub­li­can race al­most ex­actly right. Change Re­search’s June 8-10 poll showed Mr. Ste­wart with a 42-41 lead over Mr. Gille­spie, with Mr. Ste­wart’s sup­port­ers more com­mit­ted to show­ing up.

But Change Re­search’s suc­cess was cou­pled with a to­tal botch in the Demo­cratic pri­mary, where the or­ga­ni­za­tion had for­mer Rep. Tom Per­riello eas­ily win­ning. In­stead Lt. Gov. Ralph Northam tri­umphed handily.

Pat Reilly, a spokes­woman for the firm, said re­searchers build their sur­vey pool on­line, chiefly from Face­book. That al­lowed them to gather a much broader sam­ple size than any of the tra­di­tional phone line sur­veys, which have a huge no-re­sponse rate.

“Ste­wart’s core sup­port was among Don­ald Trump’s big­gest fans — those who rated Trump a 9 or 10 out of 10 — where he led over Gille­spie by nearly 20 points. Be­cause we were polling right up un­til the clos­ing days of the elec­tion, we were see­ing that sen­ti­ment on­line,” the spokes­woman said.

Larry Sa­bato, a po­lit­i­cal sci­en­tist at the Univer­sity of Vir­ginia, said the pri­mary polls in Vir­ginia this year “were gen­er­ally aw­ful.”

“The polls gen­er­ally created two themes that hard­ened into con­ven­tional wis­dom. First, Gille­spie was the ru­n­away fa­vorite and the GOP pri­mary wasn’t much of a race. Sec­ond, Northam and Per­riello were in a death strug­gle in a very close pri­mary that Per­riello was on the verge of win­ning. Both propo­si­tions were ab­so­lutely wrong,” he said.

The dan­gers of miss­ing polling tar­gets are myr­iad, but over­look­ing Trump vot­ers could skew per­cep­tions of the pres­i­dent’s job ap­proval and pop­u­lar­ity or, in this case, the un­pop­u­lar­ity of House Repub­li­cans’ health care bill.

“Re­searchers need to be hon­est with them­selves. They are viewed by some sig­nif­i­cant chunk of peo­ple — and usu­ally the same kind of peo­ple — as part of the prob­lem,” Mr. McKenna said. “That may or may not be de­served, but it is the re­al­ity. That af­fects re­sponse rates, drops, and all kinds of prob­lems in sur­vey re­search in pol­i­tics.”

David Pa­le­ol­o­gos, who runs the Suf­folk Univer­sity Poll, which didn’t sur­vey in Vir­ginia, said miss­ing vot­ers is a prob­lem but not a ma­jor one.

He said state pri­maries, such as the Gille­spie-Ste­wart-Wag­ner race, do pose pit­falls, with a large num­ber of un­de­cided vot­ers and more than a bi­nary choice be­tween two can­di­dates. A lot of Ste­wart vot­ers could have been hiding in other cat­e­gories.

“If you have a higher un­de­cided and you have a can­di­date like Trump or a sup­porter of Trump’s, it’s easy to mask your sup­port in the un­de­cided col­umn,” he said.

Mr. Pa­le­ol­o­gos, though, said that is prob­a­bly not as big an is­sue for look­ing at Mr. Trump’s job ap­proval rat­ings, where the ques­tion is straight­for­ward and there aren’t a lot of ways to re­spond.

“When you have a bi­nary ques­tion where it’s ap­prove or dis­ap­prove and the sam­ple frame is cor­rect and all the other boxes are checked on the de­sign, poll­sters are pretty close to the out­come,” he said.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.