What went wrong in this year’s pres­i­den­tial polls?

Daily Local News (West Chester, PA) - - NEWS - By Emily Swan­son and Thomas Beaumont

Don­ald Trump’s vic­tory came as a sur­prise to many Amer­i­cans, the na­tion’s poll­sters most of all.

Head­ing into Elec­tion Day, most na­tional sur­veys over­stated what will likely be a nar­row pop­u­lar vote ad­van­tage for Hil­lary Clin­ton and led many to be­lieve she was a shoo-in to win the Elec­toral Col­lege.

“The polls clearly got it wrong this time,” the Amer­i­can As­so­ci­a­tion for Pub­lic Opin­ion Re­search said Wed­nes­day in a state­ment. The as­so­ci­a­tion tra­di­tion­ally as­sesses the state of pub­lic polling after each elec­tion cy­cle, and al­ready has a com­mit­tee in place to do so again this year.

“I think it was an im­por­tant polling miss. It would re­ally be gloss­ing over it to say that it was a typ­i­cal year,” said Court­ney Kennedy, di­rec­tor of sur­vey re­search at the Pew Re­search Cen­ter.

For now, it’s im­pos­si­ble to know for cer­tain what ex­actly went wrong for poll­sters this year — and, as votes are still be­ing counted, ex­actly how far off they were. Some fac­tors poll­sters will ex­am­ine:

How big a miss?

Al­though most polls through­out the 2016 cam­paign showed Clin­ton run­ning ahead of Trump, in the fi­nal two weeks of the cam­paign her ad­van­tage nar­rowed in many na­tional sur­veys, as well as in states such as Penn­syl­va­nia and Michi­gan. Her ap­par­ent lead fell within many sur­veys’ mar­gins of sam­pling er­ror.

Kennedy said poll­sters may ul­ti­mately not have had a his­tor­i­cally large miss on the na­tional pop­u­lar vote, but thinks there was a sys­tem­atic over­rep­re­sen­ta­tion of Clin­ton’s sup­port and un­der­rep­re­sen­ta­tion for Trump’s.

She says peo­ple some­times ex­pect too much of elec­tion polls, which “are not de­signed to pro­vide ex­tremely ac­cu­rate re­sults.”

Lee Miringoff, di­rec­tor of the Marist Col­lege In­sti­tute for Pub­lic Opin­ion, says that av­er­ages of pub­licly avail­able polls some­times give a false sense of cer­tainty in a can­di­date’s lead.

“You’re tak­ing im­pre­cise es­ti­mates and throwing them all to­gether with the hope of elim­i­nat­ing er­ror,” he says.

Shy Trump vot­ers?

Trump’s cam­paign fre­quently pointed to the pos­si­bil­ity that pub­lic polls were miss­ing some of his base of sup­port, and some poll­sters say that might have played a role in the polling miss.

“One of the big­gest prob­lems that polls face nowa­days is that peo­ple don’t want to par­tic­i­pate in them at all,” said Pa­trick Mur­ray, di­rec­tor of the Mon­mouth Univer­sity Polling In­sti­tute. He plans to use voter data to find out if cer­tain types of peo­ple were less likely to par­tic­i­pate in his sur­veys.

At Pew, Kennedy said it ap­pears that there was a seg­ment of Trump’s sup­port base that was not re­spond­ing to polls, which she called “fun­da­men­tally a dif­fi­cult chal­lenge to fight.” But, she said, it’s un­likely vot­ers were ly­ing about their sup­port.

Turnout

Harold Clarke, a po­lit­i­cal sci­en­tist at the Univer­sity of Texas at Dal­las who reg­u­larly con­ducts polling, said one of the short­falls in the pres­i­den­tial pre­dic­tion was a prob­lem that has plagued sur­vey sci­ence for decades.

“We’ve got to fil­ter our sur­veys as we try to pick out those peo­ple that are re­ally go­ing to vote,” he said. “We all have the prob­lem of not get­ting likely vot­ers right.”

Mur­ray said poll­sters are us­ing likely voter mod­els that might have worked in the past, but may no longer. He sug­gested that pub­lic poll­sters should take a les­son from cam­paigns and con­sider putting out a range of num­bers re­flect­ing dif­fer­ent turnout sce­nar­ios in­stead of a sin­gle num­ber that sug­gests too much cer­tainty in where the horse race stands.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.