Failed polls in 2016 call into ques­tion a pro­fes­sion’s pre­cepts


— The polls blew it.

U.S. sur­vey com­pa­nies and me­dia or­ga­ni­za­tions that col­lec­tively pre­saged a close Hil­lary Clin­ton vic­tory now face an au­topsy on how they got it so wrong after a year suf­fused by polls, ag­gre­gates of polls and even real-time pro­jec­tions of the vote on Elec­tion Day.

While the pre­dic­tions gave some ob­servers a sooth­ing sense of cer­tainty, ac­tual vot­ers still pos­sessed the ca­pac­ity to shock. Don­ald Trump’s com­mand­ing per­for­mance de­fied the fi­nal sur­veys of the Amer­i­can elec­torate, which broadly pre­dicted a Clin­ton win of 2 to 4 per­cent­age points.

“It’s harder and harder to poll to­day, to get a sam­ple that looks like the elec­torate,” said Kar­lyn Bow­man, a pub­lic opin­ion an­a­lyst at the Amer­i­can En­ter­prise In­sti­tute in Wash­ing­ton. “We’ve seen epic fails.”

Tues­day’s re­sults were just the lat­est high-pro­file pre­dic­tive fail­ure around the world, fol­low­ing on the heels of mis­lead­ing sur­veys on the Colom­bia peace deal ref­er­en­dum this year and Greece’s bailout ref­er­en­dum in 2015. Sur­veys were ren­dered in­ac­cu­rate by new forms of tech­nol­ogy and com­mu­ni­ca­tion and po­lit­i­cal ques­tions un­like any seen in re­cent his­tory. The in­ac­cu­racy called into ques­tion a ba­sic gauge of sen­ti­ment used by politi­cians, cit­i­zens and fi­nanciers.

“The anger is stronger than any of us re­ally ex­pected,” said Me­gan Greene, chief econ­o­mist at Man­ulife As­set Man­age­ment in Bos­ton, which han­dles money for in­sti­tu­tional in­vestors such as pen­sions and foun­da­tions.

In the U.S., ques­tions linger about how to slice the elec­torate and how to weight un­der-rep­re­sented de­mo­graph­ics — whether by eth­nic­ity or lo­ca­tion or po­lit­i­cal af­fil­i­a­tion — while Amer­i­cans in­creas­ingly with­draw from sur­vey par­tic­i­pa­tion and view poll­sters them­selves through a po­lit­i­cal lens.

Fi­nal tal­lies pro­duced by CBS News, FiveThir­tyEight, Fox News, Wall Street Jour­nal-NBC News and Wash­ing­ton Post-ABC News all pre­dicted a rel­a­tively safe 4-point win for Clin­ton. Only slightly less wrong were polls by Bloomberg Pol­i­tics and New York Times’s Up­shot, which es­ti­mated a Clin­ton vic­tory by 3 points. Ras­mussen Re­ports called for a 2-point Clin­ton tri­umph.

A few got it right: The In­vestor’s Busi­ness Dai­lyTech­noMet­rica Mar­ket In­tel­li­gence poll and a tally by USC Dorn­sife-Los Angeles Times were among the rare out­fits to call the elec­tion for Trump, by 2 and 3 points.

Peter Wool­ley, a pro­fes­sor of com­par­a­tive pol­i­tics at Fair­leigh Dick­in­son University in Florham Park, New Jersey, said a key part of the dif­fer­ence be­tween ex­pec­ta­tions and the re­sults was that peo­ple sim­ply ex­pected sur­veys to be too pre­cise. Wool­ley is a past di­rec­tor of the PublicMind polling in­sti­tute at the university.

“Polling is a sci­en­tific method to ar­rive at an es­ti­mate,” he said early Wed­nes­day. “We tend to over-re­port the ac­cu­racy of the poll, and tend to for­get very quickly that it’s an es­ti­mate within a range.”

J. Ann Selzer, an Iowa poll­ster who con­ducts sur­veys for Bloomberg Pol­i­tics, said her trade en­tered un­charted ter­ri­tory this year as it at­tempted to deal with the spread of wire­less com­mu­ni­ca­tion and a de­mo­graph­i­cally volatile elec­torate.

“There was a lot of ex­per­i­ment­ing this year with the types of ques­tions they were us­ing, the types of method­olo­gies they were us­ing,” she said in an in­ter­view at Bloomberg News head­quar­ters in New York. “There’s the con­tin­u­ing bar­rier of the lack of land­lines, the ero­sion of land­lines. In the old days, if we knew your land­line phone num­ber we knew where you lived and that was fan­tas­tic for poll­sters. Now it’s very dif­fi­cult.”

Turn­ing points in the race hap­pened at poorly timed mo­ments, she said.

“Day by day, things hap­pened that would break the poll that was cur­rently in the field,” she said. “Even since Sun­day, when most polls were done, things changed.”

Poll­sters have been deal­ing with a de­cline in par­tic­i­pa­tion for decades. The share of house­holds that agreed to par­tic­i­pate in a tele­phone sur­vey by the Pew Re­search Cen­ter steadily dropped to 14 per­cent by 2012 from 43 per­cent in 1997.

When the phone is an­swered, many are un­will­ing to speak plainly.

“Let’s say you hold a con­tro­ver­sial opin­ion,” said Greg Val­liere, the Wash­ing­ton-based chief global strate­gist at Hori­zon In­vest­ments LLC, a Char­lotte fi­nan­cial ad­viser. “Are you go­ing to tell a com­plete stranger that calls you up at 8 in the evening, or will you keep it to your­self?”

Polls abound. For many me­dia or­ga­ni­za­tions, they serve not only as tools for their own use, but as call­ing cards and valu­able fran­chis- es as the Internet erodes their author­ity and earn­ing power. In re­cent years, statis­ti­cians such as Nate Sil­ver, founder of ESPN’s FiveThir­tyEight, have achieved a sort of nerdy celebrity.

Mean­while, the grand­fa­ther of U.S. opin­ion polling, Wash­ing­ton-based Gallup Inc., has pulled back. Four years ago, Gallup en­dured its third polling de­feat in four cy­cles and walked away from pres­i­den­tial horse-race polling al­to­gether. An in­ter­nal re­view found that four ar­eas of the method­ol­ogy con­trib­uted to the mis­taken pre­dic­tions: likely-voter es­ti­mates, re­gional rep­re­sen­ta­tion, weight­ing of race and eth­nic­ity, and out­reach to land­line-phone users.

Among high-pro­file fail­ures this year were the Michi­gan Demo­cratic pri­mary and the Colom­bian peace deal.

In the March 8 pri­mary, Clin­ton had been fa­vored by about 20 points in sur­veys, only to lose to Ver­mont Sen. Bernie San­ders by al­most 20,000 votes. In Colom­bia, vot­ers in Oc­to­ber re­jected a peace deal with rebels after sev­eral sur­veys in the prior week had pre­dicted the ref­er­en­dum would pass by 60 per­cent or more.

At the same time, poll­sters have got­ten some rel­a­tively un­de­served crit­i­cism. The U.K.’s June vote to leave the Euro­pean Union, of­ten called a sur­prise re­sult, ac­tu­ally was largely deemed too close to call by opin­ion polls. While mar­kets priced in a vote for “Re­main,” tra­di­tional tal­lies were much closer to the end re­sult for a Brexit win.

Trump has been out­spo­ken on polling re­sults through­out the cam­paign, at times laud­ing those that project he has more sup­port, while de­rid­ing oth­ers for shoddy method­ol­ogy or bias.

“I do think a lot of the polls are pur­posely wrong,” he said Tues­day on Fox News. “I think a lot of the polls are phony — I don’t even think they in­ter­view peo­ple.”

Bow­man of the Amer­i­can En­ter­prise In­sti­tute said the ev­ery pre­cept must be re­ex­am­ined.

“I don’t think the busi­ness is par­tic­u­larly in­tro­spec­tive, but it needs to be go­ing for­ward,” she said. “This has been a busi­ness that’s told us so much about Amer­ica. ... To lose that go­ing for­ward would be a real prob­lem.”


Sup­port­ers of pres­i­den­tial can­di­date Hil­lary Clin­ton react as exit poll re­sults ar­rive dur­ing an elec­tion night rally on Tues­day in New York City.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.