The AP’s Mission to Save the Exit Poll

Faced with rising costs and a greater degree of difficulty, the Associated Press is researching new methods of surveying voters during elections.

Election day volunteer Vicki Groff places a sign to direct voters to a polling station at Kenilworth School February 28, 2012 in Phoenix, Arizona. 
Getty Images
Scott Bland
Add to Briefcase
See more stories about...
Scott Bland
Aug. 12, 2015, 1:01 a.m.

Exit poll­sters do their work in a few short days, but the in­form­a­tion they glean dom­in­ates the polit­ic­al con­ver­sa­tion for years.

That’s be­cause the sur­veys con­duc­ted out­side vot­ing pre­cincts are the single-best tool for un­der­stand­ing Amer­ic­an polit­ics. They gen­er­ate a treas­ure trove of data — about who votes, who they voted for, and why — that tell the story of how each elec­tion was won and lost.

But now, as more voters cast bal­lots early or by mail and polling those voters be­comes more costly and po­ten­tially less re­li­able, the exit polls are get­ting more ex­pens­ive, too — and might even be los­ing some of their pre­dict­ive power. Rising costs already con­vinced the Na­tion­al Elec­tion Pool, the me­dia con­sor­ti­um that con­ducts U.S. exit polls, to aban­don some exit polls in non­battle­ground states in 2012, leav­ing ob­serv­ers without a pres­id­en­tial elec­tion’s worth of data in rap­idly chan­ging states like Geor­gia and Texas.

(RE­LATED: Should Tele­vi­sion Net­works Use Na­tion­al Polls to De­term­ine Who Par­ti­cip­ates in Pres­id­en­tial De­bates?)

In re­sponse, the As­so­ci­ated Press is try­ing to build the exit poll of the fu­ture.

Last month, the me­dia or­gan­iz­a­tion won a $250,000 Knight News Chal­lenge grant to con­duct ex­per­i­ments aimed at build­ing a poll that’s both re­li­able and cost-ef­fect­ive in 21st-cen­tury con­di­tions.

Over the next 15 months, the news or­gan­iz­a­tion and its polling part­ners will use the money to test on­line sur­vey meth­ods dur­ing real cam­paigns, from the 2015 Ken­tucky gov­ernor’s race through next year’s pres­id­en­tial elec­tions. By the time it closes the books on the 2016 elec­tions, AP hopes to have iden­ti­fied sur­vey meth­ods that it can re­com­mend to the NEP, which in­cludes the main broad­cast and cable news net­works, for use in the of­fi­cial exit poll in fu­ture elec­tions.

The exit poll’s main chal­lenge arises when voters avoid the tra­di­tion­al exit and don’t ac­tu­ally go to the polls on Elec­tion Day. (In­stead, rising num­bers vote by mail or cast a bal­lot in per­son dur­ing an early vot­ing peri­od.) The NEP has to sup­ple­ment its tra­di­tion­al in­ter­views out­side voter pre­cincts with tele­phone polling to reach them, and phone polling is get­ting both more dif­fi­cult and more ex­pens­ive. Reg­u­la­tions gov­ern­ing calls to cell phones have driv­en up poll­sters’ costs, and nose­diving re­sponse rates to phone sur­veys — of­ten less than one-in-10 people called par­ti­cip­ate in a sur­vey — threaten the ac­cur­acy of the sur­veys at the same time by mak­ing it less likely to con­tact a truly ran­dom sample of people.

“In the last dec­ade or so, people have really changed the way they vote,” said Dav­id Pace, AP’s Wash­ing­ton news ed­it­or. “It used to be every­body voted in their neigh­bor­hood pre­cinct and exit polls worked great. “¦ But one-third of voters voted ab­sent­ee in 2012, totally out of the reach of exit poll­sters” stand­ing out­side pre­cincts. “That’s a trend that’s been grow­ing for eight-to-10 years,” Pace con­tin­ued.

(RE­LATED: Hil­lary Clin­ton’s White Male Voter Prob­lem)

That led the NEP to con­duct sup­ple­ment­ary phone polls in 35 states in 2012, Pace said. But that raises the cost is­sue of phones.

“It’s really not fin­an­cially sus­tain­able,” said Pace. “So we’re look­ing ahead and ex­per­i­ment­ing be­cause, if this trend keeps go­ing, it’s go­ing to make it harder and harder and more and more ex­pens­ive to ad­apt exit polls to the cur­rent en­vir­on­ment.”

That’s where re­l­at­ively cheap but less proven on­line polling comes in. A poll­ster can’t con­tact a ran­dom sample of In­ter­net users the same way they can ap­proach ran­dom voters out­side a pre­cinct or dial ran­dom tele­phone num­bers. But re­search­ers are in­vest­ig­at­ing dif­fer­ent ways to ap­prox­im­ate the vot­ing pop­u­la­tion on­line, where ask­ing them ques­tions costs a frac­tion of con­duct­ing a phone poll.

(RE­LATED: Map: Where the 2016 Can­did­ates Raised Their Money)

One of AP’s polling part­ners, GfK, has already been run­ning ex­per­i­ments com­par­ing on­line polling with the reg­u­lar exit polls. In 2014, the firm blen­ded res­ults from dif­fer­ent on­line pan­els (one built to look like a rep­res­ent­at­ive sample of the pop­u­la­tion and one in which re­spond­ents simply vo­lun­teered to par­ti­cip­ate) to pro­ject the res­ults of Sen­ate and gubernat­ori­al races in Geor­gia and Illinois.

The on­line sur­vey per­formed com­par­ably to the reg­u­lar exit poll in those states, with a slightly smal­ler av­er­age er­ror, ac­cord­ing to slides AP and GfK presen­ted at the Amer­ic­an As­so­ci­ation for Pub­lic Opin­ion Re­search con­fer­ence this spring.

“Like every in­dustry, we’re fa­cing cost pres­sure,” said GfK man­aging dir­ect­or An­nie Weber. “One of the big moves is to move away from some of the more tra­di­tion­al modes of reach­ing re­spond­ents — in-per­son or even by tele­phone is in­creas­ingly chal­len­ging today. So we’re try­ing to ex­plore how to do high-caliber, con­sist­ently ac­cur­ate polling on­line.”

On­line polling was an ana­thema to many sur­vey re­search­ers in its in­fancy, but the vari­ous pres­sures fa­cing the in­dustry has sparked in­tense in­terest in re­li­able on­line polling.

“The ques­tion is less clear than it used to be,” said Steve Koczela, the pres­id­ent of Bo­ston-based MassINC Polling Group and a former pres­id­ent of the New Eng­land chapter of AA­POR. “On­line, you gain a num­ber of things. People can re­spond whenev­er they want, and there’s vari­ous ad­vant­ages to ad­min­is­ter­ing a poll like that. But you do lose that re­la­tion­ship to the sci­ence that polling was built on [with opt-in pan­els]. Yet that re­la­tion­ship is be­ing frayed by a lot of oth­er things, too.”

AP’s re­search pro­ject is sep­ar­ate from the reg­u­lar exit poll, ac­cord­ing to Joe Lenski, who su­per­vises the NEP exit poll as an ex­ec­ut­ive vice pres­id­ent at Edis­on Re­search. But Lenski said he, too, re­cog­nizes the need to ex­plore new meth­ods and tech­niques. The NEP as a whole has done its own re­search and ex­per­i­ments in re­cent years aimed at up­dat­ing and stream­lin­ing the exit poll, es­pe­cially since rising costs con­vinced the pool to aban­don some states 2012.

Exit-polling “is be­com­ing an is­sue not just in the United States but in coun­tries around the world,” Lenski said. “Some of the par­tic­u­lar is­sues the U.S. has are unique, but is­sues with exit-polling in gen­er­al is not unique.”

This story has been up­dated to cla­ri­fy Koczela’s quote.

What We're Following See More »
Possible Active Shooter at Alabama Military Installation
46 minutes ago
Senate Procedural Vote Now Coming on Wednesday
47 minutes ago
SCOTUS to Hear Sports Betting Case
1 hours ago

"The U.S. Supreme Court has given new life to New Jersey's challenge to a federal sports betting ban, with the high court announcing Tuesday that it hear an appeal of federal court decisions that have blocked the state's plans. That extends a six-year effort led by Gov. Chris Christie to allow expanded gambling at Monmouth Park." The NFL, NCAA, and other popular sports leagues had opposed the sports betting there and at other New Jersey locations.

More Chemical Weapons Activity Spotted at Syrian Base
2 hours ago

"The Pentagon said Tuesday the United States has seen chemical weapons activity at Syrian air base used in past chemical attack." A Pentagon spokesman confirmed what the White House first said Monday night: that "Bashar Assad appears to be taking some of the same actions he took before a chemical weapons attack on his own people in April."

Ethics Committee Looking into Conyers, Lujan
5 hours ago

"The House Ethics Committee acknowledged Monday it is investigating Reps. Ben Ray Luján, John Conyers and House staffer Michael Collins. The panel did not disclose details of its inquiry. Since the probe was referred to the House committee by the Office of Congressional Ethics, details of the OCE’s reports are expected to be made public August 9."


Welcome to National Journal!

You are currently accessing National Journal from IP access. Please login to access this feature. If you have any questions, please contact your Dedicated Advisor.