Why the GOP Still Doesn’t Have Its Act Together on Polling

Divisions remain within the party about the best way to proceed — and the best practices to follow.

Voters stand in a long line at the Supervisor of Elections office in West Palm Beach, Florida November 5, 2012. Palm Beach County Supervisor of Elections Supervisor Susan Bucher is one of five supervisors in heavily populated counties who has allowed in-person absentee voting after Florida Republican Governor Rick Scott refused to extend early voting. 
Steven Shepard
Add to Briefcase
See more stories about...
Steven Shepard
Sept. 12, 2013, 4:15 p.m.

Ever since Elec­tion Day 2012, when Re­pub­lic­ans lost the pres­id­en­tial pop­u­lar vote for the fifth time in six cycles, in ad­di­tion to two Sen­ate seats and sev­en House seats, the party has been en­gaged in an ex­tens­ive eval­u­ation of its poll­sters and the mil­lions of dol­lars of sur­vey re­search com­mis­sioned by its can­did­ates and party com­mit­tees each year. That ef­fort — spear­headed by the Re­pub­lic­an Na­tion­al Com­mit­tee and the Na­tion­al Re­pub­lic­an Con­gres­sion­al Com­mit­tee, which are look­ing to re­as­sert them­selves as power play­ers in D.C. — is a small but telling part of a tenu­ous pro­cess that seeks to trans­form the GOP in­to a party equipped to win among a di­ver­si­fy­ing elect­or­ate.

It’s an un­der­tak­ing that, at times, has been as rocky as the party’s over­all rebrand­ing ef­fort. NR­CC of­fi­cials and poll­sters in­volved in the pro­cess this year por­trayed the meet­ings as largely pos­it­ive, but they have been marked by more de­bate than has been pre­vi­ously re­por­ted, and by per­sist­ent dis­agree­ment among some that the party’s polls last year were all that far off.

“There’s a little bit of “˜Oh, we’re still go­ing to do it this way,’ “ said Jon McHenry of North Star Opin­ion Re­search in Al­ex­an­dria, Va., while not­ing that such sen­ti­ment was not the con­sensus of the NR­CC get-to­geth­ers. “We all had a stake in mak­ing sure the NR­CC did things in the best way pos­sible.”

Some poll­sters, though, are reti­cent to give up too much autonomy to the na­tion­al party or to spill trade secrets to rivals. While poll­sters work to­geth­er in some cases — polling for the cam­paign com­mit­tees’ in­de­pend­ent-ex­pendit­ure units is one not­able ex­ample — they of­ten com­pete against one an­oth­er when it comes to can­did­ate busi­ness. “We are totally on board,” said Kel­ly­anne Con­way, pres­id­ent and CEO of the polling com­pany, inc./Wo­man­Trend, who at­ten­ded a gath­er­ing at the RNC. “But as I sat in that meet­ing, I thought, “˜Wait a second, if we do all this, you’re ac­tu­ally cut­ting in­to my com­pet­it­ive ad­vant­age.’ “

Moreover, the un­der­ly­ing real­it­ies that are chal­len­ging poll­sters — for ex­ample, that tech­no­logy is mak­ing some voters harder to reach by land­line tele­phone — aren’t that dif­fer­ent from the GOP’s struggles to reach cer­tain seg­ments of the elect­or­ate. In fact, some of the party’s polling prob­lems in 2012 were rooted in er­ro­neous as­sump­tions about the com­pos­i­tion of the elect­or­ate. Most Re­pub­lic­an poll­sters who spoke with Na­tion­al Journ­al throughout this year said they thought the elect­or­ate would be more fa­vor­able to their party — mean­ing whiter, older, and over­all more Re­pub­lic­an — than the voters who showed up to cast bal­lots on Elec­tion Day. It was a vi­cious cycle for the party: In­cor­rect as­sump­tions res­ul­ted in falsely op­tim­ist­ic poll num­bers dur­ing the cam­paign, and those num­bers served to re­in­force the GOP’s pro­jec­tions about turnout.

But turnout rates in 2012 ac­tu­ally fell among white voters. Some on the right have sug­ges­ted that goos­ing turnout among the party’s base would be a more ef­fect­ive strategy than try­ing to ap­peal to minor­it­ies or young­er voters. The GOP es­tab­lish­ment has taken the op­pos­ite view: The RNC’s Growth and Op­por­tun­ity Re­port, which party of­fi­cials hoped would serve as a blue­print for a polit­ic­al comeback, spe­cific­ally en­dorsed com­pre­hens­ive im­mig­ra­tion re­form as a start­ing point for ap­peal­ing to His­pan­ics and re­com­men­ded ef­forts to reach out to young people.

In many ways, the polling re­boot is em­blem­at­ic of this broad­er de­bate with­in the party: Does the GOP need to ad­apt to chan­ging demo­graph­ics to win na­tion­al elec­tions, and, if so, to what ex­tent?

The NR­CC and the RNC have both pro­duced lists of re­com­mend­a­tions. Most GOP poll­sters who spoke with Na­tion­al Journ­al en­dorsed the pro­pos­als as a set of best prac­tices, al­though they differed on the size and scope of the party’s polling prob­lems. Some said the re­com­mend­a­tions are in­com­plete, em­phas­iz­ing some meas­ures over oth­ers for fix­ing the party’s polls. Oth­er GOP poll­sters who wer­en’t in­volved in the NR­CC pro­cess see the ef­fort as too in­su­lar. And most of the re­com­mend­a­tions would res­ult in high­er costs that would have to be passed along to cli­ents.

The over­rid­ing theme of these con­ver­sa­tions, however, is of a party try­ing to play catch-up with its Demo­crat­ic rival when it comes to sur­vey re­search and data. That this de­fi­cit has come amid broad­er, longer doldrums for the GOP is hardly co­in­cid­ent­al.

“Demo­crats are at the gradu­ate level,” one Re­pub­lic­an poll­ster groaned this sum­mer. “We’re still in kinder­garten when it comes to the be­ha­vi­or­al side of voter opin­ion.”


When Rep. Greg Walden, R-Ore., the chair­man of the NR­CC, stood be­fore the House GOP Con­fer­ence on May 15 to present re­com­mend­a­tions for the party’s polling in the 2014 cycle, his re­port rep­res­en­ted the cul­min­a­tion of an un­pre­ced­en­ted months-long re­as­sess­ment spear­headed by the com­mit­tee.

A more de­tailed memor­andum had gone out to most of the party’s ma­jor polling firms two days be­fore Walden’s present­a­tion, GOP sources said. Both the memo and Walden’s present­a­tion con­tained a series of best prac­tices for GOP sur­vey re­search­ers de­veloped after months of meet­ings and in­put with poll­sters used by the NR­CC and Re­pub­lic­an House can­did­ates.

Some Re­pub­lic­ans have faul­ted the party’s polling — more spe­cific­ally, the mod­el­ing that built as­sump­tions around what the 2012 elect­or­ate would look like — for their poor per­form­ance last year. Had the polling more ac­cur­ately re­flec­ted the elect­or­ate, it could have en­abled cam­paigns, the cam­paign com­mit­tees, and oth­er out­side groups to make bet­ter-in­formed de­cisions to max­im­ize the value of their in­vest­ments, they ar­gue.

But oth­ers deny the prob­lem is that ser­i­ous. They con­tend the party’s polling (or their polling, at least) was ac­cur­ate — un­til some races slipped away at the end. While in­tern­al polls two weeks be­fore the elec­tion may not jibe with elec­tion res­ults, they were ac­cur­ate at the time, these poll­sters con­tend. And al­though polling im­me­di­ately be­fore the elec­tion would pro­duce more-ac­cur­ate num­bers, it would have little value oth­er than to cre­ate ex­pect­a­tions; the sorts of de­cisions that polls in­form, such as tele­vi­sion ads and oth­er in­vest­ments, are already made be­fore the fi­nal week­end of the cam­paign.

The Re­pub­lic­ans’ con­gres­sion­al com­mit­tee led a com­pre­hens­ive re­view aimed at ad­dress­ing many of the chal­lenges fa­cing poll­sters on both sides of the aisle. Amer­ic­ans are be­com­ing harder to reach by tele­phone, par­tic­u­larly land­lines. And as Re­pub­lic­ans con­tin­ue to struggle to at­tract non­white voters, who tend to be even more dif­fi­cult to reach than whites, in­clud­ing the right num­ber of minor­ity voters and in the right pro­por­tion is es­sen­tial to pro­du­cing an ac­cur­ate poll.

To ac­com­plish that, the two en­tit­ies driv­ing the GOP re-brand­ing have each made a num­ber of re­com­mend­a­tions. The NR­CC asked Re­pub­lic­an polling firms to use lar­ger sample sizes, call more cell phones, and con­duct bi­lin­gual in­ter­views in dis­tricts with sig­ni­fic­ant His­pan­ic pop­u­la­tions. The RNC autopsy also asked the party’s sur­vey re­search­ers to ex­per­i­ment with re­gis­tra­tion-based sampling meth­ods — call­ing phone num­bers ran­domly se­lec­ted from a known list of re­gistered voters — that are already used by some, but not all, cam­paign poll­sters. “The thing is, with a lis­ted sample, you know so much more about these people,” said Alex Lun­dry, a poll­ster with Tar­get­Point Con­sult­ing in Al­ex­an­dria, Va., and the head of Mitt Rom­ney’s 2012 data team. “That helps en­rich the ana­lys­is of the com­pletes you have; it helps you do a bet­ter ana­lys­is of non­re­sponse bi­as.”

The con­ven­tion­al ar­gu­ment for us­ing re­gis­tra­tion-based sampling cen­ters on poll­sters’ at­tempts to identi­fy which voters are likely to turn out on Elec­tion Day. Some poll­sters will start with the sample that in­cludes only voters whom pub­lic re­cords show cast bal­lots in re­cent elec­tions, provided they were eli­gible to do so. That, the voter-list poll­sters ar­gue, can re­duce the num­ber of non­voters in their samples.

Lun­dry’s ar­gu­ment for re­gis­tra­tion-based sampling takes that fur­ther. Com­plet­ing an in­ter­view off a voter file gives you demo­graph­ic in­form­a­tion — and, po­ten­tially, con­sumer data that can be matched to it. Fur­ther­more, be­cause you have more in­form­a­tion about the people with whom you’re speak­ing, you also know more about the people whom you aren’t reach­ing as ef­fect­ively by tele­phone.

But some firms are hold­ing onto ran­dom-di­git di­al­ing for sampling, most not­ably Pub­lic Opin­ion Strategies, the firm that polled for Rom­ney last year. At an event in Wash­ing­ton earli­er this month, Bill McIn­turff, a part­ner at POS, de­fen­ded ran­dom-di­git di­al­ing and said that us­ing voter files for sampling could lead to the same un­der­es­tim­a­tion of cer­tain blocs of voters that plagued the party’s 2012 polls.

“When you look at the voter file of [people] who ac­tu­ally voted, what you find is, it doesn’t match the exit polls,” McIn­turff said.

“It’s much older, whiter, less eth­nic; and my fight in­side the party’s been, “˜Yeah, be­cause those are the folks they can match [to voter files]!’ That means there are thou­sands of voters that you can’t match that you’re miss­ing. Guess what: They’re poorer, young­er, and eth­nic and non­white.”¦ There’s a lot of oth­er folks who can vote that don’t get matched on a voter file be­cause we can’t pick ‘em out be­cause we don’t have a phone num­ber.”

While the fight over sampling meth­ods rages — the RNC re­port re­com­men­ded side-by-side stud­ies of both meth­ods — the NR­CC’s sug­ges­tion of lar­ger sample sizes may help solve the prob­lem. It would al­low cam­paigns and oth­er groups to have a bet­ter, more sig­ni­fic­ant un­der­stand­ing of harder-to-reach demo­graph­ic groups, such as young people, minor­it­ies, and in­de­pend­ents. Moreover, the seg­ments of these co­horts reached may sig­ni­fic­antly dif­fer from the pop­u­la­tion of the co­hort at large. In oth­er words, young­er voters who par­ti­cip­ate in phone polls (par­tic­u­larly land­line phone polls) dif­fer from the over­all young­er elect­or­ate — it’s thought they are more likely to be liv­ing with their par­ents — to a great­er de­gree than seni­ors who par­ti­cip­ate in polls dif­fer from the over­all seni­or elect­or­ate. Be­cause young people, minor­it­ies, and in­de­pend­ents tend to be un­der­rep­res­en­ted to be­gin with, the prac­tice of weight­ing the value of their re­sponses to meet a set threshold may ac­tu­ally in­crease the er­ror in the poll.

The party has re­com­men­ded in­creas­ing the num­ber of cell-phone in­ter­views be­cause, for more than a third of Elec­tion Day 2012 voters, their mo­bile device was their only tele­phone. This was also true in key battle­ground states. Some 34 per­cent of Ohio voters, 32 per­cent of Vir­gin­ia voters, 35 per­cent of North Car­o­lina voters, and 34 per­cent of Flor­ida voters were cell-only. Cell-only Amer­ic­ans also tend to be young­er and dis­pro­por­tion­ately non­white. Un­der­rep­res­ent­ing them in Re­pub­lic­an polls is likely to over­state the GOP can­did­ate’s bal­lot po­s­i­tion.

GOP poll­ster Glen Bol­ger wrote in a Novem­ber 2012 memor­andum that the party must start sur­vey­ing more voters us­ing cell phones. “We have been do­ing cell-phone in­ter­view­ing for sev­er­al cycles now, but really in­creased it this year with cell-phone in­ter­view­ing not only at the na­tion­al and statewide level, but also at the con­gres­sion­al level. What we do know now is that 20 per­cent of the in­ter­views with cell phones was not enough,” Bol­ger wrote.

Mean­while, con­duct­ing more in­ter­views in Span­ish is de­signed to get a more ac­cur­ate de­pic­tion of His­pan­ic voters and their per­cent­age as a share of the elect­or­ate, ac­cord­ing to Bol­ger. It isn’t enough for Re­pub­lic­ans to just in­crease the num­ber of His­pan­ics in their sur­veys. Boost­ing the per­cent­age of Lati­nos who com­plete a phone sur­vey con­duc­ted in Eng­lish may only ex­acer­bate a poll’s de­fi­ciency: Those who would be more com­fort­able an­swer­ing ques­tions in Eng­lish may be­have dif­fer­ently at the bal­lot than His­pan­ics who don’t want to par­ti­cip­ate in an Eng­lish-only sur­vey.

Each of these re­com­mend­a­tions rep­res­ents a re­sponse to changes in demo­graph­ic trends and in the way people com­mu­nic­ate with one an­oth­er. But each re­com­mend­a­tion also rep­res­ents a sig­ni­fic­ant in­crease in sur­vey costs that would be passed on to cli­ents.

The NR­CC, in ad­voc­at­ing for these more ex­pens­ive pro­to­cols, says it is will­ing to pay high­er prices for more-re­li­able polling. “I think there is an ac­know­ledg­ment based on all of these things that cost is go­ing to go up a little bit, but I think there’s also an ac­know­ledg­ment from us on the con­sumer side of this and from our can­did­ates and cam­paigns that the cost part of this is fine so long as the data and the stuff you’re get­ting back is ac­cur­ate,” said NR­CC Polit­ic­al Dir­ect­or Rob Simms.Still, not every Re­pub­lic­an polling firm is singing from the same hym­nal. Some poll­sters said their res­ults were ac­cur­ate, and that the ree­valu­ation pro­cess is based on the in­cor­rect as­sump­tion that, be­cause the party’s stand­ard-bear­er lost when his poll­ster felt con­fid­ent about his chances, the party’s polls on the whole were faulty.

One GOP poll­ster de­scribed a con­fer­ence call with Neil Ne­w­house, Bol­ger’s part­ner at Pub­lic Opin­ion Strategies and the Rom­ney cam­paign’s lead poll­ster, the Fri­day be­fore the elec­tion in which Ne­w­house said that each of the 28 latest pub­lic polls showed Rom­ney lead­ing among in­de­pend­ents. The reas­on the en­tire GOP polling in­dustry was asked to come to­geth­er, the poll­ster said, was be­cause the Rom­ney cam­paign thought it had a plaus­ible path to vic­tory based on faulty data. “I don’t think if Mitt Rom­ney had won based on POS’s polling, they’d be in that room,” the poll­ster said.


Re­pub­lic­ans hope their ef­forts to im­prove the party’s polling will help close the over­all data gap. “What’s im­port­ant to re­mem­ber is that cam­paigns are already in­creas­ing the amount of data that they look at,” said Brock Mc­Cle­ary, the NR­CC’s polling dir­ect­or dur­ing the 2012 elec­tion cycle and the founder of a new firm that fo­cuses on auto­mated phone polling. “Polling should al­most be seen in the con­text of an en­tire op­er­a­tion of data.”

Obama’s 2012 cam­paign provides an in­ter­est­ing mark­er for the GOP’s ef­forts. Re­pub­lic­ans clearly would like to rep­lic­ate the Demo­crats’ suc­cesses, but it’s less clear that the NR­CC- and RNC-led polling re­com­mend­a­tions will close that gap.

Lun­dry’s po­s­i­tion with the Rom­ney camp gave him a unique, in­side per­spect­ive on cam­paign data in 2012. “I strongly be­lieve that we had ac­cess to about 90 to 95 per­cent of the data tech­no­logy and tools the Obama cam­paign had,” Lun­dry said, ref­er­en­cing postelec­tion re­ports on the op­pos­i­tion’s op­er­a­tions. “The ex­traordin­ary dif­fer­ence was a lack of in­teg­ra­tion.”

Lun­dry thinks the polling re­boot comes up short in eras­ing that dis­par­ity en­tirely, but he cited re­ports that the NR­CC would be­gin us­ing polls to bet­ter in­form its tele­vi­sion ad­vert­ising spend­ing. “I think that’s ex­traordin­ar­ily im­port­ant,” he said. “TV’s the biggest line item in the budget but the least data-driv­en.”

Lun­dry’s latest pro­ject, the data-track­ing firm Deep Root Ana­lyt­ics, is set to provide Re­pub­lic­an can­did­ates and out­side groups with even more-de­tailed in­form­a­tion — culled an­onym­ously from view­ers’ cable set-top boxes — to in­form ad buy­ing and tar­get­ing of spe­cif­ic demo­graph­ics. In a sep­ar­ate in­ter­view in Au­gust, Lun­dry said he hopes to in­teg­rate that in­form­a­tion with polling at the des­ig­nated mar­ket-area level. The New York Times re­por­ted last month that Deep Root Ana­lyt­ics was work­ing for New Jer­sey Gov. Chris Christie’s reelec­tion cam­paign. “We’re try­ing to have as com­pre­hens­ive a look at the paid-me­dia eco­sys­tem and try­ing to un­der­stand what’s work­ing and what’s not,” Lun­dry said.

The NR­CC also ac­know­ledged a lack of in­teg­ra­tion between its data and de­cision-mak­ing.

“It’s not that Demo­crats have the mono­poly on really good data, and we’re flounder­ing in the wil­der­ness, if you will,” said the NR­CC’s Simms. “I think where the Demo­crats have ex­celled, par­tic­u­larly in [2012] and even [2008], is cre­at­ing the in­fra­struc­ture to truly util­ize the data and have the data drive their de­cision-mak­ing pro­cess for everything the cam­paign does. And that’s our goal, that’s our mo­tiv­a­tion. So the stuff we’re do­ing with polling, for ex­ample, is start­ing with the data side, where we’re go­ing to mod­el our dis­tricts in-house, which is go­ing to af­fect how we poll them, which is go­ing to af­fect how we mi­cro-tar­get them, which is go­ing to af­fect how we run the cam­paigns. So, for us, the ul­ti­mate ob­ject­ive is for all of our can­did­ates to run the best, most data-driv­en cam­paign they can run. And I think the Demo­crats really ex­celled at that last cycle and em­braced it. And I’m not sure that we did.”

Set­ting up the ne­ces­sary in­fra­struc­ture is easi­er if you’re a pres­id­ent with four years to plot your reelec­tion cam­paign. Since the even­tu­al 2016 GOP pres­id­en­tial nom­in­ee will have less than a year to build a cam­paign ap­par­at­us, the role as change agent has fallen to the RNC and NR­CC. And — so far at least — their in­volve­ment has some Re­pub­lic­ans cau­tiously op­tim­ist­ic about their party’s chances to turn things around, des­pite the dif­fer­ences among its con­sult­ant class. “Only crisis com­pels change,” Mc­Cle­ary said.

What We're Following See More »
Rosenstein Holds Presser On Russian Indictments
2 days ago
U.S. Indicts 13 Russian Nationals For Election Interference
2 days ago

The indictment, filed in the District of Columbia, alleges that the interference began "in or around 2014," when the defendants began tracking and studying U.S. social media sites. They "created and controlled numerous Twitter accounts" and "purchased computer servers located inside the United States" to mask their identities, some of which were stolen. The interference was coordinated by election interference "specialists," and focused on the Black Lives Matter movement, immigration, and other divisive issues. "By early to mid-2016" the groups began supporting the campaign of "then-candidate Donald Trump," including by communicating with "unwitting individuals associated with the Trump Campaign..."

Gates Said to Be Finalizing a Plea Deal
2 days ago

"Former Trump campaign adviser Rick Gates is finalizing a plea deal with special counsel Robert Mueller's office, indicating he's poised to cooperate in the investigation, according to sources familiar with the case. Gates has already spoken to Mueller's team about his case and has been in plea negotiations for about a month. He's had what criminal lawyers call a 'Queen for a Day' interview, in which a defendant answers any questions from the prosecutors' team, including about his own case and other potential criminal activity he witnessed."

Another Defeat for Immigration Legislation in the Senate
3 days ago

"The Senate on Thursday rejected immigration legislation crafted by centrists in both parties after President Trump threatened to veto the bill if it made it to his desk. In a 54-45 vote, the Senate failed to advance the legislation from eight Republican, seven Democratic and one Independent senators. It needed 60 votes to overcome a procedural hurdle. "

House Intel Panel Could Charge Bannon with Contempt
3 days ago

"The House Intelligence Committee has scheduled a Thursday meeting to hear testimony from Steve Bannon—but it's an open question whether President Donald Trump's former chief strategist will even show up. The White House sent a letter to Capitol Hill late Wednesday laying out its explanation for why Trump's transition period falls under its authority to assert executive privilege, a move intended to shield Bannon from answering questions about that time period." Both Republicans and Democrats on the committee dispute the White House's theory, and have floated charging Bannon with contempt should he refuse to appear.


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.