Poll sur­veys: Be­liev­ing the hype

Ex­perts weigh in on the sci­ence of po­lit­i­cal polling

The Covington News - - News -

If you be­lieve the polls, it looks like one pres­i­den­tial can­di­date has moved ahead of the other and might be head­ing to­ward a victory on Nov. 4. But can you be­lieve the polls? No­body wants to re­peat the mis­take of 1948, when voter sur­veys in­di­cated Thomas Dewey eas­ily would de­feat in­cum­bent Pres­i­dent Tru­man. The me­dia fell for it, lead­ing to one of the most em­bar­rass­ing head­lines in news­pa­per his­tory. But Rich Engstrom, as­sis­tant pro­fes­sor of po­lit­i­cal sci­ence at Ge­or­gia State Uni­ver­sity, said such a blun­der wouldn’t hap­pen to­day. “(Polling) is re­ally a sci­ence now,” he said. “We can trust polls, but they al­ways have to be taken with a grain of salt.” Maybe with more than a few grains of salt this year. The pres­i­den­tial race be­tween Repub­li­can John McCain and Demo­crat Barack Obama has been so un­pre­dictable that there is no his­tor­i­cal prece­dent. There will be more reg­is­tered vot­ers this year than ever be­fore, and some will be­long to de­mo­graphic groups that haven’t voted reg­u­larly in the past. Poll­sters may in­ter­view th­ese peo­ple about their opin­ions, but they can’t pre­dict how many ac­tu­ally will show up on Nov. 4 to vote. And be­cause this group trends to­ward younger peo­ple, some have ques­tioned whether the views of th­ese new vot­ers are re­flected ac­cu­rately in the polls. The rea­son: At least 15 per­cent of Amer­i­can adults have a cell phone but no land­line phone. Tra­di­tion­ally, poll­sters only have called peo­ple who have land­line phones, so their sur­vey sam­ple may have skewed to­ward older, more af­flu­ent vot­ers. Ac­cord­ing­tothePewRe­search Cen­ter, peo­ple who have only a cell phone are more likely to be young, male, non­white and lean to­ward vot­ing Demo­cratic. That sounds like a group that would fa­vor Obama. But a Pew study last sum­mer com­pared cell-only vot­ers with a “weighted” sam­ple of land­line vot­ers and found no sig­nif­i­cant dif­fer­ence. Engstrom said poll­sters weight the data by mak­ing ad­just­ments to cor­rect any po­ten­tial bi­ases. But th­ese as­sump­tions are based on fac­tors that have been learned from pre­vi­ous elec­tions. “All bets are off if, all of a sud­den, a whole new de­mo­graphic group shows up to vote that never voted be­fore,” Engstrom said. Henry “Chip” Carey, as­so­ciate pro­fes­sor of po­lit­i­cal sci­ence at Ge­or­gia State, said the con­cern about cell phones has been overblown. If the sur­veys miss some cell phone users, he said, it won’t nec­es­sar­ily lead to in­ac­cu­rate pre­dic­tions of elec­tion re­sults. “Most kids have cell phones,” Carey said. “But kids are less likely to vote.” All this hand-wring­ing over cell phones may be a waste of time, ac­cord­ing to Matt Tow­ery, chief ex­ec­u­tive of­fi­cer of In­sider Ad­van­tage, a Ge­or­gia-based polling firm. “The big­gest ur­ban myth of the year is that we’re not able to poll cell phones,” he said. “It’s not true. (Cell phones) have no ef­fect on polling what­so­ever.” Tow­ery said cell phones are only a hin­drance if you’re re­ly­ing on the phone book to find phone num­bers. Most poll­sters don’t do this any­more. “We can get rep­re­sen­ta­tive phone num­bers from all age groups,” he said. “We buy data from com­pa­nies that col­lect names from voter regis­tra­tion rolls.” He said it is le­gal to do this for the pur­pose of sci­en­tific re­search, in­clud­ing po­lit­i­cal sur­veys. Tow­ery said poll­sters do “over­sam­ple” young peo­ple, mean­ing they at­tempt to call a larger pool of in­di­vid­u­als. “But that’s not be­cause of cell phones,” he said. “It’s just be­cause it’s dif­fi­cult to reach them.” He added that most polling firms now use au­to­mated sur­veys, with a com­put­er­ized voice in­stead of a live per­son at the other end of the line ask­ing the ques­tions. “It’s much more ac­cu­rate,” Tow­ery said. That’s be­cause peo­ple tend to tai­lor their re­sponses to how they per­ceive the sur­veyor. For ex­am­ple, if they think the per­son ask­ing the ques­tions is AfricanAmer­i­can, they might an­swer ques­tions about Obama dif­fer­ently than they would if they thought they were speak­ing to a white per­son. And that brings up an­other pos­si­ble vari­able in the pres­i­den­tial race. In re­cent weeks, po­lit­i­cal pun­dits have been de­bat­ing

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.