Polling predictors reflect after missing the mark by a mile

Right up until the night before election day, polling institutes failed to predict the opposition Minjoo Party’s victory over the Saenuri Party.

Realmeter predicted one day before the election that the Saenuri Party would win 155 to 170 seats, the Minjoo Party 90 to 105 seats, the People’s Party 25 to 35 seats, the Justice Party 5 to 10 seats and independents 8 to 12 seats. But according to the National Election Commission, the Minjoo Party won 123 seats in the 300-member National Assembly and the Saenuri Party won 122 seats. The People’s Party won 38 seats, the Justice Party six and independents took 11 seats.

None of the 934 polling results submitted by various polling institutes to the National Fair Election Survey Deliberation Commission by March predicted the Minjoo Party’s win.

“I apologize on behalf of the polling institutes,” said Lee Taek-soo, chairman of Realmeter Research Center, on his Facebook page Thursday. “The fact is, it has become increasingly difficult to conduct election polls by calling landline telephones, and even exit polls failed to predict correctly the party to win the largest in the election.”

The last time polls varied this much from the vote results was in the general election of 2012, when most polling results predicted that the then-main opposition Democratic United Party (DUP) would win a majority over the Saenuri Party, but in the end, Saenuri won the majority with 152 seats, while the DUP won 127. Cases where candidates were elected contrary to poll predictions were also common in the election.

In central Seoul’s Jongno District, many polls announced that Saenuri Party’s Oh Se-hoon would win the district over Minjoo’s Chung Sye-kyun. But Chung outstripped Oh by 12 percent from early on in the vote counts.

In Suncheon of South Jeolla, polls predicted that Lee Jung-hyun of the Saenuri Party was likely to lose to Minjoo’s Roh Kwan-kyu. But Roh lost the election in the district.

“For future elections and polling, I propose using the anonymous phone number system, because collecting polls via landline telephones is no longer viable,” Realmeter Chairman Lee added in his Facebook post.

In the anonymous phone number system, polling institutes would be able to call citizens on their cellphones with alternative numbers provided by the government, in order to protect the identities of the voters. Only political parties are allowed to conduct polls in this way, and election law does not allow the parties to publicize the polls.

“Polling institutes are barred from getting the cellphone numbers of the voters, because doing so would violate the Personal Information Protection Act,” said Kim Chul-eung, president of the Korean Statistical Society.

“Polling institutes using automatic-answering-systems dropped the overall response rate of voters,” added Kim Young-won, professor of statistics at Sookmyung Women’s University.

In the elections so far, polling institutes conducted surveys on the voters by calling them on landline phones. The institutes used random-digit-dialing (RDD) method to reach the households and the voters. Less than 50 percent of Seoul area residents have registered landline phone numbers.

“Many have done away with landline phones, and even those who answer are usually housewives and the elderly,” said Park You-sung, professor of statistics in Korea University. “These polls based on landline phones no longer represent the voters.”

BY LEE SO-AH, RHEE ESTHER [chung.juhee@joongang.co.kr]