In Buk-gu Gap, Busan, where a National Assembly by-election is underway, two polls were conducted on the 1st to 3rd. In a survey commissioned by SBS to polling firm Ipsos, support stood at Ha Jung-woo of the Democratic Party of Korea (38%), Park Min-sik of the People Power Party (26%), and independent Han Dong-hoon (21%), in that order.
By contrast, a survey commissioned by Busan MBC and conducted by Hangil Research showed a significant difference in both support levels and rankings: Ha Jung-woo (34.3%), Han Dong-hoon (33.5%), and Park Min-sik (21.5%). Why did such differences arise in polls conducted in the same area during the same period?
◇ In ARS surveys, conservative support comes out higher… "Shy conservatives respond more comfortably"
First, the two polls differed in methodology. The SBS–Ipsos poll was conducted with 100% wireless telephone interviews. By contrast, the Busan MBC–Hangil Research poll was 84.3% wireless ARS and 15.7% landline.
Wireless ARS is conducted by listening to an automated voice prompt and pressing buttons to answer. Because many people may hang up after hearing only a robotic voice, the response rate is relatively low. There is a greater chance that the voting preferences of those highly interested in politics, who do not hang up even after hearing a robotic voice, are more reflected.
There is also an analysis that ARS surveys reflect conservative voting preferences more. That is because respondents can reveal conservative leanings more comfortably than in telephone interviews where an interviewer asks directly. A polling expert said, "Since the martial law incident, 'shy conservative' voters who hide their conservative leanings tend to respond more comfortably to ARS surveys, which are not conducted by a person."
In fact, in the Busan MBC–Hangil Research poll conducted via ARS, the combined support for conservative-leaning candidates (55%) was higher than in the SBS–Ipsos poll that used telephone interviews (47%).
The two polls also differed widely in contact rates. The contact rate refers to the ratio at which an interviewer actually reaches a respondent when calling. A poll with a low contact rate can be seen as one where respondents with high interest in politics are more likely to have answered. Conversely, a poll with a high contact rate can be seen as one where even those less interested in politics may have answered.
In the Buk-gu Gap poll conducted by SBS–Ipsos, the contact rate was 25.3% and the response rate was 14.4%. For Busan MBC–Hangil Research, the contact rate was 44.5% and the response rate was 5.3%. A political insider said, "A majority of voters who actually go to the polls are those less interested in politics," adding, "It is important to read where the so-called hidden vote is headed."
◇ "Polls do not necessarily match election results" "Some answers differ from true intentions"
Polling experts say poll results should be used only as an indicator to read overall trends and should not be overinterpreted.
Cho Jin-man, a professor in the Department of Political Science and International Relations at Duksung Women's University, said, "Poll results do not necessarily match election outcomes," adding, "It is enough to take only the results of a few credible institutions as reference."
Um Kyung-young, head of the Zeitgeist Research Institute, said, "Poll results are not data," adding, "If people answered differently from their true intentions or avoided answering, it cannot become data."
A political insider also said, "Poll results that come out within or around the margin of error early in a race have little significance."
Meanwhile, the SBS–Ipsos poll surveyed 503 men and women aged 18 or older living in Buk-gu. The sampling error is ±4.4 percentage points at a 95% confidence level. The Busan MBC–Hangil Research poll surveyed 584 men and women aged 18 or older living in Buk-gu. The sampling error is ±4.1 percentage points at a 95% confidence level. For details, refer to the website of the National Election Survey Deliberation Commission.