"I'll Do the Work, Just Give Me 20 Million Won"…'Secret Temptation' Public Opinion Poll [Broken Wind Vane]②

Public Opinion Manipulation Can Occur During Polling Process
Various Methods Like Number Shuffling and Weight Manipulation
Trustworthiness of Public Opinion Polls Declines

Editor's NoteThe suspicions surrounding 'political broker' Myeong Taegyun are shaking the political world. Claiming to be a public opinion poll expert, Myeong approached power by utilizing polls as a barometer of public sentiment. In the process, he is suspected of manipulating polls. Taking this opportunity, we examine the reality and countermeasures of poll manipulation in three parts. ① Myeong Taegyun's approach to power through polls ② How was the manipulation carried out?

# A parliamentary aide, Mr. A, received a 'suspicious proposal' from a polling company ahead of an election. The company asked for 20 million won to perform a 'refinement process.' What exactly did they mean by 'refinement'?

Having experienced several elections, aide Mr. A recalled a proposal he once received from a polling company amid the controversy over Myeong Taegyun's poll manipulation. A representative of the company, claiming to have about 30 million data points, tempted him by saying they could run polls using respondents with high response rates to produce favorable results. There was also a proposal to classify only supporters of a specific political party. Mr. A, who declined the offer, recalled, "During elections, people can be easily swayed due to their desperate feelings."

In an era of political polarization where people hesitate to express political opinions, polls serve as a crucial channel to gauge public sentiment. Depending on the results, a political hopeful can become a frontrunner, or an established politician can fall to an uncompetitive candidate. As political parties actively use polls for nominations and other purposes, the importance of polls has increased. Consequently, the risk of distortion and manipulation has also risen.

What are the methods of poll manipulation?

According to insiders in the polling industry and political circles, poll results can vary depending on pre-survey stages such as so-called 'number mixing' and 'question and order effects.'

'Number mixing' is a method where polling companies conduct surveys by mixing phone numbers they already have with 'safe numbers'?virtual numbers provided by the Central Election Poll Deliberation Commission (CEPDC) received from telecom companies?or numbers prepared through random digit dialing (RDD). The mixed numbers may include 'verified numbers' whose political leanings are already known or mobilized individuals such as family members of candidates whose support is confirmed. When the sampling population itself is manipulated in this way, polls can yield results close to the desired conclusion. The polling industry regards such acts as criminal offenses.

Although monitoring of polls is conducted to prevent such manipulation, there are sufficient incentives to attempt it. The surveillance network is loose, and penalties are lenient. According to Article 108 of the current Public Official Election Act, election-related poll data must be stored for six months. However, after this period, disposing of the data does not violate the law. After six months, evidence of manipulation disappears. Moreover, unless the CEPDC requests data submission, the polls are not subject to verification. A polling official said, "(Whether manipulation occurred) can only be known by examining the raw data," adding, "Experienced fraudsters have ample opportunity to tamper." A CEPDC official also stated, "We monitor polls, but during election periods, many polls are conducted, and some are registered even 30 minutes before publication, so some cases may be missed."

'Question effect' and 'question order effect' are not illegal but can influence election polls. The industry often attributes these issues to surveyors' mistakes or lack of skill, but since public opinion can change depending on question design, management is necessary. For example, if a question that certain political-leaning respondents might dislike is inserted, those respondents may hang up the phone, effectively excluding similar-leaning respondents from the poll.

Additionally, polls conducted only on some candidates under the guise of hypothetical two-way matchups or including 'unification-related questions' in candidate support surveys can influence results. When unification-related questions are included, support rates among supporters of the unified candidate rise, but supporters of unrelated candidates may hang up. In the last presidential election, even with almost identical survey designs, differences in such questions led to drastically different poll results. Kim Bongshin, director of Metavoice, argues, "Current CEPDC regulations on poll questions are mechanical," and calls for "a contextual review system for questions."

There are also concerns that question order affects results. For instance, in party support surveys, whether questions about presidential approval come before or after party support questions can change poll outcomes. If the president is mentioned first, respondents answer with awareness of the ruling party, but if presidential questions come later, the poll purely reflects inter-party opinion.

Manipulation possibilities remain even after polls are conducted. The industry points to methods of tampering with poll results post-survey and manipulating weighting values. Polls with low response rates often have respondent compositions that differ significantly from actual population ratios. To correct this, a process called 'weighting' adjusts the data to match population ratios. According to CEPDC's 'Election Polling Standards,' weighting adjustments are made by gender, age, and region, with limits set between 0.7 and 1.5 times. However, considering significant opinion differences between regions and generations, arbitrary adjustment of weights can lead to different poll outcomes.

Polls Facing a Crisis of Credibility

Regardless of manipulation, doubts have grown about whether polls truly represent public opinion.

Until now, the polling industry and academia have analyzed that respondents' attitudes vary depending on the polling agency. Known as the house effect, it was believed that the commissioning organization's political leaning influences respondents. Depending on whether the client is a conservative or progressive media outlet, respondents' attitudes may differ. Recently, however, attention has shifted more to polling companies themselves, viewing them not just as survey contractors but as political actors. A representative example is 'Yeoronjocha Kkot,' known for its pro-opposition stance. Conservative supporters show strong resistance to polls conducted by this company, operated by Kim Eo-jun, raising concerns about the house effect.

Not only companies but also mobilizing respondents can distort polls. For example, in party nomination-related polls, candidates ask supporters to answer calls starting with '02-XXX-XXXX' or to accept calls from '02 or 070' numbers via text messages. Considering low response rates, such encouragement alone can have a significant mobilization effect. Going further, some candidates have been prosecuted for inciting supporters to provide false responses in specific low-response age groups.