여론조사는 현대 정치 선거 과정에서 필수 불가결한 요소로 간주된다. 그런데, 언론이 조사를 취사선택하여 여론을 재구성하는 과정에서 객관성과 공정성의 원칙을 준수하는지에 대해서는 여러 차례 의문이 제기되었다. 본 연구는 언론이 정파성에 따라 특정 여론조사를 더 높은 확률로 보도하는지, 만약 그렇다면 그러한 편향성의 효과가 전사회적 차원에서 어느 정도 수준에 해당하는지 알아보고자 하였다. 이를 위해 제20대 대통령 선거운동 기간 이루어진 311개 여론조사 결과를 보도한 4,020건의 언론 기사를 분석하였다. 본 연구는 언론이 정파성에 따라 여론조사 결과를 선택적으로 보도하는 가설적인 상황을 ‘선택적 보도’ 행위로 명명하고, 이를 조사기관 편향에 따른 선택적 보도와 조사 결과에 따른 선택적 보도로 나누어 그러한 보도 편향이 실재하는지 알아보았다. 베이지안 상태공간 모형과 계층적 푸아송 회귀를 이용한 분석 결과, 조사기관의 편향, 조사 결과에 나타난 격차가 보도 확률과 유의미한 연관이 있었음에도 불구하고, 이를 언론사의 당파성에 따른 선택적 보도라고 보기는 어렵다는 결론에 도달하였다. 본 연구에서 제시된 분석 결과와 분석 방법은 언론과 여론조사의 신뢰 회복을 위한 향후 연구에 도움을 줄 수 있을 것으로 기대된다.
Public opinion polls are essential tools in modern democratic elections, serving as a barometer of public sentiment and influencing political discourse. Despite their importance, concerns are growing regarding the objectivity and fairness of how news media report on these polls. This study investigates whether Korean news outlets selectively report election poll results, focusing on the interplay between polling institutions’ biases, polling outcomes, and media coverage during the 20th presidential election. It evaluates whether such selective practices are ideologically motivated and examines their implications for public trust in media and polling institutions. The analysis is based on a dataset of 4,020 news articles covering 311 polls conducted during the election period. Employing Bayesian state-space models and hierarchical Poisson regression, the study explores correlations between polling institutions’ “house effects” (systematic biases favoring particular candidates), poll outcomes, and the frequency of media reporting. Selective reporting is defined as the strategic inclusion or exclusion of poll results, driven by factors such as media preferences, organizational limitations, or structural constraints. The findings reveal significant relationships between institutional biases and reporting frequencies but do not provide conclusive evidence of partisan motives behind these patterns. Instead, the results suggest that other dynamics, including journalistic routines and resource availability, may play a larger role. Key observations include: (1) Polls with pronounced house effects tend to receive greater media attention, although the direction of reporting does not consistently align with the ideological leanings of media outlets. (2) Large national outlets, including prominent newspapers and broadcasters, demonstrate broader and more balanced reporting practices. Conversely, regional newspapers show greater reliance on specific biased sources, reflecting potential resource constraints or strategic focus. (3) The impact of poll outcomes, such as the margin between candidates, on reporting frequency is minimal, suggesting that selective reporting may be influenced more by institutional dynamics than by poll results alone. This research contributes to the study of media bias in electoral contexts by utilizing advanced Bayesian methods to analyze complex, multi-layered data. By quantifying the relationships between poll characteristics and media behavior, it provides a nuanced understanding of how structural and operational factors influence reporting practices. These findings challenge traditional narratives that attribute selective reporting solely to ideological biases, emphasizing instead the broader systemic factors at play. In an era of information abundance and growing skepticism toward both media and polling institutions, transparency and accountability are critical for maintaining public trust and ensuring informed democratic participation. The study highlights the importance of fostering institutional integrity and calls for enhanced standards in both polling methodologies and media practices. Moreover, it provides a methodological foundation for future comparative research across diverse media and electoral systems, underscoring the value of interdisciplinary approaches in understanding evolving information ecosystems. By shedding light on the multifaceted nature of selective reporting, this study encourages policymakers, media organizations, and researchers to promote fairness and objectivity in electoral coverage, safeguarding the integrity of democratic processes.