Large-scale quantitative evidence of media impact on public opinion toward China

The US-China relationship is marked by intense rivalry, with public opinion significantly influencing political action in the US. Since personal experience with China is limited for most Americans, media representations heavily shape their perceptions. This study focuses on The New York Times' role in shaping American attitudes towards China. While not causally conclusive, the analysis is grounded in a causal understanding of the flow of information from media to public opinion. Existing literature presents conflicting perspectives on the media-public opinion relationship. Some scholars emphasize media's formative influence on public opinion, while others highlight the public's influence on media through consumer demand and selective exposure. This study aims to contribute a "big data" perspective by analyzing a large dataset spanning 50 years of NYT reporting and survey data on American views of China. Public opinion surveys consistently show a generally unfavorable view of China among Americans, often linked to concerns about human rights, economic and military power, and the political system. However, a distinction is often made between the Chinese state, people, and culture, with the latter two viewed more favorably. The New York Times, as a highly influential newspaper, is chosen for its potential impact on elite and public opinion, despite its acknowledged biases.

The study reviews existing literature on the relationship between media and public opinion, noting the conflicting perspectives. Some studies emphasize media's ability to shape public opinion, while others focus on the influence of consumer demand and selective exposure in shaping media coverage. The authors note that these studies often rely on relatively small datasets. The existing research also highlights various factors influencing attitude formation, including personal experiences, social networks, and the influence of foreign opinions on American perspectives.

The study employed a large-scale quantitative approach. The core dataset consisted of 267,907 China-related articles from The New York Times, spanning from 1970 to 2019. These articles were analyzed using BERT, a deep neural network-based natural language processing model, to determine the sentiment (positive, negative, neutral) expressed towards China across eight topical domains: ideology, government administration, democracy, economic development, marketization, welfare and well-being, globalization, and culture. This media sentiment data was then compared with longitudinal data from 101 cross-sectional surveys of American public opinion toward China. A linear regression model was used to assess the relationship between media sentiment (in current and preceding years) and public opinion in a given year. The model included lagged values of media sentiment to account for the inertia in public opinion. Model selection involved constraints to ensure non-negative coefficients (assuming positive articles have positive/neutral impact and negative articles have negative/neutral impact) and sparsity (limiting the number of non-zero coefficients).

The analysis revealed a consistent interest in China by The New York Times throughout the period, with at least 3,000 articles published annually. Media sentiment exhibited variations across the eight topical domains. For instance, negative articles consistently outnumbered positive articles in areas like ideology, government, economy, and welfare, while positive sentiment was more common in domains such as globalization and culture. Regression analysis revealed a significant relationship between media sentiment and public opinion. Specifically, the sentiment expressed in The New York Times concerning Chinese culture and democracy in one year explained 53.9% of the variance in American public opinion on China the following year. Further analyses showed that incorporating additional media sentiment variables from different topics and years further improved the model's explanatory power, with eight topics and different time lags explaining 65.9% of the variance.

The study's findings support the hypothesis that media sentiment significantly influences public opinion on China. The strong explanatory power of The New York Times' coverage highlights the significant role of elite media in shaping public perceptions. The results, while correlational and not causally conclusive, suggest a substantial impact of media framing on public attitudes. The authors acknowledge potential limitations, such as the lack of detailed analysis of intermediary processes through which media influence public opinion, the potential influence of elite communication bypassing traditional media, and the limited scope of newspapers analyzed. The model's explanatory power, while substantial, does not encompass all factors affecting public opinion.

The study demonstrates a significant correlation between The New York Times' coverage of China and subsequent American public opinion. The use of a large dataset and advanced NLP techniques provides strong quantitative evidence for the impact of media framing on public perceptions of China. Future research could explore causal relationships more rigorously through experiments or incorporate broader media sources. Further investigation into the role of social media and elite direct communication in shaping public opinion would also enhance understanding.

The study primarily focuses on The New York Times, potentially limiting the generalizability of findings. The analysis is correlational, not causal, limiting definitive conclusions about the direction of influence. The study also acknowledges potential biases within the NYT's reporting and the influence of other factors on public opinion formation, such as social media and direct communication by political elites. Finally, while the model explains a substantial portion of the variance in public opinion, it doesn't capture all influential factors.