A Network Analytic Approach to Audience Behavior and Market Structure: The Case of China and the United States

Abstract

This study adopts a network analytic approach to understand media audiences in relation to media markets, bridging the literature on audience behavior and media economics. Using audience data in the Chinese and U.S. markets, we apply multi-level measures to compare audience fragmentation patterns, a key indicator of market structure, across television channels. Drawing on McQuail's four–stage fragmentation model, we find the Chinese television market exhibits the Core-Peripheral model where a few channels dominate the marketplace and the rest are viewed by niche segments of the audience. In contrast, the U.S. market represents the Pluralism model with extremely high levels of audience duplication across channels, suggesting overlapping patterns of exposure throughout the market rather than isolated segments.

Notes

¹Each province in China has its own local market in which numerous other provincial channels and the municipal television services operate and compete against each other locally. These local markets take up one third of the total advertising market shares in China leaving CCTV channels and nationally available provincial satellite channels the rest. Conceivably, these local markets tend to segment audience flow between channels in the national market.

²The larger standard deviation in the centrality scores for the Chinese network supports this interpretation. In the U.S. network, we see much less variance indicating greater equality in centrality scores across all nodes. This analysis also sheds light on the density comparisons. The mean scores are computed from normalized centrality scores, representing the percentage of possible links that are present for each given node. The extremely high mean value for the U.S. network is consistent with the high overall density score, whereas the lower mean value is consistent with a lower density score in the Chinese network.

³This is not a conventional t test as the measures are not independent—networks by definition are dependent. To clarify, the mean centrality scores are compared to an empirical sampling distribution created with 10,000 random permutations of the network. The magnitude of the test statistic, conservative significance level, and the consistency with the differences in centralization and density scores suggests that we can accept the difference as significant, albeit with some reservations.

Additional information

Notes on contributors

Elaine Yuan

Elaine Yuan (Ph.D., Northwestern University, 2007) is an Associate Professor in the Department of Communication at University of Illinois at Chicago. Her research interests include social implications of new media and audience research.

Thomas Ksiazek

Thomas Ksiazek (Ph.D., Northwestern University, 2010) is an Assistant Professor in the Department of Communication at Villanova University. His research interests include digital media use, audience behavior, journalism studies, and network analysis.