Network Agenda-Setting
- Introduced by
- Lei Guo & Maxwell McCombs
- Year
- 2011
- Type
- Network-analytic extension of agenda-setting theory
- Family
- Agenda-setting research
What does it do?
Network agenda-setting (NAS), also called third-level agenda setting, extends classic agenda-setting theory by proposing that news media transfer to the public not only the salience of issues (first level) and of attributes (second level), but the very web of associations among issues and attributes.
Introduced by Lei Guo and Maxwell McCombs, the method represents the media agenda and the public agenda as networks and tests whether the media's bundling of elements is reproduced in the public's mind.
The intuition behind the method
Traditional agenda setting asks whether the media make audiences think certain issues are important. The network model adds: the media also teach audiences which things go together. If coverage repeatedly mentions an economic issue alongside a particular candidate, audiences come to associate the two as well. NAS captures this by drawing two networks — one of how elements co-occur in news coverage, one of how they co-occur in people's minds (from surveys) — and asking how similar the two network structures are. A high correspondence means the media transferred not just salience but the relational picture, the bundle of connections.
History
Maxwell McCombs and Donald Shaw's 1972 Chapel Hill study launched agenda-setting research by showing that issue salience in news predicted public issue salience. Attribute (second-level) agenda setting followed in the 1990s. Around 2011, Lei Guo and McCombs proposed the network agenda-setting model as a third level, arguing that media transfer relational structures, not just isolated salience; Guo's 2012 methodological article in the Journal of Broadcasting & Electronic Media laid out how to apply social network analysis and QAP to the problem. Subsequent studies extended NAS to elections, social media, and cross-platform agenda dynamics.
When to use it
Use network agenda-setting when your interest is in how bundles of issues or attributes are linked together in media coverage and in audience cognition, rather than the salience of single elements in isolation. It is appropriate when you can measure co-occurrence in both arenas — for example, content-analyzed news plus an open-ended or association survey — and when theory suggests the relational structure, not just the ranking, carries the effect. It assumes co-occurrence is a valid proxy for cognitive association and that the element set is comparable across arenas. It is unnecessary when first- or second-level salience rankings already answer the question, and it is unreliable when the public-agenda network is built from too few respondents or sparse association data.
Limitations and considerations
- Treats co-occurrence as a stand-in for cognitive association, an assumption that is plausible but not directly observed.
- Constructing a public agenda network requires association-type survey data that are harder to collect than salience rankings.
- Network correlation establishes correspondence, not causation; lagged or experimental designs are needed for directional claims.
- Results depend on how elements and co-occurrence windows are defined, and dense networks can yield inflated correlations.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Captures the relational structure of agendas — which elements are bundled — that ranking-based agenda-setting measures ignore.
- Unifies issue-level and attribute-level agenda setting within a single network framework.
- Uses QAP, a permutation test designed for the dependence in network data, giving statistically defensible matrix comparisons.
- Connects mass-communication theory to the well-developed toolkit of social and semantic network analysis.
Applications
- Testing whether the way news bundles candidates and issues during a campaign predicts how voters mentally associate them.
- Studying how attribute networks for a country, brand, or policy in coverage map onto public perceptions.
- Analyzing intermedia agenda setting by comparing the network structures of different news outlets or platforms.
- Examining agenda flow between social media and traditional news through network correspondence over time.
- Cross-cultural studies comparing media and public agenda networks across national media systems.
How it works
How it works
The analyst specifies the elements that form the nodes of both networks — these can be issues or candidates (object-level NAS) or attributes of a single object (attribute-level NAS) — and defines them consistently for the media content and the audience measure.
For each agenda, a co-occurrence weight a_ij is built by counting how often elements i and j appear together within the same unit — the same news story for the media agenda, or the same respondent's set of mentioned considerations for the public agenda.
Formula
a_{ij} = \#\{u : i \in u \wedge j \in u\}The co-occurrence counts populate two symmetric matrices, which are treated as weighted networks: the media agenda network and the public agenda network, each describing how strongly every pair of elements is linked in that arena.
Because the cells of a network are not independent, the two matrices are compared with the Quadratic Assignment Procedure (QAP), which correlates the vectorized matrices and tests significance by randomly permuting node labels thousands of times to build an empirical null distribution.
Formula
r_{QAP} = \mathrm{corr}\big(\mathrm{vec}(M_{\text{media}}), \mathrm{vec}(M_{\text{public}})\big)A significant, positive QAP correlation is interpreted as evidence of network agenda-setting: the relational structure of the media agenda is mirrored in the public agenda. Temporal designs (lagged media-to-public correlations) and individual-level extensions strengthen the inference about the direction of transfer.
Common pitfalls
- Comparing the two matrices with an ordinary correlation test instead of QAP, which ignores network dependence and overstates significance.
- Building the public network from too few respondents, producing an unstable structure that spuriously matches or mismatches the media network.
- Using mismatched element sets or co-occurrence windows across the two agendas, so the networks are not truly comparable.
- Reading a significant cross-sectional correlation as proof that media caused the public's associations without temporal ordering.
Frequently asked questions
How is network agenda setting different from second-level agenda setting?
Second-level (attribute) agenda setting concerns the transfer of salience for individual attributes of an object — making certain characteristics prominent. Network agenda setting concerns the transfer of the relationships among elements: which issues or attributes are linked to which others. The first asks whether an attribute is salient; the third asks whether the entire web of associations in the media is reproduced in the public's cognition, which network analysis is purpose-built to measure.
Why use QAP instead of a normal correlation to compare the networks?
The cells of a network matrix are not statistically independent — they share nodes — so the usual assumptions behind a Pearson correlation's significance test are violated, and ordinary p-values are too small. The Quadratic Assignment Procedure computes the matrix correlation but tests it against a null distribution built by randomly permuting node labels many times, which respects the network's dependence structure and yields valid significance levels.
How do you build the public agenda network?
Typically from open-ended or association survey data: respondents list the issues, attributes, or considerations they associate with a topic, and the analyst counts how often each pair of elements is mentioned together across respondents. Those co-occurrences form the weighted public agenda network, which is then compared to the media network built the same way from coverage. Adequate sample size and a well-defined element set are essential for a stable public network.
Related methods
Sources
Guo, L. (2012). The application of social network analysis in agenda-setting research: A methodological exploration. Journal of Broadcasting & Electronic Media, 56(4), 616–631.
Open DOI referenceCopied.McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36(2), 176–187.
Open DOI referenceCopied.