Structural Bias Coding Protocol
This instrument operationalizes media bias as a set of observable, codeable omissions and asymmetries. It does not ask whether a story "feels" biased. It asks whether the structural conditions that produce distorted coverage — press-release dependence, PR mediation, resource inequality, coaching asymmetry, false balance, and suppressed context — are present, absent, or unverifiable. Inter-rater reliability (IRR) controls are built into every item.
Each construct below is defined empirically, not impressionistically. Coders apply observable criteria, not intuition. Contested codes must be resolved before final scoring.
Press-release dependence
Coded as present when ≥50% of factual claims in the story can be traced to institutional communications, PR-drafted materials, or official handouts without independent corroboration.
Official-source saturation
Proportion of named/attributed sources who are institutional spokespersons, government officials, or corporate representatives vs. those with independent or adversarial standing.
Resource asymmetry
One or more parties in a binary-framed story has materially unequal access to legal counsel, crisis communications, public relations infrastructure, or research capacity relative to another party.
Media-training asymmetry
One party has demonstrably more experience with media interviews, coached messaging, on-camera preparation, or spokesperson infrastructure than the other. Inferred from institutional capacity and prior media history.
Legal intimidation capacity
One party has access to legal threats, NDA enforcement, or litigation strategy that constrains the other party's ability to speak publicly or fully.
Binary scaffolding
The story imposes a two-sided frame on a situation involving structural hierarchy, power differentials, or more than two materially distinct positions, creating a false impression of equality between parties.
False balance / bothsidesism
Both "sides" are given equivalent rhetorical weight despite significant empirical, institutional, or evidentiary asymmetry between them. Distinct from legitimate balance when parties are genuinely comparable.
PR firm / lobby mapping
Whether the intermediary communications infrastructure shaping the public narrative has been identified, disclosed, or investigated. Unmapped intermediaries make source evaluation impossible.
Asymmetry disclosure
Whether the story disclosed to the audience any relevant power, resource, or experience asymmetries that would allow readers to evaluate the relative credibility of sources.
Omission of suppressed context
Presence of relevant prior settlements, NDAs, lobbying history, funding relationships, or institutional conflicts of interest that the story failed to surface or acknowledge.
Headline / lede framing
Whether headline and opening paragraph establish a frame that systematically favours one interpretation of contested facts before the body of the story presents evidence.
Source diversity index
Whether sources represent a range of institutional positions, power levels, and proximity to the issue, or whether sourcing is clustered within one type of actor (e.g., government only, corporate only).
| Construct | Operationalization | Scale | IRR method |
|---|---|---|---|
| Press-release dependence | Count of claims traceable to institutional PR material ÷ total factual claims | 0–3 + evidence | Cohen's κ (≥0.70) |
| Source power distribution | Ratio of official/institutional sources to independent/adversarial sources | 0–3 + evidence | Cohen's κ (≥0.70) |
| Resource asymmetry | Qualitative assessment of legal, PR, financial, and experiential capacity gap between parties | 0–3 + evidence | Consensus coding + documentation |
| Media training gap | Inferred from institutional type, prior media presence, spokesperson infrastructure | 0–3 + evidence | Consensus coding + documentation |
| False balance | Presence of equivalent rhetorical weight despite documented empirical asymmetry | 0–3 + evidence | Cohen's κ (≥0.70) |
| Asymmetry disclosure | Binary: disclosed / not disclosed, with degree noted | 0–3 + evidence | Cohen's κ (≥0.70) |
| Suppressed context | Absence of verifiable prior record, NDA, lobbying, or conflict-of-interest information | 0–3 + evidence | Dual-coder + adjudication |
| Headline framing | Direction and content of frame established in headline and lede; pre-coded frame typology applied | 0–3 + evidence | Cohen's κ (≥0.70) |
Inter-rater reliability benchmark: Cohen's κ ≥ 0.70 for all quantitative items. Items falling below this threshold require codebook revision and re-coding before analysis. Percent agreement alone is not sufficient — chance agreement must be corrected (Cohen, 1960; Krippendorff, 2004). For items requiring consensus coding, all disagreements must be documented and resolved before the record is finalized. See Landis & Koch (1977) for κ interpretation benchmarks used here.