Qualitative Content Analysis

Manifest content analysis focuses on explicit, surface-level meanings: what is directly stated or shown. For example, counting the frequency of certain words or explicit statements (e.g., 'the word pain appears 47 times'). Latent content analysis interprets underlying or implicit meanings—what the content suggests, implies, or reveals about values, attitudes, or worldviews. For example, analyzing how media portrayals of obesity implicitly convey moral judgment or shame. Manifest analysis is more objective; latent analysis requires interpretive judgment. Most QCA combines both: manifest categories (what is said) with latent interpretation (what it means).

Both identify patterns in qualitative data, but differ in structure and purpose. Thematic analysis identifies broad themes or patterns without necessarily creating a rigid coding scheme; it is more inductive and exploratory. Content analysis creates an explicit, systematically applied coding scheme with defined categories and decision rules; it is more structured and allows frequency counting. Content analysis is better for large datasets requiring consistency; thematic analysis is often faster for exploratory research. Some consider TA a type of content analysis; others view them as distinct. Use TA for exploratory pattern-finding; use QCA for systematic, transparent categorization with frequency data.

Develop a clear definition for each category with inclusion/exclusion criteria. Have at least one second coder independently code a subset of data (10–20% minimum). Calculate inter-rater reliability: percent agreement for simple schemes, Cohen's kappa or Krippendorff's alpha for more complex ones (target >0.70). Resolve disagreements through discussion, refining category definitions if needed. Regular coder meetings throughout analysis maintain consistency. Document your coding decisions and reasoning (audit trail). Validity: ensure your categories logically relate to your research question and provide meaningful interpretation of patterns. Pilot-test your scheme on a small sample before full analysis.

Yes. Purely inductive QCA develops categories from data. Purely deductive QCA applies predetermined categories from theory or prior research. Many studies use hybrid approaches: start with deductive categories grounded in theory, then add inductive codes that emerge from the data. This can strengthen analysis by combining theoretical rigor with data-responsiveness. Be explicit about which approach (or combination) you used and justify your choice.

Initial inductive coding often produces 30–100+ codes from 15–20 interviews or documents, which are then grouped into 10–25 categories. The goal is not to minimize codes but to create a coherent hierarchy: specific codes nested within broader categories. Merge codes when they represent the same concept or produce redundancy (no purpose in keeping them separate). Split categories if they are too broad and contain heterogeneous meanings (e.g., 'barriers' might split into 'structural barriers' and 'individual barriers'). Audit your scheme: if a code appears in only one document, question whether it deserves a code or is an outlier.