What content analysis can tell us
Political speeches, party platforms, legislative debates, and policy documents contain more material than a reader can easily compare from memory. Content analysis is a way to examine that material using explicit, repeatable rules. A researcher might ask how often speakers discuss housing, how a document describes a policy problem, or whether a set of speeches emphasizes government action or individual responsibility.
The method can combine counting with close reading. A count may show that a theme appears more often in one collection of documents than another. Reading the passages can help explain what the theme means in each setting. Neither step, on its own, reveals a speaker’s private motives or proves how an audience understood the message.
Start with a question and a defined collection
A useful project begins with a question narrow enough to guide decisions. Instead of asking, “What do politicians say about the economy?” ask how a specified group of speeches describes the causes of unemployment. Define which speakers or institutions are included, what kinds of texts qualify, and the period covered. State whether you are examining prepared remarks, transcripts of spoken words, official documents, or a mixture of these. A prepared speech and a question-and-answer exchange may not be directly comparable.
Decide how texts will enter the collection before looking for a preferred pattern. You might include every eligible document within a defined period or select documents using a stated rule. Record where each text came from and note missing or incomplete items. If only easily accessible speeches are included, the findings describe that collection, not necessarily everything the speakers said.
Build a codebook that another reader could use
A codebook lists the categories to be recorded and explains when each applies. First choose a unit of analysis: an entire document, a paragraph, a sentence, or a distinct claim. The unit should match the question. Document-level coding can identify broad themes, while sentence-level coding may capture how a policy is framed. Set rules for quotations within speeches, repeated slogans, headings, and passages that address several topics.
Make each category concrete. A code called “economic concern” is vague; a rule such as “mentions jobs, wages, prices, or household costs as a public problem” gives a coder more guidance. Include examples and borderline cases. Specify whether a passage can receive multiple codes. For instance, a sentence about housing costs and wages may need both “housing” and “employment,” unless the project requires a single primary topic.
Test the codebook on a small portion of the collection, then revise unclear definitions before coding the full set. Keep a record of changes so that earlier passages can be reviewed under the final rules. Categories should serve the research question rather than force every interesting observation into a count; an “other” code and brief notes can preserve material the initial scheme did not anticipate.
Check consistency without mistaking it for certainty
If more than one person is coding, give them the same instructions and ask them to code some of the same material independently. Compare their decisions and discuss disagreements. If coders repeatedly differ over whether a passage names a problem or proposes a solution, the codebook may need sharper definitions. Even when working alone, a researcher can recode a portion later and inspect changes in judgment.
Agreement matters, but it does not guarantee that the categories capture the intended idea. Coders may consistently apply a poor definition. To assess whether the codes are useful, return to the original question and read passages assigned to each category. Ask whether they genuinely illustrate the concept and whether important counterexamples were missed. Consistency and meaningful interpretation are related, but distinct, goals.
Interpret the results in context
When comparing counts, pay attention to the size and type of each text. Ten mentions in a long policy document do not have the same meaning as ten mentions in a short address. Depending on the question, report both raw counts and a measure relative to document length, and explain what each measures. An absent term does not always mean an absent idea: speakers may use synonyms, implication, or references that depend on surrounding context.
Present findings with enough detail for readers to understand the choices behind them: the collection, inclusion rules, unit of analysis, code definitions, and how disagreements were handled. Use short passages where helpful to show how a code works, while avoiding quotations that distort the larger text. Content analysis is most informative when its conclusions remain proportional to its evidence: it can identify patterns in the material examined, while leaving room for other texts, interpretations, and explanations.
