Should Qualitative Coding Be Inductive, Deductive, or Both?

Should Your Qualitative Codes Come From the Data or Existing Theory?

You are preparing to analyze interviews, observations, or documents. Should you develop codes by examining what participants say, begin with categories derived from existing literature, or use both?

This decision is often presented as a choice between inductive and deductive coding. Inductive coding is commonly described as allowing codes to emerge from the data, while deductive coding applies concepts identified in advance.

That distinction is useful, but it can be misleading when treated as an absolute division. Researchers never approach data without prior knowledge, and theoretically informed studies may still identify unexpected meanings.

The challenge is determining how existing concepts should influence the analysis without either imposing premature interpretations or ignoring relevant theoretical knowledge.

Inductive, Deductive, and Combined Coding Can All Be Appropriate

Understanding Inductive, Deductive, and Combined Qualitative Coding

What Is Qualitative Coding?

Qualitative coding involves assigning labels to segments of data to identify features relevant to the research question and analytic purpose. These labels may describe experiences, actions, meanings, processes, or concepts.

For example, consider a participant who says:

"I started using AI to prepare lesson materials, but I always check the output because sometimes it gives references that do not exist."

A researcher might initially code this statement as AI-assisted lesson preparation, verification of AI outputs, or concerns about fabricated references.

These codes highlight different aspects of the same account. Their relevance depends on what the study seeks to understand.

Coding is not identical to analysis. Depending on the methodology, codes may serve as tools for organizing material, comparing cases, examining meanings, developing categories, or constructing interpretive themes.

Some qualitative approaches do not use conventional segment-by-segment coding as their central analytic procedure. Therefore, decisions about coding should follow the broader selection of an appropriate qualitative analysis approach.

What Is Inductive Qualitative Coding?

Inductive coding develops analytic labels primarily through engagement with the collected material rather than applying a comprehensive set of predetermined categories.

The researcher examines transcripts, field notes, or documents and identifies features that appear relevant to the research question. Codes are then developed, reconsidered, combined, separated, or refined as analysis progresses.

For example, a researcher investigating teachers' experiences with generative AI might initially encounter repeated accounts of uncertainty about whether AI-generated teaching materials are reliable.

Rather than beginning with a predetermined category called technological trust, the researcher may develop codes such as checking factual accuracy, verifying references, comparing AI outputs with textbooks, and withholding unverified material from students.

Further analysis might examine relationships among these practices and the meanings teachers attach to them.

Importantly, inductive coding does not mean that codes arise independently of the researcher. What researchers notice is influenced by their questions, disciplinary knowledge, assumptions, and interpretive positioning.

Induction is therefore better understood as an analytic orientation that gives substantial priority to engagement with the data, not as a guarantee of interpretation free from theoretical influence.

What Is Deductive Qualitative Coding?

Deductive coding begins with concepts, categories, or theoretical constructs identified before or during analysis from an existing framework, prior research, or clearly specified analytic questions.

These concepts provide an initial structure for examining the data.

For example, a researcher studying technology adoption may use constructs from an established technology acceptance framework to develop preliminary categories concerning perceived usefulness and perceived ease of use.

Interview material is then examined for evidence relevant to these categories.

Deductive coding can be particularly useful when the research aims to examine how an existing framework applies in a particular setting, compare evidence across cases using common conceptual categories, or investigate predefined implementation dimensions.

However, using a theoretical framework does not mean that every participant statement must fit one of its categories. A rigid framework may overlook experiences that challenge or extend existing explanations.

Researchers should distinguish between using theory to guide attention and forcing data to confirm theoretical expectations.

What Is Combined Inductive and Deductive Coding?

Combined coding, sometimes described as hybrid coding, incorporates both theoretically informed categories and codes developed through engagement with the data.

The researcher may begin with a preliminary coding framework while allowing additional codes to capture relevant material that does not fit the existing structure.

For instance, a study of university teachers' AI adoption may initially use established constructs relating to perceived usefulness and ease of use. During analysis, participants may repeatedly discuss concerns about academic integrity, professional identity, or institutional accountability that are insufficiently represented in the initial framework.

The researcher can develop additional codes for these issues, examine their relationship to existing categories, and consider whether the original framework requires refinement.

This is not necessarily an equal mixture of induction and deduction. A study may be predominantly deductive with limited inductive expansion, or predominantly inductive while using theoretical concepts as sensitizing resources.

The balance should be determined by the analytic purpose rather than an expectation that combining approaches is automatically superior.

How Do the Three Coding Strategies Compare?

These are general tendencies. The actual procedures must be consistent with the specific analytic methodology.

Does Inductive Coding Mean You Must Avoid Reading the Literature?

No. Avoiding relevant literature is not a universal requirement of inductive analysis.

Researchers inevitably bring prior knowledge to their work. Reading methodological and substantive literature can help establish the research problem, identify important debates, and justify the study's contribution.

Some methodological traditions encourage particular ways of engaging with prior literature to avoid prematurely imposing existing explanations. Others explicitly use theoretical knowledge as part of the interpretive process.

The more defensible objective is to recognize and critically examine the influence of prior knowledge rather than claiming that the researcher has none.

Reflexivity is particularly important because coding decisions reflect how researchers understand the material and the questions they ask of it.

Does Deductive Coding Require a Fixed Codebook?

Not necessarily, although a structured codebook is often useful when predefined categories are central to the analysis.

A deductive coding framework may specify code names, definitions, inclusion criteria, exclusion criteria, and illustrative examples. These details can support consistent application, particularly in structured content analysis or team-based framework analysis.

However, the extent to which a codebook should be fixed depends on the methodology. Some approaches allow refinement of categories when evidence reveals ambiguities or missing concepts.

The question of whether to develop a qualitative codebook before data collection should therefore be answered in relation to the intended analytic approach, not through a universal requirement.

Can Inductive Coding Use a Codebook?

Yes. An inductively developed coding framework may be documented in a codebook as codes and their definitions become clearer.

For example, researchers conducting conventional qualitative content analysis may begin by examining the data and subsequently organize emerging codes into a structured framework.

The presence of a codebook does not, by itself, make the analysis deductive. What matters is how the codes were developed and how they function within the analytic process.

Conversely, not every approach requires codebook standardization. Reflexive thematic analysis, as articulated by Braun and Clarke, treats coding as a situated and interpretive practice rather than a procedure whose quality depends on independent coders applying identical definitions.

How Does Coding Strategy Relate to Thematic Analysis?

Thematic analysis is not a single uniform procedure. Different traditions adopt different assumptions about coding, theme development, researcher subjectivity, and analytic quality.

Reflexive thematic analysis may be more inductive or more deductive, depending on the research question and theoretical orientation. In either case, themes are actively developed through interpretation rather than simply discovered as pre-existing categories.

Codebook-oriented thematic approaches may use more structured procedures and predefined categories. Coding reliability approaches may emphasize consistency across coders, although these assumptions should not be transferred uncritically to reflexive thematic analysis.

Consequently, saying that a study used thematic analysis with inductive coding is not a complete methodological description. Researchers should identify the specific approach and explain how coding contributed to theme development.

How Does Coding Strategy Relate to Qualitative Content Analysis?

Qualitative content analysis provides a useful example of how different coding orientations can operate within a related methodological family.

Hsieh and Shannon distinguish conventional, directed, and summative approaches to qualitative content analysis.

Conventional content analysis commonly develops coding categories from the data. Directed content analysis uses existing theory or research to guide initial codes. Summative content analysis begins with identifying and examining particular words or content before interpreting their contextual meanings.

These approaches are not interchangeable, and summative analysis should not be reduced to simply counting words.

The distinction illustrates why the purpose of coding matters more than the presence or absence of predefined labels.

What About Abductive Qualitative Analysis?

Abductive analysis involves developing or reconsidering explanations through engagement between empirical observations and theoretical ideas, often prompted by surprising or puzzling evidence.

Rather than treating analysis as movement exclusively from data toward concepts or from theory toward data, abductive reasoning may involve repeatedly examining observations in relation to possible explanations.

For example, a researcher may expect teachers who perceive AI as useful to adopt it readily. Interviews might instead reveal that some highly positive teachers avoid using AI because they perceive conflicts with professional responsibilities.

This unexpected pattern may prompt reconsideration of the initial explanation and engagement with alternative theoretical concepts.

Abduction should not be treated as merely another name for combining inductive and deductive codes. It concerns a logic of explanatory reasoning that may involve theoretical reinterpretation, rather than only the technical source of coding categories.

Should You Decide the Coding Strategy Before Data Collection?

Usually, you should establish the intended coding orientation before collecting data, particularly when the study is designed to examine a specific framework or when the coding strategy influences interview questions and sampling.

However, you do not necessarily need to determine every code in advance.

A predominantly deductive study may require carefully developed prompts addressing predefined concepts. A predominantly inductive study may prioritize open accounts that allow participants to introduce relevant issues not anticipated by the researcher.

Combined approaches may need both forms of evidence.

The important relationship is between the coding strategy and the design of interview questions. Questions should provide opportunities to examine the intended concepts without making participants' responses artificially conform to them.

Do Multiple Coders Make One Strategy More Rigorous?

Not inherently. The value and purpose of multiple coders depend on the methodology.

In some structured coding approaches, multiple coders may independently apply categories, discuss ambiguities, and assess coding consistency. These procedures may be appropriate when the analysis relies on stable code definitions.

In reflexive thematic analysis, coding differences may instead reflect legitimate interpretive perspectives. Requiring coding agreement as a universal indicator of quality would be inconsistent with the assumptions of that approach.

Researchers should therefore decide whether multiple coders are needed based on the analytic purpose rather than treating inter-coder agreement as a universal requirement.

What Makes a Coding Strategy Methodologically Rigorous?

Rigor is not established simply by describing codes as inductive, deductive, or hybrid.

Researchers should explain how codes were developed, what role existing theory played, how coding decisions were examined, and how the analysis moved from coded material toward its findings.

In a structured coding approach, clear definitions, appropriate documentation, and systematic application may be especially important. In a reflexive interpretive approach, the coherence and depth of interpretation, researcher reflexivity, and engagement with the dataset may be more relevant.

Regardless of orientation, findings should be grounded in a careful examination of the material. Researchers should consider evidence that complicates their initial interpretations rather than selecting only extracts that support preferred conclusions.

Three Ways to Code Interviews About Teachers' Use of Generative AI

Inductive Coding: Begin With the Participant's Account

The researcher reads the statement closely and develops provisional codes reflecting features of the account.

• Saving preparation time
• Hesitation about AI-assisted assessment
• Uncertainty about accountability
• Concern about inaccurate generated content

These codes may be revised as additional interviews are examined. The researcher does not assume in advance that every account must fit a particular technology adoption framework.

Deductive Coding: Begin With an Existing Framework

Suppose the researcher is examining technology adoption through predefined constructs, including perceived usefulness and perceived risk.

The statement may be coded under perceived usefulness because the participant identifies time savings. Concerns about assessment and accountability may be examined under an appropriately defined risk-related category.

However, the researcher should not automatically force accountability concerns into an unrelated construct simply because the initial framework lacks a suitable category.

Combined Coding: Use Existing Concepts While Developing Additional Codes

The researcher begins with theoretically informed categories concerning usefulness and perceived risk but also develops a new code for uncertainty about professional accountability.

Across interviews, this code may become analytically important if participants describe responsibility as a distinct consideration that is not adequately explained by the initial framework.

None of these coding decisions is automatically correct simply because it follows a particular orientation. The researcher must justify the codes in relation to the study's purpose and the meaning of the material.

Furthermore, these are illustrative initial coding decisions, not completed qualitative findings. A defensible analysis would require sustained engagement with the broader dataset, attention to variation and context, and further interpretive work.

Common Misconceptions About Inductive and Deductive Coding

How to Choose the Coding Strategy That Fits Your Study

Begin by identifying the relationship your study intends to establish between existing knowledge and the empirical material.

If the primary objective is to understand participants' experiences without organizing the analysis around a predetermined conceptual framework, an inductive orientation may be suitable.

If the study explicitly examines established constructs or applies a predefined framework, a deductive orientation may provide a clearer analytic structure.

If the purpose involves examining existing concepts while remaining open to additional categories, a combined strategy may be defensible.

Once you have selected an orientation, explain it explicitly in the methods section. A useful description identifies the source of initial codes, how additional codes were developed, the role of existing theory, and how coded material contributed to the final interpretation.

For a combined approach, you might explain that the initial coding framework was informed by selected theoretical constructs, while additional codes were developed to capture relevant meanings not adequately represented by those constructs.

Avoid claiming that codes emerged naturally without researcher involvement. Equally, avoid suggesting that deductive coding mechanically verifies theory.

The most defensible coding strategy is the one that supports the analytic purpose while making its assumptions and procedures open to scholarly examination.

Before Choosing Inductive, Deductive, or Combined Coding, Check:

Frequently Asked Questions About Qualitative Coding Strategies

Choose the Coding Logic That Serves Your Research Question

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.