How Should Researchers Think About Race and Ethnicity as Research Variables?

What Does a Race or Ethnicity Variable Actually Tell You?

Race and ethnicity appear routinely in questionnaires, demographic tables, administrative datasets, statistical models, and subgroup analyses. Their familiarity can make them seem like straightforward participant characteristics: collect a category, enter it into the dataset, compare the groups, and interpret the difference.

The methodological reality is considerably more complicated.

Race and ethnicity are socially and historically situated classifications whose definitions, categories, and meanings vary across societies and over time. They may nevertheless be highly relevant to research because racialization, ethnicity, culture, migration, discrimination, socioeconomic conditions, environmental exposures, access to institutions, and other social processes can shape people's experiences and outcomes.

The question is therefore not simply whether researchers are allowed to collect race or ethnicity. It is what these variables are intended to represent in a particular study and whether the measurement and interpretation actually support that purpose.

Use Race and Ethnicity When They Serve a Clear Research Purpose

Race and Ethnicity Need a Conceptual Rationale, Not Just a Checkbox

A demographic variable becomes scientifically useful when researchers know why they are collecting it and what role it plays in the study. Race and ethnicity should not be exceptions to that principle.

Before choosing categories, ask what the variable is intended to accomplish. Is it needed to describe the sample? Evaluate inclusion? Examine disparities? Investigate experiences of discrimination or racialization? Meet a funder's reporting requirement? Study cultural or ethnic identity? Assess whether an intervention reaches different populations? These purposes are not interchangeable.

Race Should Not Be Treated as a Simple Biological Classification

Race has often been treated historically as though racial categories divide humans into biologically discrete groups. Contemporary scientific guidance cautions strongly against that interpretation.

The National Academies' framework for genetics and genomics research describes race as a sociopolitically constructed system of classification and specifically recommends that researchers not use race as a proxy for human genetic variation. NIH likewise states in its current inclusion policy that the federal racial and ethnic categories it uses are social-political constructs and should not be interpreted as anthropological classifications.

This does not make race irrelevant to research. Quite the opposite: social classifications can have material consequences. Racism, discrimination, residential segregation, unequal access to resources, differences in environmental exposure, institutional practices, and other processes associated with racialization can influence health, education, employment, and many other outcomes.

Ethnicity Is Also More Complex Than a Fixed Demographic Label

Ethnicity commonly refers to forms of shared identity or affiliation associated with culture, heritage, language, nationality, geographic origin, traditions, or other dimensions. Exactly what ethnicity means varies across contexts.

Two people assigned to the same broad ethnic category may differ considerably in language, migration history, cultural practices, socioeconomic circumstances, national origin, and self-identification. Conversely, people may identify with multiple ethnicities.

Researchers should therefore avoid assuming that an ethnic label measures culture itself. If the hypothesis concerns language, migration experience, dietary practices, cultural beliefs, or another specific mechanism, measuring that characteristic directly may be more informative than treating ethnicity as its substitute.

Ask What Race or Ethnicity Is Doing in Your Analysis

A variable can play several roles in research. Race or ethnicity might be used descriptively, as a sampling or stratification variable, as an exposure or predictor in a model, as the basis for examining heterogeneity, or as part of an analysis of disparities.

Each role requires different reasoning.

If the variable is descriptive, researchers should explain how participants were categorized. If it is used analytically, the rationale should be stronger: what theoretical or empirical relationship is being investigated, and what does the variable represent in that analysis?

Automatically adding race to a regression model as a generic “control variable” can be particularly difficult to interpret. Statistical adjustment changes the estimand and can obscure rather than clarify relationships when the causal role of the variable and related social processes have not been considered.

A Group Difference Does Not Explain Itself

Suppose researchers observe that an outcome differs between racial groups. The statistical comparison establishes a difference under the conditions of the study. It does not establish that race itself caused the difference.

Potential explanations might involve differences in exposure to discrimination, neighborhood conditions, wealth, educational opportunities, health care access, occupational conditions, environmental hazards, institutional treatment, migration histories, or numerous other factors. Which mechanisms are plausible depends on the research question.

Categories Depend on Place, Time, and Purpose

There is no single worldwide classification of race and ethnicity suitable for every research project. Categories used by a U.S. federal agency may be poorly suited to another country, and categories meaningful in one historical period may change over time.

Even within a country, researchers may need different levels of specificity for different purposes. A broad category such as “Asian,” for example, can combine populations with substantially different national origins, migration histories, languages, socioeconomic circumstances, and experiences.

This is one reason researchers should consider when demographic categories oversimplify the people being studied.

Self-Identification Is Often Preferable When Identity Is What Matters

When race or ethnicity is collected as an aspect of participant identity, self-identification generally avoids having researchers assign participants to categories based on appearance, name, or assumption. Researchers should also consider whether participants may select multiple categories and whether more detailed self-description is appropriate.

The appropriate method still depends on purpose. Historical records, administrative datasets, clinical systems, and secondary datasets may use classifications that researchers cannot redesign. In those cases, investigators should report how the original data were collected rather than implying that inherited categories were chosen specifically for the current study.

Official Categories May Be Required for Particular Research

Researchers sometimes have limited discretion because a funder, government agency, registry, or other authority requires specific categories.

For example, the current NIH inclusion policy requires applicable NIH-funded clinical research to address inclusion by race and ethnicity. NIH's 2025 revision references the U.S. Office of Management and Budget's revised Statistical Policy Directive No. 15 categories and explicitly characterizes the classifications as social-political constructs. NIH also encourages collection of greater detail that can be aggregated into required categories.

Implementation can lag behind policy revision. NIH stated in August 2025 that award recipients should continue reporting inclusion data using the older 1997 categories until further notice, despite the revised federal standards. This illustrates why researchers should verify the current operational instructions rather than assuming that a newly announced classification has already replaced every existing reporting system.

Such requirements are policy-specific. They should not be generalized into universal categories for all research or all countries.

Broad Categories Can Conceal Important Heterogeneity

Aggregating participants may provide enough observations for analysis and make results easier to report. But aggregation can also conceal meaningful differences among populations placed under the same umbrella label.

A broad racial or ethnic category may contain people with different languages, geographic origins, immigration histories, socioeconomic circumstances, cultural affiliations, and experiences of discrimination. A group-level average can therefore obscure substantial within-group variation.

Researchers should decide the appropriate level of aggregation based on the research question, available data, statistical precision, participant privacy, and interpretability. Disaggregation is not automatically better if it produces unstable estimates or risks identifying participants.

Race and Ethnicity May Intersect With Other Characteristics

People do not experience race or ethnicity independently of age, sex, gender, disability, socioeconomic circumstances, geography, language, migration status, and other dimensions of life.

This matters because a broad group comparison can obscure considerable variation within each category. It also means researchers should be cautious about treating race or ethnicity as a sufficient explanation for a complex social pattern.

Thoughtful analysis may sometimes require measuring the more specific mechanisms implicated by the research question rather than expecting one demographic variable to carry the explanatory burden.

How the Same Racial Difference Can Support Very Different Interpretations

The racial variable can be informative without being treated as the causal mechanism. In fact, recognizing that distinction often generates a more useful research question.

Common Mistakes When Using Race and Ethnicity in Research

Know Why You Need the Variable Before Choosing the Categories

A useful starting point is to write one sentence explaining why race or ethnicity matters to your study. If that sentence is difficult to write, the variable may not yet have a clear conceptual role.

Finally, consider why diversity in the sample matters for the particular research question. The purpose is not to populate a demographic table with as many categories as possible. It is to ensure that participant selection, measurement, analysis, and interpretation are aligned with the knowledge the study is supposed to produce.

Before Using Race or Ethnicity in Your Study, Check the Rationale

Questions About Race and Ethnicity as Research Variables

Use the Categories to Study a Question, Not as an Explanation by Themselves

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.