03 · What You Need to Know
Underrepresentation Has to Be Judged Against a Research Purpose
The word “underrepresented” is sometimes used as though it describes a permanent property of a demographic group. Methodologically, that is too imprecise. A population is underrepresented relative to something: a target population, the people affected by a condition, the population expected to use an intervention, an analytical objective, or a body of research.
Without specifying that benchmark, the label tells readers very little.
Underrepresentation Is Not the Same as Being a Numerical Minority
Suppose a population makes up 10% of the target population and approximately 10% of a probability sample. It remains numerically smaller than the other 90%, but that alone does not make it underrepresented relative to population composition.
Now suppose the researchers need to estimate an outcome separately for that 10% and the study includes only a handful of participants from the group. Proportional representation may still provide insufficient information for the planned subgroup analysis.
These scenarios show why representation and representativeness answer different questions. Numerical presence, proportional similarity to a population, and analytical adequacy should not be collapsed into a single concept.
A Group Can Be Present and Still Be Underrepresented
Researchers sometimes treat inclusion as a yes-or-no question. If at least some participants from a population appear in the dataset, the group is described as represented.
Presence can certainly matter, but it may not be enough.
A subgroup with very few observations may produce estimates with substantial uncertainty. Qualitative research may include one or two participants from a population without capturing important variation in experiences. A clinical study may enroll members of a group but still provide too little information to assess outcomes relevant to that population.
The relevant question is therefore not only “Are they there?” but also “Does their participation provide the evidence this study needs?”
Underrepresentation Can Begin Before Recruitment
Low enrollment is often attributed to participant willingness. That explanation can overlook the many filters that operate before a person ever receives an invitation.
Eligibility criteria can exclude populations systematically. Recruitment may occur through institutions or communication channels that do not reach them. Research sites may be geographically inaccessible. Participation may require transportation, time away from work, childcare, particular technology, language proficiency, or repeated visits.
The National Academies' review of barriers to clinical research representation emphasizes that barriers operate at multiple levels, including individual studies, research institutions, communities, and the wider research system. It also cautions against assuming that low participation among racial and ethnic populations simply reflects unwillingness to participate.
This is why researchers should first examine who actually gets an opportunity to enter the research before interpreting enrollment patterns as participant preference.
Eligibility Criteria Can Systematically Remove Relevant Populations
Some underrepresentation originates in the protocol itself. Upper age limits, exclusion of people with disabilities or common comorbidities, language requirements, technology requirements, or other restrictions may remove populations that are part of the eventual real-world population.
Many exclusions are scientifically or ethically justified. The problem arises when restrictions are weakly connected to the research question or when their cumulative effect creates a population substantially narrower than the one researchers later discuss.
Reviewing whether inclusion and exclusion criteria systematically narrow the sample is therefore one part of addressing underrepresentation.
Study Procedures Can Produce Underrepresentation Without Explicitly Excluding Anyone
Imagine that everyone in the target population is formally eligible, but participation requires four weekday visits to a distant research site. People with inflexible employment, caregiving responsibilities, transportation difficulties, mobility limitations, or limited financial resources may be less able to participate.
No criterion excludes them. The design still creates unequal access.
Similar barriers can arise through inaccessible research materials, complicated consent procedures, digital-only participation, lack of accommodations, lengthy visits, or inflexible scheduling. Research on accessible study design has shown how people with disabilities can be implicitly excluded when recruitment, consent, intervention, or measurement procedures are inaccessible.
Underrepresentation can therefore emerge from the interaction between people's circumstances and what the study requires of them.
Persistent Underrepresentation Can Become an Evidence Gap
A single narrowly defined study does not need to answer every question for every population. The larger concern arises when the same populations remain poorly studied across many investigations.
CIOMS notes that historical exclusion of populations considered vulnerable, including children, women of reproductive age, and pregnant women, contributed to limited information about the diagnosis, prevention, and treatment of conditions affecting those populations. The ethical concern is not merely unequal enrollment. Repeated exclusion can mean unequal access to the knowledge produced by research.
This distinction is important. Underrepresentation matters most when it changes what the evidence base can reliably tell us.
Underrepresentation Can Limit Subgroup-Specific Conclusions
If an important population contributes very little data, researchers may be unable to determine whether an intervention, association, or outcome differs for that group.
That does not mean researchers should automatically run subgroup analyses for every demographic characteristic. Such analyses require a substantive rationale and adequate information. Rather, when subgroup-specific inference is genuinely important, recruitment should be designed with that objective in mind.
For example, a study intending to analyze outcomes meaningfully by sex or gender needs sufficient relevant participation for the intended analysis. Merely recording the variable after recruitment does not create the necessary information.
Underrepresentation Can Affect Applicability Without Making the Study Invalid
It is important not to overcorrect. A study does not become invalid merely because every population is not proportionally represented.
Some research deliberately studies narrowly defined groups. Qualitative studies may use purposive sampling rather than population-proportionate recruitment. Early mechanistic studies may require tightly controlled populations. Studies designed for subgroup comparisons may intentionally oversample smaller populations.
The methodological concern is a mismatch between the evidence and the claim. If a study contains little information about a population, conclusions about that population should remain appropriately cautious.
Underrepresentation Is Also an Ethical Question
The Belmont Report connects justice with fair participant selection and the distribution of research burdens and benefits. CIOMS similarly argues that equitable research requires attention to both who bears the burdens of participation and who benefits from the knowledge produced.
This creates two possible injustices. Some disadvantaged populations may be recruited disproportionately because they are convenient or have limited alternatives. Other populations may be excluded so consistently that evidence relevant to their needs remains inadequate.
Equitable participation therefore does not mean simply recruiting more people from every group. It requires examining whether the distribution of participation makes sense scientifically and ethically.
There Is No Universal Percentage for Adequate Representation
Researchers sometimes want a threshold: must every group constitute at least 10% of the sample? Should sample proportions match census percentages? Is equal representation preferable?
There is no universal answer.
For population estimation, proportionality and the sampling design may be important. For subgroup comparisons, researchers may need substantially more participants from a smaller population than proportional sampling would provide. For qualitative research, adequacy depends on the methodology, phenomenon, sampling logic, and informational needs rather than population percentages.
Population underrepresentation
A group participates less than would be expected relative to a relevant target or affected population.
Analytical underrepresentation
A group contributes too little information to support an important analysis or inference, even if its sample proportion is not unusually low.
Sometimes the Appropriate Response Is Deliberate Oversampling
If a smaller population is central to the research question, proportional recruitment may yield too few participants for meaningful analysis. Researchers may then intentionally recruit that population at a higher rate.
This does not necessarily make the design less rigorous. Deliberate oversampling of an underrepresented group can be methodologically appropriate when planned around a clear analytical objective and accounted for correctly in analyses that estimate population quantities.
Watch Out
Do not assume that low enrollment proves that a population is unwilling to participate. Before making that claim, examine whether people were eligible, reached by recruitment, invited, able to access the study, and given participation procedures that were realistically feasible.
