What If the Population You Want to Study Is Extremely Rare or Hard to Reach?

What If the People You Need Are Exceptionally Difficult to Find?

Some research questions concern populations that are easy to define but remarkably difficult to study. Perhaps only a small number of eligible people exist. They may be geographically dispersed, absent from conventional records, protected by institutional gatekeepers, reluctant to identify themselves, or connected through social networks that outsiders cannot readily enter.

This can happen when studying uncommon occupations, rare clinical conditions, highly specialized professionals, geographically isolated communities, people with unusual experiences, or populations whose identities or activities make participation sensitive.

The immediate temptation may be either to abandon the question or to recruit whoever you can find and hope the sample is sufficient. Neither response is automatically appropriate.

Rare and hard-to-reach populations require a more deliberate question: What exactly makes this population difficult to study, and what sampling and recruitment strategy is defensible given that constraint?

A Difficult Population Requires a Different Recruitment Logic

Why Some Research Populations Are So Difficult to Study

Rare and hard to reach do not mean the same thing

A population may be difficult to study for several quite different reasons. Treating all of these situations as the same recruitment problem can lead to poor methodological decisions.

A rare population has relatively few members who satisfy the characteristics relevant to the research question. A hard-to-reach population may be larger but difficult for researchers to identify, contact, engage, or recruit. A hidden population is often one for which no adequate sampling frame exists and whose members may be reluctant to identify themselves because of privacy, stigma, marginalization, legal concerns, or other sensitivities.

The terminology is not perfectly standardized across disciplines. Reviews of hard-to-reach population research caution against treating diverse populations as though they share a single set of characteristics merely because researchers find them difficult to reach.

These conditions can overlap. A population can be both rare and hidden, for example, which makes recruitment particularly demanding.

Diagnose the access problem before choosing a sampling method

Do not begin by deciding, "I will use snowball sampling because my participants are hard to find." Start by identifying what makes them hard to find.

Ask whether the main constraint is numerical rarity, geographical dispersion, absence of a sampling frame, institutional gatekeeping, unwillingness to self-identify, distrust of researchers, communication barriers, participant burden, or some combination of these.

The answer changes what a sensible strategy looks like. If eligible specialists are rare but publicly identifiable through professional directories, expanding recruitment across institutions or regions may be enough. If participants gather at identifiable venues, venue-based or time-location approaches may be possible. If membership is hidden but people are connected through social networks, network-based recruitment may be more useful.

In other words, "hard to reach" describes a research problem. It does not prescribe one universal solution.

Check whether the population is genuinely rare

Researchers sometimes infer rarity from difficulty finding participants. Those are not equivalent.

Suppose you are studying university instructors who have formally incorporated generative AI into assessment policies. You contact two departments and find only six eligible instructors. That does not establish that the population is rare. Your recruitment frame may simply be narrow.

Before concluding that few eligible people exist, investigate what is known about the population. Depending on the topic, useful information may come from administrative records, professional organizations, registries, published studies, government statistics, community organizations, research sites, or preliminary enquiries.

This extends the earlier question of whether your target population is actually accessible. A population may appear rare because only a small fraction is visible through your present access route.

Expand the recruitment frame when appropriate

If the population is genuinely uncommon, recruiting from one convenient site may be unrealistic. Expanding the frame can sometimes make the project feasible without changing the research question substantially.

Depending on the study, that could mean recruiting across multiple schools, hospitals, universities, clinics, professional organizations, geographical areas, online communities, or other appropriate settings. Multisite recruitment increases the pool of potentially eligible participants and may also reduce dependence on a single gatekeeper.

Expansion has costs. Additional sites may require separate permissions, coordination, travel, staff, data-management procedures, or ethics-related amendments and approvals depending on the jurisdiction and study. A wider recruitment frame therefore improves feasibility only if the additional complexity remains manageable.

Community and organizational partnerships may be essential

For some populations, access depends less on finding a larger list and more on establishing a credible relationship with people who already have access and trust.

Community organizations, professional associations, service providers, advocacy groups, clinics, local leaders, and other intermediaries may help researchers understand how a population is organized, where potential participants can appropriately be reached, what language is acceptable, and what barriers may discourage participation.

Research involving marginalized or hidden populations may require formative work before recruitment begins. Literature on respondent-driven and other hard-to-reach sampling approaches emphasizes understanding the target community and, in some applications, identifying trusted or well-connected initial participants.

Partnerships should not be treated merely as mechanisms for obtaining bodies for a sample. They can affect the ethical acceptability, cultural appropriateness, trustworthiness, and practical success of the research relationship.

Snowball sampling can open networks, but it has limitations

Snowball sampling is commonly used when researchers cannot easily identify all members of a population. Researchers begin with a small number of eligible participants or contacts, who then help identify or refer other potential participants. Recruitment proceeds through social connections.

This can be useful when membership is difficult to identify externally but people within the population know one another.

The apparent simplicity of snowball sampling can be misleading, however. Participants with larger social networks may become more visible. Recruitment may remain concentrated within socially similar clusters. Isolated members may never be reached. People may also avoid referring others when the topic is sensitive or when doing so could compromise privacy.

These limitations matter particularly when researchers want population-level estimates. A network-generated sample should not automatically be described as representative simply because recruitment continued until the desired number was obtained.

Respondent-driven sampling is more structured than ordinary snowball sampling

Respondent-driven sampling, commonly abbreviated as RDS, was developed as a structured form of chain-referral sampling for populations that are difficult to enumerate and recruit through conventional frames.

RDS typically begins with selected initial participants, often called seeds. Participants receive a limited number of recruitment opportunities and recruit peers from the target population through their networks. Recruitment then proceeds in waves. Information about participants' network connections or network size is used in the analytical framework.

RDS has been used extensively in research involving hidden and marginalized populations, particularly in public health. It should not, however, be treated as a magic conversion of a convenience sample into a representative probability sample. Its inferential properties depend on assumptions about social networks and recruitment processes, and methodological reviews continue to identify limitations. Recent reviews note that evidence is insufficient to regard RDS as the universally best method for hard-to-reach populations.

Researchers considering RDS should therefore understand its assumptions, implementation requirements, recruitment diagnostics, and analysis rather than using the label for ordinary participant referrals.

Time-location sampling may work when the population gathers somewhere observable

Some populations lack a complete list but regularly gather at identifiable locations. Depending on the research context, these might include service sites, community locations, events, establishments, or other venues.

Time-location sampling, sometimes called time-space or venue-based sampling, uses combinations of locations and times as part of the sampling process. Instead of requiring a list of every individual in the target population, researchers identify where and when eligible people are likely to be present and sample through those venue-time units.

The approach can be useful for populations that are difficult to enumerate but visible in recurring places. It is less suitable when important members never attend the identified venues. As with other strategies, the recruitment mechanism shapes who has an opportunity to enter the study.

Purposive or targeted recruitment may fit some research questions better

Not every study of a rare population seeks a population prevalence estimate or statistically representative sample.

Qualitative research may deliberately recruit participants with particular experiences, roles, perspectives, or characteristics because those cases can provide information relevant to the research question. A study of a rare professional role, for example, may use purposive sampling to identify information-rich participants across selected settings.

Targeted approaches can also seek particular subgroups within a heterogeneous hard-to-reach population. Reviews of hard-to-reach sampling methods describe targeted sampling as an approach that first examines the population and then deliberately recruits across identified subgroups.

The methodological justification should follow the research purpose. A purposive sample may be entirely appropriate for understanding experiences or processes while being inappropriate for estimating population prevalence. The mistake is not using a nonprobability method. The mistake is making claims the sampling method cannot support.

Do not confuse snowball sampling with respondent-driven sampling

Because both use social referrals, the terms are sometimes used interchangeably. They should not be.

A smaller achievable sample may require a different research question

Suppose the analysis originally planned for your quantitative study requires several hundred participants, but careful investigation shows that only 80 eligible people are likely to be reachable during the project. The methodological response should not be to collect 40 and proceed as though nothing changed.

You may need to expand recruitment, change the design, reconsider the planned analysis, narrow the population, modify the research question, or conclude that the intended study cannot presently be conducted as designed.

For qualitative research, a small population does not automatically create the same problem because adequacy is evaluated according to the methodological approach and informational needs rather than a universal numerical threshold. Still, the population must contain enough appropriate cases to address the question credibly.

The central principle is alignment: the sample you can obtain, the method you use, and the claim you intend to make must fit one another.

Rarity can make confidentiality more difficult

Rare populations create an ethical problem that can be easy to overlook: participants may be identifiable even after obvious identifiers such as names are removed.

Imagine interviewing the only three female chief executives in a narrowly defined local industry, the only specialist physician of a particular type at a hospital, or the sole employee occupying an unusual role in an organization. Reporting combinations of job title, institution, age, location, or experience may allow knowledgeable readers to infer who participated.

Researchers should therefore think beyond removing names. Recruitment procedures, quotations, demographic tables, case descriptions, small cell counts, and combinations of contextual details can all create disclosure risks. The appropriate protections depend on the population, study design, data, applicable ethics requirements, and promises made to participants.

Difficulty reaching a population can itself introduce bias

Hard-to-reach populations are rarely equally hard to reach in every segment.

The most socially connected members may be easiest to recruit through referrals. People attending services may be easier to reach than those disconnected from services. Those comfortable disclosing a sensitive identity may be more visible than those who conceal it. Online recruitment may favor people active on particular platforms.

Consequently, successfully obtaining the desired sample size does not prove that the sampling problem has been solved. You should ask who your strategy is likely to reach easily, who remains difficult to reach, and whether those differences matter for the research question.

More recruitment effort does not automatically solve a coverage problem

If a sampling strategy systematically misses part of the population, simply extending recruitment may produce more participants from the same reachable segment.

This is particularly important when a researcher says, "I will keep collecting until I have enough." Enough for what? A larger sample may increase the amount of information or improve statistical precision under appropriate conditions, but it does not automatically remove selection processes created by the recruitment method.

Sometimes the appropriate solution is not more recruitment but a different recruitment channel, an additional site, another sampling method, or a narrower interpretation of the population represented by the evidence.

When the Eligible Population Almost Disappears After Screening

The revised project remains meaningful. What changed was not the value of the phenomenon but the alignment among the research question, available population, sampling strategy, and intended conclusions.

Common Mistakes When the Population Is Difficult to Reach

Adapt the Study to the Reason the Population Is Difficult to Reach

If preliminary recruitment planning suggests that your population is rare or hard to reach, do not immediately abandon the research question. First diagnose the constraint precisely. The solution to geographical dispersion may be multisite recruitment. The solution to a missing sampling frame may involve network-based or venue-based methods. Institutional access may require partnerships. A genuinely tiny population may require a different question or design.

Then ask whether the resulting strategy still produces evidence capable of answering the question you intend to ask.

This final possibility matters. Some worthwhile questions simply require networks, sites, funding, time, or methodological infrastructure that you do not presently have. Recognizing that situation is part of determining whether a research question is genuinely feasible, not evidence that the question itself lacks value.

Before Studying a Rare or Hard-to-Reach Population

Frequently Asked Questions About Rare and Hard-to-Reach Populations

Hard to Reach Does Not Mean Impossible to Study

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.