03 · What You Need to Know
How to Respond When Research Access Does Not Cover the Intended Population
What Does It Mean for the Accessible Population to Be Narrower?
The target population is the collection of individuals or other entities about which the research intends to make claims. The accessible population consists of those who can realistically be identified or reached through the study's recruitment arrangements.
The accessible population is narrower when some eligible members of the target population cannot enter the recruitment process.
For example, a national study of teachers may have access only to teachers employed in selected school divisions. A clinical study may target patients across different care settings but recruit exclusively from specialist hospitals.
The difference concerns population coverage, not simply the number of people who eventually participate.
A researcher might contact every eligible individual in four participating institutions and still have a substantially narrower accessible population than originally intended.
The distinctions among target population, accessible population, and study sample should therefore be established before evaluating the problem.
Why Might the Accessible Population Be Narrower?
Limited accessibility can arise from institutional permission, geography, incomplete registers, financial constraints, recruitment methods, ethical restrictions, and the characteristics of the population itself.
Some restrictions are unavoidable within a particular project. Others may result from recruitment decisions that can be modified.
| Source of restricted access |
Example |
Possible response |
| Institutional access |
Only selected universities approve recruitment. |
Seek additional sites or qualify institutional coverage. |
| Geographic restrictions |
Recruitment occurs only in urban areas. |
Consider geographically distributed sites or alternative recruitment channels. |
| Incomplete sampling frame |
Eligible individuals are missing from a register. |
Update, supplement, or reconstruct the frame. |
| Digital access barriers |
Online-only recruitment excludes people with limited connectivity. |
Offer suitable offline or assisted recruitment options. |
| Language barriers |
Recruitment materials are available in only one language. |
Consider appropriately translated and validated materials. |
| Ethical or legal restrictions |
Some records or populations cannot be accessed under existing approvals. |
Seek appropriate authorization or narrow the feasible inference. |
| Resource constraints |
Recruitment cannot cover all intended regions. |
Prioritize coverage strategically and reassess achievable precision. |
Identifying the cause matters because an incomplete administrative list requires a different remedy from institutional nonparticipation or an ethically justified restriction.
Is a Narrower Accessible Population Automatically a Research Problem?
No. The consequences depend on the intended estimand and whether inaccessible population members differ in ways relevant to that estimand.
Suppose a study investigates a particular psychological mechanism that plausibly operates across several settings. Evidence from a restricted population may still be informative, although broader applicability requires justification.
By contrast, a study estimating the national prevalence of inadequate internet access would face a substantial problem if recruitment excluded institutions with limited connectivity.
The same accessible population can therefore be adequate for one research question and inadequate for another.
Researchers should assess whether differences between accessible and target populations affect the intended inference rather than assuming every mismatch is equally consequential.
What Should You Examine First?
Begin by identifying the population members or settings that cannot enter the study.
Then ask whether their absence is likely to affect the outcome, exposure, relationship, or intervention effect being investigated.
For example, a study of educational AI adoption that excludes institutions without established digital infrastructure may overrepresent favorable implementation conditions.
Relevant differences might include institutional resources, geographic location, socioeconomic characteristics, disability-related access needs, or prior experience.
Researchers should use available administrative statistics, census information, institutional records, or previous research to examine these differences.
Where direct comparison is impossible, the uncertainty should be acknowledged rather than replaced with an assumption of similarity.
Should You Try to Expand the Accessible Population?
Expanding coverage is often the most direct response when important target population groups are missing and additional recruitment is feasible.
For example, a university study might recruit additional institutions representing different geographic locations, institutional types, and levels of digital infrastructure.
Researchers might also introduce alternative recruitment channels for eligible individuals who cannot be reached through existing institutional systems.
However, adding sites should follow a defensible selection strategy.
Recruiting whichever institutions respond most quickly may increase the number of sites without adequately addressing the characteristics that originally limited coverage.
The objective should be to improve the relevance of the accessible population, not merely increase the number of participants.
Can You Improve Coverage Without Adding More Study Sites?
Sometimes. Important exclusions may occur within participating sites rather than between sites.
For example, an online survey may reach only students with active institutional accounts, even though other eligible students are enrolled at the same university.
Improving the sampling frame, providing accessible survey formats, or using additional approved contact methods may increase coverage.
Researchers should distinguish frame undercoverage from nonresponse.
Students missing from the sampling frame have no chance of selection through that frame. Students who receive an invitation but decline participation contribute to nonresponse.
The remedies differ, and addressing one does not automatically correct the other.
Should You Change the Target Population to Match the Accessible Population?
Sometimes, but the revision should follow a genuine change in the research objective.
Suppose the original question concerns all undergraduate students nationally, but recruitment is possible only at four universities.
Redefining the target population as students at those four institutions changes the intended scope of inference.
That may be defensible if the research is reframed as an institutional or multisite investigation.
However, it does not answer the original national prevalence question.
Researchers should document the revision and explain whether it affects the research question, sampling strategy, hypotheses, or estimand.
A population definition should not be altered after results are examined merely to make the observed sample appear representative.
Can You Retain the Original Target Population?
Yes. Researchers can retain a broader population of substantive interest while acknowledging that their study directly covers a narrower accessible population.
For example, a study may investigate an issue relevant to national higher education while reporting empirical estimates specifically for participating institutions.
The discussion may consider broader implications, but it should distinguish observed findings from extrapolations requiring additional assumptions.
Retaining the original target population can be useful when the broader research question remains important and the limitations are transparent.
However, researchers should not label estimates from a restricted sample as national estimates without a defensible inferential basis.
Can a Larger Sample From the Accessible Population Solve the Problem?
Not necessarily. Increasing sample size may improve precision for the population represented by the recruitment procedure.
It does not automatically eliminate systematic differences between accessible and inaccessible population members.
For example, surveying 5,000 students from well-resourced urban universities may provide highly precise estimates for those settings while offering limited evidence about students at under-resourced rural institutions.
The distinction between precision and bias is fundamental.
Researchers should not assume that a larger sample automatically eliminates sampling bias.
Can Statistical Weighting Help?
Weighting may help address certain differences between the observed and target populations when relevant population information is available.
For example, calibration weights might align the sample with known population distributions of institutional type, geographic location, or educational level.
However, weighting requires defensible assumptions about the relationship between measured characteristics, selection, and the study outcome.
If entire target population subgroups are absent from the observed data, their outcomes cannot generally be recovered without additional modeling assumptions or external evidence.
Weights may also become unstable when some population characteristics are poorly represented.
Researchers should examine weight distributions, overlap, and sensitivity to adjustment choices rather than assuming weighting automatically restores representativeness.
What Is the Role of Positivity and Overlap?
Population adjustment and transportability methods often require adequate overlap between the characteristics represented in the study and those present in the target population.
Suppose an AI tutoring intervention is evaluated only among students with reliable internet access.
If the target population includes students without reliable connectivity, the observed data may provide little direct evidence about intervention effects for that subgroup.
Statistical models may extrapolate beyond the observed support, but such estimates depend on assumptions that cannot be verified from the excluded group.
Researchers should therefore assess whether relevant combinations of characteristics in the target population are represented in the study.
When overlap is inadequate, additional recruitment may be preferable to relying on increasingly complex adjustments.
How Does the Problem Differ for Descriptive and Causal Research?
For descriptive research, the central concern may be whether the accessible population provides an appropriate basis for estimating a population mean, proportion, or distribution.
For causal research, the concern may be whether an intervention effect estimated in the study population can be transported to the target population.
For example, a randomized AI tutoring trial conducted in selected universities may credibly estimate an assignment effect within those institutions under appropriate assumptions.
However, the effect may differ in institutions with limited infrastructure if connectivity or implementation conditions modify the intervention's effect.
Randomization within the study does not automatically establish generalizability beyond it.
Researchers should distinguish internal causal identification from population transportability.
Can Sensitivity Analysis Help When Inaccessible Groups Cannot Be Studied?
Yes. Sensitivity analysis can examine how conclusions might change under plausible assumptions about the missing population.
For example, researchers could estimate how a population prevalence would change if inaccessible students had substantially lower or higher rates of AI use than accessible students.
Such calculations do not establish the true outcomes of inaccessible members.
They reveal how strongly the intended conclusion depends on assumptions about those members.
Where possible, plausible ranges should be informed by independent data or previous research.
Researchers should clearly distinguish hypothetical sensitivity scenarios from empirical estimates.
How Can You Quantify the Potential Coverage Problem?
For a simple descriptive mean, the target population can be divided into accessible and inaccessible components.
The difference between the accessible and target population means is:
This expression illustrates that the consequences of restricted coverage depend on the proportion excluded and how different the excluded population is in the outcome.
It assumes the accessible population mean is known or appropriately estimated. Sampling error and nonresponse may introduce additional uncertainty.
When Should You Consider a Different Sampling Strategy?
A different sampling strategy may be appropriate when the original recruitment arrangement systematically excludes groups essential to the intended inference.
For example, an online-only survey of digital exclusion may require alternative methods to reach people without reliable internet access.
A national institutional study might require a broader school or university sampling frame rather than recruitment solely through existing professional contacts.
Researchers should consider alternative approaches when no complete sampling frame exists.
However, switching recruitment methods can introduce new selection mechanisms and measurement differences.
Changes should therefore be planned, documented, and evaluated rather than assumed to improve representativeness automatically.
What If the Research Is Qualitative?
A narrow accessible population may be appropriate in qualitative research when participants have experiences directly relevant to the research question.
For example, researchers investigating the implementation of AI assessment policies may purposively recruit academic leaders who have managed those policies.
Statistical representativeness may not be the objective.
Nevertheless, restricted access can limit the range of perspectives available, particularly when institutions or gatekeepers exclude dissenting or less visible participants.
Researchers should examine whether the accessible cases provide sufficient experiential variation and conceptual relevance.
Claims of transferability should be supported by contextual description rather than unsupported population-wide assertions.
When Should the Study Be Redesigned?
Redesign may be necessary when the accessible population cannot support the central research question and the problem cannot be addressed through feasible recruitment or defensible analysis.
For example, a national prevalence study based exclusively on a small convenience sample of institutions may require a different sampling strategy if nationally representative estimation is essential.
Similarly, a study of digital exclusion that cannot recruit digitally excluded participants may be unable to answer its principal question adequately.
Possible responses include revising recruitment channels, changing the estimand, narrowing the research question, or postponing data collection until appropriate access can be secured.
The decision should follow the inferential requirements of the study, not merely the desire to preserve the original protocol.
How Should Limited Population Access Be Reported?
Methods sections should define the original target population, describe the accessible population, and explain the restrictions determining who could enter the study.
Report the institutions, geographic areas, recruitment channels, and relevant eligibility conditions.
Where possible, compare accessible and target population characteristics using suitable external information.
Describe any recruitment expansions, weighting adjustments, or sensitivity analyses and their assumptions.
In the discussion, distinguish findings directly supported by the study from conclusions requiring broader generalization.
Transparent reporting is preferable to silently redefining the population after recruitment.
