What Should You Do When Selected Participants Cannot Be Contacted During Probability Sampling?

What Happens When Randomly Selected Participants Cannot Be Reached?

You have selected participants through a valid probability sampling procedure and begun recruitment. Some invitations are delivered successfully, but others bounce back. Several telephone numbers are inactive, some individuals never answer, and a few institutions cannot confirm whether selected participants are still affiliated with them.

Should you continue attempting contact, remove these individuals from the sample, or randomly select replacements?

These decisions matter because probability sampling establishes a known selection mechanism, but the people who ultimately provide data may differ systematically from those originally selected.

The central challenge is to manage noncontact without confusing it with ineligibility, changing the sampling procedure informally, or assuming that a random initial selection guarantees an unbiased responding sample.

Retain Selected Units, Follow a Documented Contact Protocol, and Assess Nonresponse

How to Manage Noncontact Without Compromising Probability Sampling

Why Does Noncontact Matter in Probability Sampling?

Probability sampling determines how eligible population units are selected. It does not guarantee that every selected unit will be successfully contacted or will participate.

Suppose 500 teachers are randomly selected from a verified institutional registry. If only 350 provide usable responses, the original 500 may constitute a valid probability sample, but the 350 respondents are not necessarily equivalent to a new simple random sample of the population.

Participation may depend on characteristics related to the research outcomes. Teachers with limited internet connectivity, heavy workloads, or low engagement with institutional communication may be less likely to respond.

If those characteristics are associated with the variables being measured, the responding sample may produce biased population estimates.

Importantly, a low response rate does not establish the magnitude of nonresponse bias, and a high response rate does not guarantee its absence. The relationship between response behavior and study outcomes is consequential.

What Is the Difference Between Noncontact, Refusal, and Ineligibility?

These outcomes should be distinguished because they have different methodological implications.

An unanswered email does not establish refusal. An inactive telephone number does not establish ineligibility. Similarly, failure to locate a selected person does not necessarily mean that the person is outside the target population.

Researchers should establish disposition categories before recruitment and apply them consistently.

How Many Contact Attempts Should Researchers Make?

There is no universally appropriate number of contact attempts.

The number and timing should reflect the population, recruitment mode, study burden, ethical approvals, and evidence about effective contact strategies.

A survey administered through institutional email might use an initial invitation followed by a limited number of reminders. A telephone survey may require attempts at different times or days to reach participants with different schedules.

Repeated attempts should be purposeful rather than excessive. Researchers must respect explicit refusals, requests to stop contact, institutional restrictions, and applicable privacy requirements.

A predefined protocol should specify the permitted contact channels, timing, maximum attempts, procedures for updating contact information, and circumstances requiring recruitment to stop.

Consistency improves transparency, although tailored follow-up may be justified when it is planned and documented.

Should Researchers Use Different Contact Methods?

Multiple contact modes may improve coverage when selected participants differ in their access to communication channels.

For example, a researcher studying teachers' digital competence should consider whether relying exclusively on online invitations could make teachers with weaker digital access less likely to participate.

Where authorized and feasible, alternative contact methods might include institutional email, postal invitations, telephone calls, or coordination through designated institutional contacts.

However, switching modes may introduce measurement differences if the questionnaire is administered differently across channels.

The recruitment protocol should therefore distinguish contact mode from response mode and assess whether any differences could influence the data.

Researchers should not obtain personal contact information from unauthorized sources merely to increase response rates.

What If the Contact Information Is Outdated?

Outdated contact details should be investigated without automatically excluding the selected unit.

For example, a teacher may remain eligible but have changed institutional email addresses. An updated address may be available through an authorized administrative record.

The researcher should verify whether the participant still meets the eligibility criteria and whether updated contact information can lawfully and ethically be used.

If the sampling frame contains widespread inaccuracies, the issue may extend beyond individual noncontact and indicate a broader frame-quality problem.

Correcting outdated sampling-frame records should follow consistent eligibility and documentation rules.

Where contact information cannot be updated, the selected unit should receive the appropriate unresolved-contact disposition rather than being silently removed.

Should Unreachable Participants Be Replaced?

Not automatically.

Replacing an unreachable participant with the next available person can undermine the original probability selection mechanism. The replacement may have a different probability of inclusion or may belong to a population segment that is easier to contact.

Even selecting a replacement randomly does not necessarily eliminate nonresponse bias. The original noncontact process may remain systematically related to the study outcomes.

Additional probability samples can sometimes be drawn using a documented design that preserves calculable inclusion probabilities. However, this is not equivalent to informal one-for-one substitution.

The methodological consequences of replacing nonrespondents in probability sampling should be considered before any supplemental selection occurs.

How Should Contact Attempts and Outcomes Be Recorded?

Researchers should maintain a secure contact-disposition log linked to the selected sampling units.

Depending on the study, useful information may include the original sampling identifier, eligibility status, contact channels used, dates or rounds of attempts, delivery failures, successful contact, refusal, and completion status.

Where permissible, researchers may also retain limited administrative characteristics useful for evaluating nonresponse, such as school sector or geographical region.

The log should distinguish selected units from completed responses. It should also distinguish confirmed ineligibility from unknown eligibility.

Personal contact details should be stored securely, access should be restricted, and information should be retained only as permitted by the approved research protocol and applicable requirements.

How Do You Calculate a Response Rate?

A response rate describes the proportion of eligible sampled units that provide qualifying responses, subject to the specific disposition definitions used.

For a simple situation in which eligibility is established for every selected unit, one basic calculation is:

Real surveys may contain partial interviews, cases with unknown eligibility, and different categories of nonresponse. The American Association for Public Opinion Research provides standardized response-rate definitions that account for such distinctions.

Researchers should specify which response-rate definition they use and report enough disposition information for readers to understand the calculation.

How Can Researchers Evaluate Nonresponse Bias?

Nonresponse bias arises when the expected values of estimates based on respondents differ from the corresponding population values because of the response process.

The risk depends on whether response propensity is associated with the study outcomes, not simply on how many participants fail to respond.

Researchers can investigate potential bias using information available for both respondents and nonrespondents.

For example, a university may know the academic programs, year levels, or enrollment categories of all randomly selected students. Comparing those characteristics between respondents and nonrespondents may reveal differential participation.

However, similarity on observed characteristics does not establish that respondents and nonrespondents are similar on unobserved outcomes.

Where suitable information exists, researchers may also compare estimates with reliable external population benchmarks, examine response patterns across recruitment waves, or conduct sensitivity analyses under alternative assumptions.

Each method has limitations. Nonresponse-bias analysis should therefore describe what the available evidence supports and what remains uncertain.

Can Weighting Correct Nonresponse?

Nonresponse adjustments may reduce bias when suitable auxiliary variables are available and the assumptions underlying the adjustment are credible.

For example, if response rates differ between known strata, researchers may adjust the base weights of respondents within appropriately defined adjustment classes.

A simplified adjustment factor for a class with equal base weights is the reciprocal of its observed response rate.

More complex designs may require weighted response adjustments, response-propensity models, calibration, or other methods.

Researchers should distinguish these adjustments from the original inverse-probability design weights.

Adjustments can also increase variance, particularly when some respondents receive large weights. Their potential benefits and limitations should be evaluated together.

What If Noncontact Is Concentrated in Particular Population Groups?

Noncontact may differ across geographical regions, institutional sectors, age groups, or other characteristics.

For example, teachers in remote schools may be harder to contact because institutional communication systems are less reliable.

If those schools also differ in technology access, a survey of digital teaching practices could be affected by systematic nonresponse.

Researchers should examine contact and completion outcomes across relevant frame variables when the necessary information is available.

Targeted follow-up may be appropriate when planned within the recruitment protocol and permitted by the relevant ethical requirements.

However, concentrating additional effort on underresponding groups does not automatically remove bias. The final analysis should still consider remaining differences between respondents and the target population.

When Should Recruitment Stop?

Recruitment should stop according to predefined operational and ethical criteria rather than simply when the desired number of completed questionnaires has been obtained.

Relevant considerations include the approved contact schedule, participant preferences, resource limits, and whether additional attempts are likely to produce useful responses.

If the intended number of responses is not achieved, researchers should evaluate the consequences for statistical precision and the feasibility of further probability-based recruitment.

Continuing indefinitely is not necessarily justified, particularly when repeated contact could become intrusive.

The final report should describe the achieved sample, unresolved noncontact, and limitations rather than presenting the planned sample size as though it had been fully realized.

Managing Unreachable Teachers in a Probability-Based Survey

The researcher retains all 600 original selection records and applies the protocol consistently.

Suppose the final outcomes are as follows:

Assume eligibility has been established for every selected unit. The number of confirmed eligible units is 600 − 30 = 570.

The researcher compares respondents and nonrespondents using school sector and geographical region, which were recorded in the original frame.

Suppose the analysis indicates lower participation among teachers in remote schools. Because geographical location may also be related to digital infrastructure, the researcher identifies this as a potential source of nonresponse bias.

Appropriate weighting adjustments and sensitivity analyses are considered, with their assumptions documented.

The researcher does not simply recruit 180 additional convenient teachers to restore the original target of 600 completed questionnaires.

Instead, the final report distinguishes the original probability sample from the achieved respondent dataset and explains how noncontact, refusal, and ineligibility were handled.

Common Mistakes When Selected Participants Cannot Be Reached

How to Respond to Noncontact While Preserving Research Integrity

The appropriate response begins with identifying why the selected unit has not provided data. Contact failure, refusal, uncertain eligibility, and confirmed ineligibility should not be handled as interchangeable problems.

For the methodology section, report the original number selected, contact procedures, recruitment period, disposition categories, response-rate definition, and achieved sample size.

Describe any nonresponse adjustments and the information used to support them. Where nonresponse bias cannot be ruled out, state the limitation without implying that the initial random selection eliminates the concern.

The central methodological responsibility is to preserve the distinction between probability selection and subsequent participation.

Before Closing Recruitment, Verify How Noncontact Was Managed

Frequently Asked Questions About Unreachable Selected Participants

Manage Noncontact Transparently Rather Than Replacing Participants Informally

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.