Undercoverage bias is the systematic error that occurs when your sampling frame — the actual list of people you draw your sample from — leaves out part of the population you intend to study, so whole groups have no chance of being included. If you survey student mental health using only an automated email, the students who never check their inbox or lack reliable Wi-Fi are silently excluded, and your findings paint a ‘sunny’ picture that ignores the people who may be struggling most. It is one of the most common and most overlooked threats to the validity of survey research.
This guide covers exactly what undercoverage bias is, what causes it, a worked example you can follow step by step, real-world examples across survey modes, how it differs from related sampling problems, and a practical checklist for avoiding and reducing it in your own dissertation or study.
What Is Undercoverage Bias?
We live in a world of ‘big data’, but big data is not always complete data. Undercoverage bias occurs when the “sampling frame” — the list or mechanism you actually select people from — does not match the true target population. When a specific group is left off that list, they are “undercovered”: they have a zero or near-zero probability of being sampled, no matter how large your final sample is. Because their views, circumstances and experiences never enter the data, the results are skewed in a predictable direction, and the researcher often never notices the gap.
The crucial word is frame. A frame is the operational stand-in for the population: an email list, a telephone directory, a register of households, the people walking past a particular spot at a particular time. Whenever the frame is narrower than the population — or excludes a group that systematically differs — you have undercoverage. If a study on student mental health relies on a single email survey, the frame is “students who use that email regularly”, not “all students”. The difference between those two lists is exactly where the bias lives. This is why undercoverage is treated as a recognised form of research bias rather than a mere logistical nuisance.
Undercoverage bias matters most when the excluded group differs from the rest of the population on the very thing you are measuring. This makes it a particular risk in survey-based research, where the frame is built before any data is gathered. Leaving out left-handed people from a survey about commuting habits may not bias your result at all, because handedness is unrelated to commuting. But leaving out students without reliable internet from a survey about digital wellbeing biases the result heavily, because connectivity is directly tied to the topic. Who you leave out is just as important as who you include — and the harm depends on whether the missing group is systematically different.
A Worked Example: How Undercoverage Distorts the Number
It helps to see undercoverage bias as a calculation rather than an abstract idea. Imagine a university wants to study student stress levels and decides to hand out surveys to every student walking into the campus library at 10:00 AM on a Tuesday.
- The Population: All 1,000 students at the university.
- Who the frame reaches (600 students): Library-going, on-campus, daytime students — relatively engaged, true mean stress 5.0.
- The Undercoverage (400 students): Students who work full-time, students who only take online classes, and students so overwhelmed they are not coming to campus at all — true mean stress 7.5. They have zero chance of being handed a survey.
Reported result: the survey can only ever average the 600 covered students, so it reports a mean stress of 5.0 out of 10, while the true population value is (600 × 5.0 + 400 × 7.5) ÷ 1000 = (3000 + 3000) ÷ 1000 = 6.0. Undercoverage has understated stress by a full point, and a wellbeing officer reading the report would wrongly conclude that students are coping well — missing precisely the disengaged, overworked and remote students who need support most. Notice that no amount of extra surveying at the library fixes this: the missing 400 are structurally outside the frame.
This last point is what separates undercoverage from a small sample. Collecting more responses at the same library makes your estimate more precise but no less biased — you are simply measuring the wrong population more accurately. The only real fix is to widen the frame so the missing 400 can be reached, a point we return to in the avoidance section below.
How Undercoverage Bias Happens (At a Glance)
What Are the Causes of Undercoverage Bias?
Undercoverage is rarely deliberate. It creeps in through how the frame is built and how participants are reached. Recognising these causes at the design stage is the first line of defence. The main drivers are set out below.
Problems With the Sampling Frame
A sampling frame is the list or representation of the intended audience from which the sample is chosen. If that list is incomplete, outdated or wrong, certain people are silently excluded. Individuals in remote communities, travelling people, or those without fixed addresses are frequently missing from official registers, which guarantees undercoverage before a single response is collected.
Non-Response From Certain Groups
Non-response bias occurs when chosen participants decline to take part or fail to reply. It is technically distinct from undercoverage — non-respondents were in the frame, whereas undercovered people never were — but the two interact. When non-response clusters in particular groups, for example if people from lower socioeconomic backgrounds are consistently less likely to reply, those groups end up underrepresented in just the same way, compounding any coverage gaps.
Selection Bias in Recruitment
When the method used to pick participants fails to capture the whole population, selection bias feeds directly into undercoverage. Studies that rely on convenience sampling or voluntary participation systematically over-reach the available and motivated while under-reaching everyone else, so the frame quietly shrinks to “people who happened to be easy to find”.
Exclusion Criteria
Some studies and surveys apply criteria that screen particular populations out. A health study that excludes everyone with pre-existing conditions, for instance, introduces undercoverage if those conditions are common in the real target population. Exclusion criteria are sometimes necessary, but each one narrows the frame and should be justified explicitly.
Access Restrictions
Restricted access to certain groups produces undercoverage sampling bias. An online-only poll that omits people without internet access underrepresents those with poor connectivity or those marginalised by the digital world — often the very groups whose experiences a study most needs to capture.
Language or Cultural Barriers
If survey materials are available only in one language, or fail to account for cultural sensitivities, people from other linguistic or cultural backgrounds are effectively excluded. They may never receive the invitation in a form they can engage with, or may decline because the instrument does not speak to their context, leaving a coverage gap that mirrors community divisions.
Undercoverage Bias vs Related Sampling Problems
Undercoverage is easily confused with its neighbours, and the distinctions matter because each calls for a different fix. The table below sets them side by side. For the wider family of distortions and how they interrelate, see our hub on research bias.
| Bias | Where the problem occurs | Core idea | How to fix |
|---|---|---|---|
| Undercoverage bias | The sampling frame | A group is left off the list, so it has zero chance of selection — the frame is narrower than the population. | Build a complete, current frame; use multiple modes to reach the missing group. |
| Selection bias | How the sample is chosen | Even from a fair frame, the selection method favours some people over others (e.g. convenience sampling). | Use random / probability sampling so everyone has a known chance. |
| Non-response bias | Who replies | People in the frame are invited but do not answer, and non-respondents differ systematically. | Reminders, incentives, shorter surveys, weighting. |
| Ecological fallacy | How results are interpreted | A reasoning error — drawing individual-level conclusions from group-level data, not a sampling flaw. | Match the level of analysis to the level of the claim. |
A useful rule of thumb: undercoverage is a problem with the list you draw from, selection bias is a problem with how you draw, and non-response is a problem with who answers. All three are distinct from the ecological fallacy, which is an interpretation error rather than a sampling one. Undercoverage is also different from purely cognitive distortions such as recency bias, where a researcher over-weights the most recent information — that affects judgement, not who ends up in the frame.
Examples of Undercoverage Bias
Once you know the pattern, undercoverage bias is easy to spot. Here are six everyday settings where it routinely appears, each tracing back to a frame that quietly excludes part of the population.
Telephone Surveys
Conducting a survey exclusively through telephone interviews excludes people without landlines or mobile phones, and increasingly those who screen unknown numbers. This underrepresents particular socioeconomic and age groups, so the frame — “people reachable by phone” — is narrower than the population it is meant to represent.
Online Surveys
If a survey is available only online, people with poor internet access or low digital confidence are left out. The resulting undercoverage bias is most acute for older adults, lower-income households and people in remote areas — precisely the groups often most relevant to social research.
Household Surveys
Household surveys built on address registers omit people in non-standard living situations: homeless populations, those in group or institutional accommodation, and unusual housing arrangements. Because the frame is a list of conventional households, anyone outside it is structurally undercovered.
Employment Surveys
Workplace surveys can suffer undercoverage if particular workforce groups — part-time staff or remote workers — are missing from the sampling frame. The result is a distorted picture of the workforce’s traits and experiences that over-represents office-based, full-time employees.
Geographical Surveys
A survey focused on a defined geographic area can still leave out certain neighbourhoods or regions within those boundaries — for example, areas with incomplete address data. Where the target population is not fully captured inside the chosen boundaries, undercoverage bias follows.
Phone-In Radio Shows
“Phone-in” radio programmes invite listeners to call in and share their opinions, but the only people heard are those who choose to call and happen to be listening. Excluding everyone who does not call in or tune in introduces severe coverage and self-selection bias, skewing any impression of public opinion toward an unrepresentative, vocal minority.
“Undercoverage occurs when some groups in the population are left out of the process of choosing the sample.” — Moore, McCabe & Craig, Introduction to the Practice of Statistics
How to Avoid and Reduce Undercoverage Bias
Avoiding undercoverage means working on two fronts: preventing coverage gaps when you design the frame and collect data, and correcting for any remaining gaps in analysis. The strongest studies do both, and build these safeguards in from the outset rather than bolting them on at the end. Start by developing a comprehensive sampling frame that reflects the full target population — one that is correct, current and includes every necessary subgroup.
1. Use Random and Probability Sampling Methods
Choose participants using probability methods — simple random, stratified or cluster sampling — from the most complete frame you can build. When every member of the population has a known, non-zero chance of inclusion, undercoverage is far less likely. Stratified sampling is especially useful for guaranteeing that small but important subgroups are represented.
2. Combine Multiple Data Collection Methods
Use several methods of data collection to reach people a single channel would miss. Pairing telephone polls with online surveys, in-person interviews or paper questionnaires widens the frame and lifts coverage among hard-to-reach groups. Mixed-mode designs are one of the most effective practical defences against undercoverage.
3. Apply Adjustments and Weighting
After fieldwork, use weighting to account for groups that ended up underrepresented. Where you know the true population profile, giving heavier weights to people from undercovered groups re-balances the sample toward the population. Weighting is a correction, not a cure — it cannot fully repair a group that was completely absent — but it meaningfully reduces residual bias.
4. Actively Engage Underrepresented Groups
Boost participation among hard-to-reach groups through targeted outreach: focused recruitment, community partnerships and collaboration with organisations that already have the trust of the people you need to reach. Active engagement turns groups that would have been undercovered into genuine participants.
5. Use Multilingual and Culturally Sensitive Approaches
Make survey instruments, materials and communications available in the relevant languages and sensitive to cultural context. Removing linguistic and cultural barriers brings in communities that single-language instruments would silently exclude, closing one of the most common coverage gaps in diverse populations.
6. Reach Non-Institutional Populations
Plan deliberately for people who fall outside standard registers — homeless people, those in group living arrangements, and people without permanent addresses. Alternative sampling strategies, partnerships with regional organisations and targeted field outreach are often the only way to cover these populations at all.
Why Undercoverage Bias Threatens Validity
Undercoverage bias is ultimately a threat to the reliability and validity of your conclusions, and to how far they generalise. Even a large, carefully analysed dataset gives the wrong answer if a whole group was never eligible to appear in it. For a dissertation, examiners look for evidence that you defined your population precisely, scrutinised your frame for gaps, and reflected honestly on who might have been excluded and how that could shape your findings. Acknowledging a coverage limitation, and explaining how you tried to close it, is a mark of methodological maturity rather than weakness — and it is one of the clearest signals that you understand the difference between the population you wanted and the sample you actually reached.
Looking for research help?
Research Prospect to the rescue then!
We have expert writers on our team who are skilled at helping students with their research across a variety of disciplines, guaranteeing 100% satisfaction. Learn More about how we can support your study.
Worried your sample won’t represent your population?
Our methodologists help you build a complete sampling frame, choose the right sampling method and defend your findings against coverage gaps.
