Why the way we recruit matters

A sampling technique is the procedure used to recruit participants from a population. Researchers intend their findings to apply to a particular group of people: the target population. The sample is the subset of that population who take part in the research.

Recruitment involves a trade-off. Researchers need a practical way to find participants, but those participants also need to reflect the population adequately. A representative sample has relevant characteristics that reflect those of its target population. The word relevant matters here. If age could affect the behaviour under study, an age imbalance matters more than a difference unrelated to that behaviour.

When some members of a population are more likely to be included than others, the resulting systematic distortion in sample composition is called sampling bias. It can limit generalizability: the extent to which findings apply beyond the people or circumstances studied. A large sample isn't automatically representative. Recruiting more people through the same biased route may simply reproduce the imbalance.

Snowball sampling

In snowball sampling, initial participants help identify or recruit further eligible participants through their contacts. A researcher investigating experiences of an uncommon workplace arrangement, for example, could start with a few eligible workers and ask them to pass an invitation to others.

Access is the main advantage. Through personal connections, researchers can reach people who are difficult to locate through public recruitment. Potential participants may also be more willing to consider taking part when an invitation comes from someone they trust. Recruitment can grow even without a complete population list.

That same network is the main weakness. People often know others with similar backgrounds or experiences, so recruitment may stay within a narrow social circle. Anyone outside the network has little chance of inclusion. Agreement among several participants doesn't necessarily show that their views are widespread across the population.

Asking participants to pass on an invitation, rather than disclose someone else's details, can protect privacy. A trusted referral must still be an invitation, though—not pressure to participate.

Random sampling

With random sampling, every member of the target population has an equal chance of selection. To do this, a researcher could give each person on a complete population list a unique number, then use a random-number generator to select the people to invite.

Selection doesn't depend on whom the researcher prefers or can conveniently reach. Nor does it depend on which participants volunteer first. This reduces systematic selection bias and can support generalization to the population from which the sample was drawn.

Random selection does not guarantee a representative result, however. Chance alone may leave a small sample unbalanced. The procedure also needs a sufficiently complete and accurate population list, which may be unavailable or costly to obtain. Some selected people may decline; those who take part may differ from those who refuse.

Be precise about the procedure. Choosing people who happen to be nearby is not random sampling. Random sampling concerns who enters the study, not which experimental condition participants enter.

Stratified sampling

In stratified sampling, researchers divide the population into relevant subgroups and select participants from each one. Proportional stratified sampling gives each subgroup the same proportion of the sample as it has in the population. Selection within each subgroup can then be random.

A researcher studying employees' experiences, for example, could divide a workforce by job category and recruit from each category in proportion to its presence in the workforce. Important subgroups are deliberately included rather than left entirely to chance. That's the advantage.

The diagram contrasts unrestricted random selection with proportional stratified selection. Both can use random selection. Only the stratified procedure first controls representation of the chosen subgroups.

The drawback is the information and organization needed. Researchers must identify which subgroups matter and who belongs to them. They also need to know each subgroup's proportion in the population. Recruiting the required participants from every subgroup may take time, especially if some groups are less accessible.

Stratification protects representation only for the characteristics used to create the subgroups. Matching job-category proportions doesn't automatically make a sample representative in age or experience. It doesn't ensure representation in willingness to discuss difficulties either. Biased recruitment within subgroups remains biased; stratification does not remove it.

Opportunity sampling

For opportunity sampling, researchers recruit eligible people who are readily available at the time and place of recruitment. A researcher could, for example, invite available visitors in a community building to complete a short questionnaire.

Recruitment is quick and inexpensive, with straightforward access to participants. This makes the technique useful when resources are limited or a researcher needs a feasible way to recruit for a small practical investigation.

But availability isn't neutral. The choice of building influences who can be approached, as do the day and recruitment time. People who never visit that location have no opportunity to participate. Those who are present may share characteristics relevant to the findings, so generalizing from this sample to all local residents would require caution.

The distinction to remember is that the researcher approaches accessible people. Agreeing to participate does not turn the procedure into self-selected sampling.

Self-selected sampling

In self-selected sampling, individuals volunteer in response to an invitation or advertisement. A researcher might advertise a study through a public notice and accept eligible people who respond.

A broad invitation can reach potential participants without a complete population list. Volunteers may be motivated to take part and willing to give detailed responses. Neither commitment nor completion is guaranteed, though.

The weakness lies in the decision to volunteer: it may be related to the topic. People with strong opinions may be more likely to respond, as may those with particularly positive or negative experiences. Having more available time can also make a response more likely. The advertising channel matters too. People who never encounter the invitation cannot volunteer, so a large response does not establish representativeness.

Unlike opportunity sampling, potential participants initiate contact in response to the invitation. Both techniques can be practical. They provide different routes into the sample, however, and exclude people for different reasons.

Weighing advantages against disadvantages

Judge each technique against the research purpose and circumstances. Random sampling reduces selection bias, but it needs an adequate population list. With stratified sampling, researchers secure representation of chosen subgroups at the cost of more organization and a need for population information. Snowball sampling improves access through networks, though connected people may be overrepresented. Opportunity sampling makes recruitment practical while tying inclusion to availability. Self-selected sampling attracts willing participants, but inclusion depends on noticing and responding to an invitation.

A strong judgement explains why a particular advantage or limitation matters for the intended population and the behaviour being investigated. No technique guarantees an unbiased sample. Practical accessibility is not the same as representativeness.