Home Statistical Dictionary Snowball Sampling
Snowball Sampling
A non-probability sampling method in which existing study participants recruit additional participants from among their acquaintances, useful for reaching populations with no available sampling frame, especially hidden or hard-to-access groups.
In Plain English
Snowball sampling starts with a small number of initial participants and grows the sample by asking each one, 'do you know anyone else who fits this study?', with each new wave of recruits potentially bringing in more contacts, like a snowball picking up more snow as it rolls. It's often the only realistic way to study populations that have no public list anywhere, people who use a particular illegal drug, members of a tight-knit but private community, survivors of a specific rare experience, since these groups can't be reached through a phone book or address list, but can often be reached through their own social networks.
Definition
Snowball sampling begins with an initial set of seed participants, identified by the researcher or through convenience or judgment sampling, and expands the sample through successive waves of referral, each participant is asked to nominate or introduce additional eligible individuals from their own social network, who are then recruited and asked to do the same. Snowball sampling is a non-probability method because the final sample's composition depends on the structure of participants' social networks and their willingness to refer others, neither of which is known or controlled by the researcher, units with larger, more connected, or more cooperative social networks are systematically more likely to enter the sample, but it remains one of the few practical approaches for studying populations with no accessible sampling frame, particularly hidden, stigmatized, or hard-to-reach groups where members can typically only be identified through their existing social ties to one another.
Properties
- Snowball sampling's fundamental limitation is that recruitment probability depends on an individual's position and connectedness within the underlying social network, a factor entirely outside the researcher's control and generally unobservable, which makes the resulting sample's relationship to the true target population impossible to characterize with standard design-based methods.
- Snowball sampling remains, despite this limitation, often the only practical method for reaching populations that have no public sampling frame and can only be identified through peer referral, such as people who inject drugs, sex workers, or members of small, geographically dispersed diaspora communities, which is why it persists as a standard tool in public health and sociology research.
- Respondent-driven sampling was developed specifically to address snowball sampling's uncontrolled recruitment bias, by tracking who recruited whom and each participant's network size, and using that information to construct statistical weights, it attempts to convert snowball-style recruitment into something closer to a valid probability sample, though this remains an area of active methodological debate.
At a Glance