Recruiting people for a survey can be hard. Experimenters have to consider the randomness in their sample, the size of their sample, and the resources needed to bring their sample together. The work is worth it because you can’t survey an entire population! But what if you got your participants to do most of the work for you? You can, if you use snowball sampling as a sampling method.
What Is Snowball Sampling?
Snowball sampling is a type of sampling method in which the initial participants recruit other participants until the ideal sample size is reached. This takes some of the pressure off of experimenters to find willing subjects but is not a perfect method of sampling for every survey.
Snowball sampling is also called:
- Chain sampling
- Chain-referral sampling
- Referral sampling
- Network sampling
Examples of Snowball Sampling
Here’s how this might work:
- Researchers want to take a survey among people in sex work about the harassment they have experienced in their daily life. For obvious reasons, sex workers may not immediately come forward and trust people asking about their occupations and experiences. So the researchers gain the trust of a handful of sex workers who are willing to tell their stories. The researchers then ask the sex workers to talk to other people in the industry about the survey. Now that people in the industry know that the researchers have good intentions and will protect the workers, more people come forward and take the survey.
- A group of researchers wants to take a survey of people in the furry and pup and handler community. They recruit a small group of participants through an online ad, but they don’t have a sample size large enough. Rather than running another online ad, the researchers reach out to the participants to recruit friends and community members for the survey. These are pretty tight-knit groups!
- In 2002, professor Kath Browne published a paper on snowball sampling and the methods she used to reach non-heterosexual women. During this time period, it was harder to reach this population due to social stigmas surrounding homosexuality and queerness.
Pros of Snowball Sampling
Why would a researcher choose snowball sampling?
- Certain populations are hard to reach.
- Stigmatized populations are afraid to come forward and speak about their experiences.
- Researchers do not have to run recruitment campaigns or use their resources to find participants.
- Under-researched communities may only be accessible through snowball sampling.
- Researchers may gather unique data that is only available or known through snowball sampling.
Cons of Snowball Sampling
There are a lot of reasons to choose snowball sampling over other methods. But researchers must be aware of various biases or errors that come with snowball sampling and other forms of non-probability sampling methods.
It may take extra time to recruit participants through others. Participants don’t have the same motivation that researchers have to build a sample. They may forget to recruit people or give up easily. If the survey is time-sensitive, this method may not work out.
Participants may not properly represent the entire community. Let’s say you recruit a sex worker that works with high-profile clients, and you ask the sex worker to recruit others. They may only know people who work with high-profile clients. You are not very familiar with the community and do not know how common this is, so you may only survey one type of participant while neglecting other people in the community.
There are ways to avoid this type of bias. Doing extra research on the community and screening each participant for various demographics will give you a sense of whether your sample is proportionate to the population you’re researching.
Sample sizes are only reached if the participants recruit others. If participants fail to recruit others, researchers may not reach their sample size and the survey may not be considered legitimate. Researchers have to show that their survey has reached a significant sample size and that the sample properly reflects the community.
If the risk of the cons outweighs the pros, researchers may benefit from looking at other sampling methods to build a sample.
Other Sampling Methods
Snowball sampling is a type of non-probability sampling method. This means that probability, or relying on random draws, is not used to put together the sample. Both non-probability and probability sampling methods can put together a decent sample and/or put together a sample that contains bias. When putting together a sample, consider the different methods available to you and how they impact your budget, the accuracy of your sample, and the time it will take you to put together the sample.
Other Non-Probability Sampling Methods
Judgemental Sampling
Let’s say you want to take a sample of the people within your workforce. You know some of these people very well and select them because you think they will deliver the most informed opinion. This may be true, but it is also a form of non-probability sampling that may be influenced by experimenter bias.
Convenience Sampling
Maybe you want to take a sample of people in New York. You go out onto a street and grab the first 10 people who will answer your questions. This is convenient and you may get a diverse selection of people, but you cannot be sure that your sample accurately reflects the entire population until after you’ve collected data.
Web Panels
It’s easy to throw a survey online and ask people to take it. You can run ads with a link to the survey with or without an incentive, but you can also reach people through cheaper methods. There are obvious cons to this method: you can’t properly screen who is taking the survey, you don’t know if people are lying, and the people who take the survey may do so only because they have strong opinions that most of the population doesn’t share.
Probability Sampling Methods
Simple Random Sampling
The “simplest” way to create a sample is to assign everyone in the population a number, choose a handful of numbers to create a significant sample size, and call it a day. But this isn’t always possible. Let’s say you can assign a number to every person in the phone book. You choose random numbers…and then what? Not all participants are accessible or will want to participate. It may take extra time to successfully reach the people you have randomly sampled. If you need to reach them in person, it may also take travel time and extra resources.
Cluster Sampling
Populations may naturally contain certain groups. Americans, for example, can be grouped into the state they live in, their age, or their income bracket. Cluster sampling is a sampling method that identifies these different groups and picks different groups to survey. This can be very convenient if researchers have limited resources, but multiple demographics must be considered so that everyone in the population is properly represented. Only surveying people in New York, Massachusetts, and New Jersey may not provide a proper representation of the entire U.S. population.
Stratified Sampling
Stratified sampling still separates samples into groups, but participants as they relate to those groups are chosen differently. Researchers will choose individuals within different groups to be a part of the sample. This gives researchers more control over whether different demographics are represented appropriately in the sample size.
