Now that we have talked about probability sampling, let’s focus on non-probability sampling in research. We saw that we need a sampling frame whenever we want to go for a probability sampling technique. But many times, a sampling frame is not available, or sometimes even the target population is not completely known. This is especially true when the researcher is trying to study a vulnerable or stigmatized population. For example, a researcher might be interested in risk-taking behavior among ecstasy (drug) users. Or the emotional toll on children exposed to domestic violence. Or decision making in runaway teenagers. It is unlikely that s/he will find a sampling frame. It is under such circumstances that research might go for non-probability sampling. Applied research often faces such difficulties. This type of sampling design can also be of choice due to convenience as it is sometimes easier to implement. Under non-probability sampling, the chance to be in the sample for any individual is unknown. Here the researchers use their judgment to select an individual in the sample based on pre-established characteristics.
This ability to construct their own sample can be very useful in certain circumstances such as when a researcher wants to investigate certain subcultures or closed cultures or sensitive topics. It can also be very useful when the researcher wants to study certain traits of specific groups, and where generalizability of the results is not the primary objective. The non-probability sampling can also be used when not much information is available on the subject or it is a new area of exploration. Non-probability sampling designs are great when the study aims to do an in-depth review of few cases rather than generalizability of the results.
At the same time, since there is no known chance of an individual or element to be selected in the sample, it makes this design more vulnerable to sampling error (specifically, bias). Additionally, sampling error cannot be estimated in such cases and hence element of bias cannot be completely ruled out. However, as mentioned earlier, this type of sampling design might be the only way to study some subject matters. It does have the advantage of less time and money required to complete the study.
General limitations of non-probability sampling
- Since the sample is not likely representative of the population, it is difficult to generalize the findings.
- Lower external validity.
Non-probability sampling techniques
- Convenience sampling (aka haphazard sampling)
- Purposive sampling (aka judgment sampling)
- Quota sampling
- Snowball sampling (aka respondent-driven sampling)
Convenience sampling
Convenience sampling is named so as researchers select the sample which is most convenient to her/him. That is, participants are willing and available to be studied and possess the characteristics that the researcher wants to investigate. It requires the least planning. Or convenience is a major advantage. Convenience sampling can also be good for pilot studies and testing instruments. But a major obvious disadvantage is such samples are likely biased as the researcher makes no effort to know the population. For example, if a researcher interviews people, who come to watch a basketball game, s/he leaves out anyone who did not come to watch that game and is not interested in participating. Additionally, the sample will only have people who are likely to have a strong opinion about the policy.
Because of a likelihood of bias, for the most part, researchers avoid using convenience sampling. But if it is the only available option, then researchers have to be careful and include all relevant information about the sample (such as demographics and other important characteristics) so that validity of claims can be ascertained. Additionally, providing a clear description of how, when, and why this sampling technique was used along with a sample description can help reduce the bias. Another way to decrease bias is by repeating the study with similar samples.
Advantages of convenience sampling
- Easy to obtain.
- Relatively low cost.
- Good for exploratory or pilot studies or testing instruments.
- Providing a sample description might help reduce potential bias.
Disadvantages of convenience sampling
- Results are not very generalizable.
- The sample is most likely biased.
- Lower external validity.
Purposive sampling
Under this kind of sampling technique, the researchers exercise their judgment to pick individuals who fit the research criteria and think might represents the target population. Using this sampling design requires previous knowledge of the population and the sample should possess characteristics that align with the purpose of the research (hence the name). It is, therefore, imperative that the researcher does a thorough literature review and has a complete understanding of the field before starting.
For example, if a researcher wants to study unofficial power hierarchy in a school, s/he can decide to interview union leaders, the principal, the principal’s secretary, etc. Or if a researcher wants the know about parents’ preference of twin strollers, s/he can talk to parents of twins below the age of three.
This technique is a little better than convenience sampling as research has more control over who can be in the sample. Qualitative research often uses this sampling design as it provides flexibility to look deeper into a specific problem rather than aiming for generalizability. It is extensively used for grounded theory development. It allows research to ask more ‘why’ questions rather than ‘what’ questions. It allows them to situate a behavior or traits within the context.
Unlike probability sampling, not all participants are equal here. One articulate and a well-informed participant can advance the research much better compared to a few randomly chosen ones. Hence researchers can afford to use their judgment to maximize their time and effort in advancing their purpose or objectives. There is no overall sampling strategy that informs the researcher how many individuals are needed. One has to take what they can get. An obvious disadvantage is that if the researcher makes an error of judgment, the sample will not be representative and can waste precious time and effort. A standard way to overcome such bias is to provide a detailed sample description. Additional details such as who was chosen and who was not and why can help enhance the generalizability of the results. Purposive sampling design is an excellent choice for a pilot study, or a case study (critical or intensive), or harder to find populations such as illegal migrants, refugees, drug addicts. Overall, it has a lot to offer under appropriate precautions.
Advantages of purposive sampling
- Researchers have better control over who gets selected in the sample compared to convenience sampling.
- If done properly can represent the target population.
- An excellent choice for pilot or exploratory studies, hard to find population, and case studies that need cultural explanations.
Disadvantages of purposive sampling
- The researcher needs to have prior knowledge of the population.
- An error of judgment can lead to a non-representative population.
Quota sampling
It is most similar to stratified sampling where researchers set different proportions of quota (categories) to be in the final sample and then get the sample through convenience sampling or purposive sampling. Generally, the quota is based on demographic characteristics such as age, gender, income, location, education, etc. For example, a researcher might want to study the effect of the new health policy. S/he might create a two-by-two table for gender and age so that each quota is adequately represented such that s/he has younger men, younger women, older men, and older women. Researchers can also expand this process to fit respondents to other criteria such as income, education, location, etc. They can create quotas to ensure a representative sample. But doing so does not guarantee that the sample will be unbiased on other characteristics outside the quota system.
This technique was often used in survey research as an alternative to random sampling before the 1950s as it is quick and cheaper. In 1948, the Chicago Tribune famously predicted Thomas Dewey’s presidential victory beating Harry Truman, based on quota sampling, while the votes were still being counted.
The obvious limitation is that such samples are often not representatives of the population. Because the quota sample tends to be biassed towards people you can find easily.
Some researchers, however, argue that this sampling technique was developed not to create a representative sample but rather avoid bias on key characteristics of interest. For example, a study on psychological effects of aging can include older men and women from different races, and then investigate the effects of other factors such as socioeconomic status, type of work, mental stimulation available, access to quality healthcare, etc. within these key categories.
Advantages of quota sampling
- Relatively easy to implement.
- Allows research to have better control over the composition of the sample.
- If done properly can outperform probability sampling at lower cost and hassle.
- Useful design for a targeted audience or market research.
Disadvantages of quota sampling
- Often not representative of the target population and is probability biased.
- Can give an illusion of representative of the population.
- The effectiveness of this technique is inescapably linked to the researcher’s thoroughness of analysis and inference.
Snowball sampling
As the name suggests, here the initial respondents are selected because they match the predefined criteria and then additional respondents are selected based on referrals. That is, the research asks the respondents to identifies others to be part of the sample because they are particularly knowledgeable about the topic. Snowball sampling is a good way to extend purposive sampling when the existing resources of getting new participants are exhausted. This technique is often used when the research is done on a sensitive topic or closed cultures or on the members who are stigmatized or reclusive or harder to reach. It is also used when the total size of the target population might be unknown or very few members exists. For example, snowball sampling might be a good way to understand the number of homeless people in a city, people with drug addiction, children of domestic violence, people who experienced sexual assault, etc. Sometimes, snowball sampling can also be used to build a sampling frame after which even probability sampling can be used.
However, referral methods can increase your sample size but can also take away researchers’ ability to control who will be in the sample. It also eliminates the possibility to include non-respondents and might quite possibly lead to non-observation bias or a biased sample. One way to guard against this is to get the initial respondents to be as diverse as possible.
Advantages of snowball sampling
- Best sampling technique when the target population is not known.
- Best sampling technique to get hard to reach population.
- Can be used to build a sampling frame.
Disadvantages of snowball sampling
- Researchers can quickly lose the ability to control who gets in the sample.
- It can lead to non-observation bias.
- Sometimes the valid population can only be defined after the sample is formed.
Bibliography
Bernard, H. R. (2006). Research methods in anthropology: Qualitative and quantitative approaches (4th ed.). Lanham, MD: AltaMira Press.
Dooley, D. (2001). Social research methods (4th ed.). Upper Saddle River, New Jersey: Pretence Hall.
Given, L. M. (2008). The Sage encyclopedia of qualitative research methods. Los Angeles, CA: Sage Publications.
Levin, R. I., Rubin, D. S., Siddiqui, M. H., & Rastogi, S. (2017). Statistics for management (8th ed.). Noida: Pearson India.
Cite this article (APA)
Trivedi, C. (2020, December, 12). Non-probability sampling in research. Conceptshacked. Retrieved from https://conceptshacked.com/non-probability-sampling/
