49 Sampling

Sampling is an essential component of social science research. In fact, the absence of a defined sampling method is actually a form of sampling. As mentioned in Chapter 7 (Methodology), data collection requires defining the unit of analysis and before gather information about it. In social research, that “unit” is often human beings, institutions, objects (e.g, newspapers, photographs, books) or social parameters (such as marriage, divorce, birth and death). As you will recall from Chapter 3, ethical concerns should guide any form of data collection. To guarantee quality data, you should be placing limits on the population/text you want to observe and then on the behaviour/aspect of the population you want to observe (Bhattacherjee, 2012) from the onset. This will attend to some of the issues with coverage and sampling errors discussed earlier. A guiding question can be: what is the kind of text, people, place, or behaviour that I want data about? (e.g., “2nd generation  Indian immigrants in BC” or “articles referring to Uber in 2020”). The next steps are to  think about: (1) how you will access that data, and (2) what claims you seek to make about the ‘unit.’ It is also important to consider the goal of your data collection. For example, if it is to generalize about a population, you will need to consider a large sample size (also, see discussion on research paradigm later in the chapter).

Effective Sampling Frame for Quality Data

Once a population or unit of analysis is determined, you will need to consider developing a sampling frame. Bhattacherjee (2012, p. 66) defines the sampling frame as “an accessible section of the target population (usually a list with contact information) from where a sample can be drawn”. In other words, it is a list of all those within a population (which could be institutions, households, individuals, social artifacts etc) that can be sample. It is crucial that you are able to access the units in your sampling frame (see Chapter 7, Methodology). Another important tip is that you should consider ways of motivating your units to participate in your research (if they are people). This means clearly explain to them why your research is significant to them. Doing so will help reduce non-response errors and contribute to you gathering quality data.

The Sampling Process as Critical to Quality Data

Besides motivation, there are two other common practical problems that should inform your sampling process. The first concerns the consistency of your data. This means that the context under which data is collected should be similar across all cases. In essence, how data is produced should be similar for all cases. Cases should have similar characteristics (same population, behaviour, event), and relate to the same setting (interviewing drug users in a police station vs. in their home will likely produce biases/inconsistencies). The second concerns the substance of your data, i.e., the data you have is capable of addressing your RQ in an unbiased manner. In sampling data, you ought to include any details about the data that may ‘corrupt’ it (such as its biases, your biases, and problems in how it is sourced) in field notes. Be reflexive about the major problems you note in finding your data, and any changes you have in your data collection processes throughout gathering. This latter process will improve the transparency in discussing methods, noting limitations of your data and the major turns in your research process. Common potential limitations may be lack of data on an important aspect of your research question, biases in your sampling (e.g. in non-probability sampling methods such as judgmental, convenience and snowball sampling), and other contextual interferences with primary data (e.g., the setting of an interview interrupts the flow or content of conversation). Paying attention to these details will reduce the likelihood of you having incomplete or irrelevant data at the end of data collection.

Choosing the Appropriate Sampling Technique

Because the potential of sampling error to undermine data quality is so grave, it is important that an appropriate sampling technique is selected for your study. Common sampling methods include systematic sampling, cluster sampling, quota sampling, and snowball sampling (see Bhattacherjee 2012, p. 67-68 for some details of the different sampling techniques).

As you will remember from your Research methods courses, simple random, systematic and cluster sampling are three forms of probability sampling. Probability sampling selects samples at random based on the theory of probability in order to limit any bias which can influence the probability that the data is not representative of the population (i.e. not skewed towards exceptions in a population). The key to probability sampling is that every member of a given population has a known and non-zero chance of being selected in the sample. These are difficult conditions to meet, hence, it is unlikely that you will be using probability techniques in your undergraduate work. Nonetheless, many government agencies and other large organizations employ probability techniques in their studies so if you are using secondary data, be sure to check the methodology, and, in your analysis, account for any limitation or strength offered by the sampling technique used. Accordingly, we briefly highlight four common probability sampling techniques: simple random sampling, systematic sampling, stratified sampling and cluster sampling. We also refer you to Box 8.3.3 for additional tips on sample selection.

In simple random sampling, a sample is chosen randomly from the population, all with the same probability of selection. Systematic sampling draws a sample according to a random starting point but with a fixed, periodic interval. For example, a researcher might be interested in “articles written about Uber by BC media outlets in 2020”, use simple random sampling to select the first case then sample at a fixed interval, say “every 10th article.” Stratified sampling divides the population according to categories (strata) of interest (e.g. certain demographic characteristics) then samples from within each strata. Finally, cluster sampling divides the population into clusters (e.g. “articles by province, cities, municipalities”) and then selects cases from among the clusters.

If you are collecting primary data for your undergraduate project, you are likely using a non-probability technique. Non-probability sampling uses ‘subjective’ criteria sample selection such that not everyone in a population has a known or equal chance of being selected. Examples of non-probability techniques include convenience, quota, snowball, and expert sampling. Convenience samples include all participants who meet the study’s criteria, and are willing and able to participate. Quota sampling divides the unit of analysis into exhaustive, mutually exclusive groups, and then picks a predetermined number of participants/cases from each group. Expert sampling involves the selection of participants who are known to be knowledgeable about the topic. Snowball sampling selects participants through “word of mouth” – through asking participants to help find other individuals who fit the goals of your study. While this will likely limit your sample to the size of your network (and the bias of your network), it will also ensure that the population your studying is one in which you have the most access too (since they are likely to be close to you and motivated to work with you because you share the same social network; this can likewise extend to reflecting biases that form and filter the milieu you occupy).

All of these sampling methods can be used alongside the types of analysis discussed later in the chapter. The following sections seek to parcel out unique data collection techniques and the assumptions which underpin them. Interpretive research, for instance, will often require a less regimented sampling procedure; but the researcher must be aware of why this is the case to still collect enough of the right data to make a meaningful interpretation. We will also unpack three common methods of data collection – sourcing secondary qualitative and quantitative data, survey data (including interviews), and interpretive research.

Box 8.1 – Thinking About Your Sample Design for In-depth Interviews: Some Practical Tips

Extensive in-depth interviews (that take upward of an hour) require more due diligence in sampling. This makes snowball sampling and other methods that rely on your ‘ties’ effective for finding people who are motivated to substantively participate in your research. As per Human Ethics Board stipulations, your research cannot pose a serious risk of harm to subjects, and so your interviews will not likely involve stressful topics (of which you would need more training to undertake) on a vulnerable population you have no connection to. With this in mind, you should think carefully of the people who will form the data of your research. Then, once you have found them, ensure that you outline in as much detail as possible the goals and requirements of your research.

If you are still struggling to find enough participants (which for in-depth interview Honours theses often ranges from 5-10) then consider posting ads on school bulletins. The ethics board at UBC permits the use of small stipends (gifts no more than $10) for low-risk research and allows advertisements to be printed to gain participants. Make sure to make your advertisements early, however, because the ethics board will ask that you submit these ads with the rest of your information.


Bhattacherjee, A. (2012). Social Science Research: Principles, Methods, and Practices https://scholarcommons.usf.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=1002&context=oa_textbooks



Icon for the Creative Commons Attribution-NonCommercial 4.0 International License

Practicing and Presenting Social Research Copyright © 2022 by Oral Robinson and Alexander Wilson is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, except where otherwise noted.

Share This Book