Sampling

When planning an empirical study (e.g., survey), the issue of sampling is very important. Before you start collecting your data, you need to determine how you will select the people to participate in your study.

Full or total survey vs. sample

The first question to ask is whether you need to draw a sample, or whether you will conduct a full or total survey. In a complete or total survey, you collect data on all members of the population or you already have data on all of them. This is often the case, for example, when you are working with administrative data (e.g., grade lists of all students at a university) or have user data available (e.g., sales figures in an online store). In practice, full or total surveys are usually difficult to implement because they are expensive and time-consuming. Therefore, if you are writing your bachelor's or master's thesis and want to conduct a survey, you will most likely have to define a sample.

Population and sample

As explained above, in a complete or total survey you work with all members of the population. You therefore have data on the entire population. What is the population? The population consists of all elements that are of interest for the research. This can be, for example, all persons about whom a statement is to be made with the help of a survey. A sample is a selection from the entire group of elements, i.e. a selection from the population.

Types of sampling

There are several ways in which you can draw a sample. Thus, a sampling procedure defines the way and the steps you use to select the elements from your population. Three broad groups of sampling procedures can be distinguished:

• Probability sampling
• Purposive or quota sampling
• Convenience sampling

Probability sampling

In probability sampling, each element of the population has a known, non-zero probability of being selected through a random procedure. The probabilities do not have to be equal; equal probabilities are a feature of simple random sampling. For example, you can use a list of all elements of the population and randomly select individuals from it.

An example of this would be a random selection of households from the central population register of a city. Using a computer, you can then randomly select from this register, for example, a sample of 1000 addresses in the city. You then contact these households and ask them (or a selected member of the household) to participate in your survey.

Probability sampling is often difficult to implement in practice, however, because in many cases there is no list of the population, or the selection procedure is too elaborate for smaller empirical studies.

You can conduct probability sampling in a single-stage or multistage way. In single-stage sampling, you select the elements in one step. Multistage sampling makes the selection in several steps. For example, in the first stage you randomly select 50 municipalities in a state, and in the second stage you randomly select 50 addresses from each of these municipalities.

Purposive or quota sampling

Purposive sampling selects participants according to criteria relevant to the study. Quota sampling is a related non-probability method that aims to reproduce the population distribution of selected characteristics, such as age and gender. Quota characteristics can include gender, age, educational attainment, place of residence, position in a company, and length of employment.

For example, if you are doing a survey in retail and you see that in your population there are 40% young women, 30% old women, 20% young men and 10% old men, you try to achieve this distribution in your sample as well.

Quota sampling is especially widespread in the field of market and opinion research and is also often implemented in the context of bachelor or master theses. This form of sampling is less time-consuming and less costly, so it is practical for smaller empirical studies. However, an important assumption is that you have information about your population and know how certain characteristics (e.g. age, gender, etc.) are distributed there. Matching these characteristics does not guarantee that the sample is representative in other respects, because participants within each quota are not necessarily selected at random.

Convenience sampling

The third group is convenience sampling. Here, participants are selected mainly because they are easy to reach and willing to participate. This method is often used in psychology experiments, for example when volunteers are recruited from an available group. Because the participants are not selected randomly, they may differ systematically from the population of interest.

Sample selection in online surveys

With online surveys, it is mostly more difficult to determine the sample selection in advance. In most cases, there is no list of the population from which a selection can be made. One possibility here is to repeatedly take a look at the already completed questionnaires during the course of the survey and check the distributions of quota characteristics. If, for example, you notice that older women are underrepresented, you can actively contact more people from this target group. The goal is to get as close as possible to the quota plan. However, matching the quotas does not by itself turn the sample into a probability sample.

General note

No matter which sampling method you choose, it is very important that you explain your approach clearly in your bachelor's or master's thesis. It should be clear to the reader of your thesis what the population and sample of your study are and how you have selected them.

In your paper you should answer the following questions:

• What were the target population and the final sample?
• Why were these people selected?
• How did you contact the respondents?