Quota Sampling

The target population in a sampling survey can often be divided into several subpopulations based on certain characteristics of the population units. Some common variables used to categorize population include gender, age, location, education level, and so forth. It is also common in a sampling survey situation that investigators might require certain propositions or numbers for each type of sampling unit included in the sample for various reasons. For instance, a representative sample is often considered one of the most important criteria when evaluating the quality of a sampling survey, and a sample is considered to be representative if the sample structure is the same or close to the population structure; if one is to conduct a survey where gender is considered an influential factor of the response, the investigator needs to draw a sample where the proportion of each gender is as close to the target population as possible. Nevertheless, there are also cases when investigators’ subjective judgments play an important role, for example, in a marketing survey of the acceptance level of a new perfume product, the investigator may likely consider a sample with a majority of female interviewees over males appropriate. Bear in mind, other motives or objectives are also possible.

All that being said, in reality, it has never been an easy task to randomly select a sample where the number of units belonging to each subpopulation are fixed; yet many sampling designs developed in the past were targeted toward that end. Several of these designs include stratified sampling designs, proportional allocation of sample sizes, proportional sampling, and quota sampling. Among all, quota sampling is one of the most commonly used sampling selection methods in opinion and marketing survey. The sampling method could be viewed as a different class of sampling designs that shares similar sampling principles. The practical popularity portioned to quota sampling is due mainly to its ability to provide samples with the desired numbers or proportions of each subpopulation at a limited sampling cost. Such an advantage is often compromised by the lack of legitimate statistical inference, as often criticized by statisticians. In the following sections, the general sampling principle of quota sampling and the usual methods used to construct the inference for the population quantity of interest, along with a short review of the advantages and disadvantages, are discussed. A possible manner to reach a balanced solution between sampling convenience and proper inference is also introduced.