Stratified Sampling

Stratified sampling is a process that first divides the overall population into separate subgroups and then creates a sample by drawing subsamples from each of those subgroups. Within the overall process of sampling, stratification is related to the definition of the population because it requires a prior definition of categories within the population before it is possible to draw samples from those subgroups. This general process can apply to both qualitative and quantitative research. For survey sampling, stratification ensures the degree to which preselected subgroups in the population are represented in the sample—otherwise, a process of random sampling always includes the possibility that one of these groups will be substantially over or underrepresented, simply by the luck of the draw. For example, a survey of a city where the population is evenly split between three major ethnic groups might divide the sample into thirds, with each subsample drawn from a different ethnic group to ensure that the size of each group in the sample reflects its size in the population.

In qualitative research, stratified sampling is a specific strategy for implementing the broader goal of purposive sampling. In this case, dividing the larger population into subcategories that are relevant for the research goals ensures that the data will include cases from each of these categories. The simplest kind of stratification divides the overall sample into two distinct groups, but it is also possible to create more than two groups or to draw the subsamples from different points along a continuum.

The most common reason for using a stratified approach to purposive sampling is to do systematic comparisons between the categories that define the basis for stratification. For example, an interview project might examine how parents from the lower, middle, and upper socioeconomic sectors interacted with their children's teachers, while a participant observation project might compare schools where students from low-income families had either above- or below-average performance on standardized tests, or a media analysis project might examine written work from students whose teachers had either less than 2 or more than 5 years of experience. In each of these cases, the overall goal of purposive sampling includes the need to determine the similarities and/or differences between carefully selected subsets of the larger population, and the stratification of the sample makes this comparison possible.

For qualitative research, stratification has a distinct link to quota sampling as a means of selecting cases. Thus, when the purposive selection process calls for data from subgroups in the population, the next step is to select the members of those subgroups that will make up the corresponding subsamples, which amounts to setting a quota for the size of each sub-sample. For example, if the research design calls for a total of 20 men and women to be interviewed on some topic, then the research design implies a quota of more or less 10 men and 10 women in each subgroup. Of course, the emergent data from those interviews may point toward something other than an even split between the original categories, but the key goal is still to ensure that there are a sufficient number of data sources in each subcategory.

...