Where the Rubber Meets the Road

Sociologists use data from experiments, ethnographies, survey interviews, in-depth interviews, archives, and administrative records. Analysts disagree, however, on whether probability sampling is necessary for each method. To address the issue, the author introduces eight dimensions of data collection, places each method within those dimensions, and uses that resource to assess the necessity and feasibility of probability sampling for each method. The author finds that some methods often seen as unique are not, whereas others’ unique natures are confirmed. More surprisingly, some methods for which probability sampling is rare were found to require it, whereas one for which probability sampling is usually believed to be impossible was found to easily use it. Efforts to salvage nonprobability samples and eight additional general justifications for nonprobability sampling are addressed. Advice for individual analysts and counsel for collective responses to improve research are offered.