Online Program

Defining and Measuring Hard-to-Survey Populations

*Roger Tourangeau, Westat 

Keywords:

In an era of falling response rates for household surveys, it may seem that all populations are hard to survey, but some household populations present special challenges of various sorts that make them harder to survey than the general population. One way to classify the various sources of trouble is by what survey operation they create difficulties for. I distinguish populations that are hard to sample, those whose members who are hard to identify, those that are hard to find or contact, those whose members are hard to persuade to take part, and those whose members are willing to take part but nonetheless hard to interview. These distinctions reflect the major operations in many surveys — sampling, screening, locating, contacting, recruiting, and interviewing. With any given population, problems can arise with each of these operations, making the population hard to survey. And, of course, some hard-to-survey populations present combinations of several kinds of trouble. The talk also examines various metrics for assessing the level of difficulty associated with a given population. Some of these measures reflect difficulties in one stage in the survey process but others attempt to quantify overall levels of difficulty across all of the potential sources of trouble. For example, the U.S. Census Bureau has developed a hard-to-count index based on tract-level differences in census return rates. The talk reviews these and other indexes for measuring how hard it is to survey a specific population.