Abstract

For researchers interested in the study of personal networks, measures of network composition are often obtained through the use of name generators and name interpreters. However, the cost of administering a survey with multiple name generators, in terms of time and respondent motivation, is often prohibitive. Researchers seeking to minimize respondent burden routinely turn to time saving measures, such as the use of a single name generator (i.e. the “important matters ” generator used in the General Social Survey (GSS)). We argue that the limitations of this approach are often understated. In the study of social support, multiple name generators are required to ensure that researchers sample from the full definition of support. Putting aside issues of construct validity, we compared measures of network composition and structure obtained from stand alone generators to measures obtained from a six-item multiple name generator. We found that although some single generators provided passable estimates for some measures, all single generators failed to provide reliable estimates across a broad spectrum of network measures, including key variables such as size and density. In an attempt to improve the reliability of network measures, beyond what could be obtained through single generator alternatives, and while

Citations