Social Isolation in America: An Artifact

It is generally believed that social isolation has been increasing dramatically in the U.S. over the past few decades. This is largely based on the 2004 General Social Survey (GSS), which indicated that the average American’s discussion network (the number of individuals a person discusses important matters with) had shrunk considerably. The findings were unexpected and inspired strong debate among researchers. Some even suggested the findings might be the result of an unknown technical error. To date, no one has been able to fully account for the seemingly drastic rise in isolation found in the GSS.

This study is the first to examine whether the increase in social isolation found in the GSS can be traced to the interviewers conducting the surveys (i.e., to ‘interviewer effects’). Researchers hypothesized that because the question asking participants to list people in their network required interviewers to ask a series of follow-up questions for every individual named, the latter may consciously or unconsciously have discouraged the former from providing lengthy lists. Statistical analyses revealed that indeed a few ‘outlier’ interviewers were responsible for a very high number of reports of social isolation. Eliminating these interviewers’ results reduced the finding of isolates by approximately 50%.

Interviewer effects appear to be a major issue associated with survey questions that ask participants to generate lists of names. Estimates of social connectivity and social isolation that rely on such questions may suffer from substantial bias related to interviewer measurement error. This raises the possibility that inferences made based on such estimates of social connectivity may be mistaken.