Models and Marginals: Using Survey Evidence to Study Social Networks

Abstract

Fischer (2009) argues that our estimates of confidant network size in the 2004 General Social Survey (GSS), and therefore the trend in confidant network size from 1985 to 2004, are implausible because they are (1) inconsistent with other data and (2) contain internal anomalies that call the data into question. In this note, we assess the evidence for a decrease in confidant network size from 1985 to 2004 in the GSS data. We conclude that any plausible modeling of the data shows a decided trend downward in confidant network size from 1985 to 2004. The features that Fischer calls anomalies are exactly the characteristics described by our models (Table 5) in the original article.