Networks or neighborhoods? Correlations in the use of publicly-funded maternity care in California

This study focuses on "network effects" in the utilization of publicly-funded prenatal care using Vital Statistics data from California for 1989-2000. Networks are defined using 5-digit zip codes and a woman's racial or ethnic group. Like others, we find evidence that the use of public programs is highly correlated within groups defined using race/ethnicity and neighborhoods. These correlations persist even when we control for many unobserved characteristics by including zip code-year fixed effects, and when we focus on the interaction between own group behavior and measures of the potential for contacts with other members of the group ("contact availability"). However, the richness of our data allows us to go further and to conduct several tests of one important hypothesis about networks: that the estimated effects represent information sharing within groups. The results cast doubt on the idea that the observed correlations can be interpreted as evidence of information sharing. In particular, we find estimated effects to be as large or larger among women who have previously used the program as among first-time users. (C) 2003 Elsevier B.V. All rights reserved.