Causal Relationships between Personal Networks and Health: A Comparison of Three Modeling Strategies

Area of Science:

• Social network analysis
• Health outcomes research
• Causal inference methodologies

Background:

• Previous studies show correlations between personal network characteristics and health.
• Establishing a causal link between social networks and health remains a significant research challenge.

Purpose of the Study:

• To evaluate different causal inference approaches for egocentric network data.
• To investigate the relationship between network variables and global health outcomes using longitudinal data.

Main Methods:

• Utilized three waves of data from the University of California Berkeley Social Networks Study (N = 1,159).
• Compared three statistical modeling strategies: cross-sectional ordinary least squares (OLS) regression, regression with lagged dependent variables (LDVs), and hybrid fixed and random effects models.
• Examined nine network variables and two global health outcomes.

Main Results:

• Cross-sectional and LDV models may inflate the perceived causal effects of social networks on health.
• Hybrid models indicate that network-health associations are mainly between individuals.
• Network changes do not appear to cause significant within-individual changes in health.

Conclusions:

• The findings highlight the utility of panel data for advancing network and health scholarship.
• Causal effects of social support within networks on health may be less substantial than previously assumed.
• Recommends advanced modeling techniques for more accurate causal inference in network research.