Identification and estimation of causal peer effects using double negative controls for unmeasured network...

Area of Science:

• Social Sciences
• Econometrics
• Network Analysis

Background:

• Causal peer effects are crucial for understanding social influence but are challenging to estimate in observational data.
• Unmeasured network confounding, including homophily and contextual effects, hinders accurate identification.
• Network dependence of observations further complicates the estimation of peer effects.

Purpose of the Study:

• To develop a framework for non-parametrically identifying causal peer effects in the presence of unmeasured network confounding.
• To propose a robust statistical estimator for causal peer effects.
• To provide methods for assessing the reliability of the estimated causal peer effects.

Main Methods:

• Leveraging a pair of negative control outcome and exposure variables (double negative controls) to address unmeasured confounding.
• Developing a generalized method of moments (GMM) estimator for causal peer effects.
• Establishing consistency and asymptotic normality of the GMM estimator under specific network dependence assumptions.

Main Results:

• The proposed double negative control framework successfully identifies causal peer effects non-parametrically.
• The generalized method of moments estimator demonstrates consistency and asymptotic normality under stated assumptions.
• A consistent variance estimator is provided for statistical inference.

Conclusions:

• The double negative control approach offers a powerful tool for estimating causal peer effects in complex observational network data.
• The proposed GMM estimator provides a statistically sound method for quantifying peer influence.
• This framework advances the ability to study social influence and network dynamics accurately.