Causal Graphs: Addressing the Confounding Problem Without Instruments or Ignorability

TL;DR

This paper discusses causal graphs as a method to address confounding in causal inference without relying on instruments or ignorability assumptions, providing a new perspective on causal identification.

Contribution

It introduces causal graphs as an alternative approach to handle confounding, bypassing the need for instrumental variables or ignorability assumptions.

Findings

Causal graphs can identify causal effects without instruments.

The approach offers a visual and conceptual framework for confounding control.

It broadens the toolkit for causal inference in observational studies.

Abstract

Discussion of "Instrumental Variables: An Econometrician's Perspective" by Guido W. Imbens [arXiv:1410.0163].