Causal Identification in Crime Research: Lessons From Structural Causal Models

Area of Science:

• Social Sciences
• Criminology
• Econometrics

Background:

• Observational data in social sciences often presents identification challenges.
• Causal inference requires explicit causal assumptions for valid conclusions.
• Existing methods may not adequately address potential biases like collider bias.

Purpose of the Study:

• To demonstrate how structural causal models (SCMs) and directed acyclic graphs (DAGs) can overcome identification challenges in social sciences.
• To highlight the importance of distinguishing identification from estimation.
• To improve the credibility of causal claims in applied research, using crime and criminal justice as case studies.

Main Methods:

• Utilizing structural causal models (SCMs) to represent causal relationships.
• Employing directed acyclic graphs (DAGs) to visualize and analyze causal structures.
• Applying these methods to case studies within crime and criminal justice research.

Main Results:

• SCMs and DAGs make assumed causal structures explicit.
• These graphical models aid in diagnosing potential pitfalls such as collider bias and inadmissible instruments.
• Distinguishing identification from estimation sharpens causal inference strategies.

Conclusions:

• Explicitly specifying causal models with SCMs and DAGs is crucial for valid inference.
• These methods enhance the rigor and credibility of causal claims derived from observational data.
• The approach offers practical benefits for researchers in fields like crime and criminal justice.