An introduction to instrumental variable assumptions, validation and estimation

Area of Science:

• Epidemiology
• Econometrics

Background:

• The instrumental variable (IV) method is used in economics to establish causality when unmeasured confounding is present.
• Drawing parallels between IV assumptions and randomized controlled trials (RCTs) can enhance understanding in epidemiology.

Purpose of the Study:

• To elucidate the assumptions of the instrumental variable method.
• To highlight the importance of the fourth identifying assumption for generalizability.
• To promote a deeper understanding of IV methods in epidemiological research.

Main Methods:

• The study explains the concept of an instrumental variable: a predictor of exposure with no independent effect on the outcome.
• It draws analogies between ideal instruments (like random assignment in RCTs) and naturally occurring instruments in observational studies.
• The text emphasizes the role of instruments in defining a subpopulation of 'compliers' with pseudo-randomized exposure.

Main Results:

• The instrumental variable method relies on four key assumptions, which can only be partially validated empirically.
• The fourth assumption, concerning the generalizability of effects, is crucial but often overlooked.
• Instruments facilitate exchangeability regarding unmeasured confounders within the complier subpopulation.

Conclusions:

• Epidemiological studies using instrumental variables should rigorously validate all four assumptions.
• Leveraging parallels to randomization can improve the application and interpretation of IV methods.
• Subject-matter expertise is essential for validating the underlying assumptions of instrumental variable analyses.