Causal Inference With Instrumental Variables (Part 1)

Causal Inference With Instrumental Variables (Part 1)

by Dr. Hyunseung Kang

Formal & Physical Sciences Mathematics PBMathematicsPBTProbability and statistics
🎙 Dr. Hyunseung Kang 👥 8K 📅 January 23, 2026 ⏱ 60 min 👁 319 📄 lecture 🧭 2026-08-15
Available in: English (current) Français

Keywords

causal inferenceinstrumental variablesmonotonicityrandomized encouragementidentification

Summary

This lecture, part of the Isaac Newton Institute’s ‘Foundations of causal inference’ program, provides a rigorous introduction to instrumental variables (IV) methods. Dr. Hyunseung Kang begins by reviewing strong ignorability assumptions for causal identification, highlighting two common violations: non-compliance in RCTs and unmeasured confounding. He then introduces the concept of an instrument as a variable satisfying three core conditions: relevance, independence, and exclusion. The lecture focuses on the monotonicity-based approach, motivated by the classic randomized encouragement design from Permutt and Hebel (1984) on maternal smoking and birth weight. Kang formalizes the assumptions using counterfactual notation, defining potential treatment and outcome variables. He discusses the importance of the monotonicity assumption and its role in identifying the local average treatment effect (LATE). The lecture is interactive, with questions from the audience on topics such as time-varying treatments, mediation, and falsification of the exclusion restriction. The content is technical and aimed at a graduate-level audience, providing a solid foundation for understanding IV methods.

160 words

Critical Evaluation

Value of the Information & Strength of the Argument

The lecture offers substantial value by clearly explaining the motivation and formal assumptions behind IV methods, using a concrete example (smoking and birth weight) to ground the theory. The argumentation is rigorous, building from strong ignorability to the IV framework, and carefully distinguishes between the monotonicity-based and structural approaches. The speaker emphasizes the importance of study design in defining valid instruments and addresses potential violations of assumptions. The interactive format allows for clarification of complex points, enhancing the pedagogical value.

89 words

Title / Content Match

The title accurately reflects the content, which focuses on the foundations of instrumental variables, specifically the monotonicity-based approach.

Quality & Reliability

8/10

Lecture by a recognized expert in causal inference, based on established literature (Angrist, Imbens, Rubin), with rigorous formal definitions and references to key papers. The content is technically accurate and well-structured, though it remains an introductory lecture.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

  • Angrist, Imbens & Rubin (1996) — Seminal paper on IV with monotonicity, consistent with the lecture's approach.
  • Imbens & Angrist (1994) — Introduces LATE and monotonicity, aligning with the lecture's content.

Dissenting Sources

  • Hernán & Robins (2006) — Alternative structural approach to IV, which the lecture contrasts with the monotonicity-based approach.

Contribution & Novelties

This lecture provides a clear and rigorous introduction to instrumental variables, emphasizing the monotonicity-based approach. It bridges theory and application through the randomized encouragement design, making abstract concepts accessible. The speaker’s interactive style and use of counterfactual notation enhance understanding.

Pour aller plus loin :

109 words

Radar Profile

The radar profile shows high scores in information quantity and quality, reflecting the lecture's comprehensive coverage and technical depth. The technical level is high, suitable for a specialized audience, while the reliability is strong due to the expert speaker and institutional backing.

Reliability 8/10

💬 No comments were provided for analysis.