Shapley Instruments

EI Seminar
Image of campus Woudestein

When multiple valid instrumental variables are available, researchers must decide which instruments to include in their model to achieve optimal finite sample performance. We propose a novel instrument selection method based on the Shapley value. Our approach is examined under two different settings.

Speaker
Giovanni Mellace
Date
Thursday 13 Nov 2025, 12:00 - 13:00
Type
Seminar
Room
ET-14
Location
Campus Woudestein
Add to calendar

In the first, there are many potentially weak instruments, and the goal is to select a subset that meets certain regularity conditions. In the second, we consider a heterogeneous treatment effect model and estimation of the Combined Compliers Local Average Treatment Effect (CC-LATE), where adding instruments entails a trade-off between improving precision and increasing the size of the Combined Compliers share.

See also

Statistical Methods for High-dimensional Dolatility

Mark Podolskij (University of Luxembourg)
Campus Woudestein, showcasing the flags of the School's and Institutes.

The Causal Interpretation of the AKM Estimand

Daniel Wilhelm (LMU Munich)
Campus Woudestein, showcasing the flags of the School's and Institutes.

Testing for endogeneity of irregular sampling schemes

Giulia Livieri (LSE)
Image - Graph Statistics Stock

Causal Inference and Policy Evaluation

Keynote Speaker: Alberto Abadie
Image of campus Woudestein

Adapting smart waste collection systems through operational learning

Dilay Aktas Dejaegere (KU Leuven)

Uneven buffering of U.S. maize yields against extreme heat - a time-varying coefficient panel approach

Marina Friedrich (Vrije Universiteit Amsterdam)
View of a cornfield

Handling and Exploiting Symmetry in Benders Decomposition

Christopher Hojny (Eindhoven University of Technology)
Campus from above with Mandeville Building

FinEML Conference 2026

Financial Econometrics Meets Machine Learning
Image - University of Geneva
More information

Do you want to know more about the event? Contact the secretariat Econometrics at eb-secr@ese.eur.nl.

Compare @count study programme

  • @title

    • Duration: @duration
Compare study programmes