HSE Centre for Big Data in Economics and Finance research seminar: "Exogeneity in Structural Vector Autoregressions and Local Projections with External Instruments", L. Fanelli
HSE Centre for Big Data in Economics and Finance research seminar: "Exogeneity in Structural Vector Autoregressions and Local Projections with External Instruments", L. Fanelli, 5 October 2026, 12 noon (Moscow time)
HSE Centre for Big Data in Economics and Finance invites you to the regular research seminar on October 5, 2026, at 12:00 (Moscow time) via Zoom.
Speaker: Luca Fanelli (University of Bologna)
Topic: “Exogeneity in Structural Vector Autoregressions and Local Projections with External Instruments”
Joint work with: Giovanni Angelini (University of Bologna) and Giuseppe Cavaliere (University of Bologna and Exeter Business School).
Contact info: luca.fanelli@unibo.it
Proxy-SVARs and LP-IVs typically treat instrument exogeneity as a maintained assumption. This paper provides an alternative identification strategy under which proxy contamination becomes estimable and exogeneity becomes testable with standard methods regardless of instrument strength. The authors embed policy variables, outcomes, and external proxies in a joint simultaneous-equations system, called the proxy-SEM. The model is identified without imposing proxy exogeneity by exploiting auxiliary identification information provided by the system's dynamics.
Under dynamic exclusion restrictions, lagged policy variables serve as internal instruments for contemporaneous policy variables, with relevance governed by a standard rank condition. Conditional on these auxiliary identifying restrictions being credible, the proxy-SEM recovers the covariance between the proxies and a nonsingular transformation of the non-target shocks. This information is used to construct decontaminated instruments and to test proxy exogeneity. The decontaminated instruments remain correlated with the target shocks while being orthogonal to non-target shocks.
The approach allows for asymptotic inference robust to both weak and contaminated instruments. The exogeneity test uses standard asymptotic inference under strong or weak external instruments, including just-identified settings. The test can also account for uncertainty about which member of a stated family of auxiliary identifying assumptions is correct. Oil-market and fiscal applications illustrate the empirical relevance of the approach.
The HSE Centre for Big Data in Economics and Finance, established in 2025 at the Faculty of Economic Sciences, conducts fundamental and applied research in big data analysis, develops modern econometric and machine learning methods, and implements educational programs training next-generation specialists. Learn more about the Centre's activities on its official webpage.
To participate, please register on the Centre's seminar page where the Zoom link will be provided upon confirmation.

