Robust Inference Via Heteroskedasticity in Linear Models

IMF Working Papers, May 22, 2026

Source details

Study focus: inference via heteroskedasticity in linear models commonly used for macroeconomic policy analysis where covariate endogeneity must often be addressed with limited time and data.
Framework features:
- Nests standard heteroskedasticity-based approaches.
- Allows for new non-nested restrictions.
- Does not require ex-ante regime labelling.
Primary contribution: proposal of an easily implementable weak-identification-robust test with derived sufficient conditions for validity.

Approach: heteroskedasticity-based inference in linear models that accommodates endogeneity and limited data settings.
Novelty:
- Incorporates non-nested restrictions alongside standard methods.
- Eliminates the need for prior regime labelling.
Output: a weak-identification-robust test designed to be easily implementable.
Theoretical results: sufficient conditions for the validity of the proposed test are derived.

Findings: simulation results show good size and power properties in a wide range of settings.
Implication: the proposed test performs well under diverse simulated scenarios relevant to macroeconomic applications.

Demonstrated versatility and scalability through applications to:
- Fuel-price passthrough in Sierra Leone.
- Effect of remittances on consumption in the Philippines.
- Exchange-rate passthroughs in many countries.

Subject: Balance of payments, Consumption, Exchange rates, Foreign exchange, Fuel prices, Inflation, National accounts, Prices, Remittances
Keywords: Anderson-Rubin, Consumption, Exchange rates, Fuel prices, Global, Heteroskedasticity, IMF working papers, Inflation, regime labelling, Remittances, simulation result, two-step inference, views of the IMF, weak identification