Measuring the Unmeasurable: Unraveling the complexities of real-time output gap estimation

This paper evaluates seven output gap models for real-time estimates, based on three criteria: stability of estimations on new observations, data revisions and/or methodological changes; inflation forecasting accuracy; and potential output response to structural economic shocks. Results confirm no single model outperforms across all criteria. Structural VARs exhibit superior inflation forecasts but show high instability, while semi-structural models produce more theoretically consistent potential output responses. To overcome this trade-off, we propose a novel clustering approach to pool models based on their real-time performance, yielding improved estimates. Our findings highlight the value of this method for enhancing real-time output gap measurement and informing monetary policy decisions.