In this presentation, Kaustubh addresses the imperative to enhance nowcasting models for Indian GDP in the post-pandemic era. He brings together a diverse set of high-frequency economic indicators (HFIs) into multiple factors—nominal, survey-based, labor-market, and real economic activity—to produce GDP growth nowcasts. Unlike existing models that primarily focus on aggregate estimates, he evaluates each HFI’s contribution by analyzing its impact on nowcast revisions following new data releases.

In addition, the COVID-19 pandemic introduced outliers in HFIs, distorting model parameters and reducing forecasting accuracy. To address this, Kaustubh discusses incorporating the Oxford Stringency Index and proposes a novel sigmoid-based data transformation that minimizes model sensitivity to large shocks. This approach enables models to handle unexpected events more robustly without overreacting. This methodology improves nowcasting models’ resilience to outliers, providing valuable insights during volatile periods.

In this presentation, Juan Rubio-Ramírez introduces a novel inference algorithm for structural vector autoregressions (SVARs) identified by sign restrictions. The key aspect of the algorithm is to move beyond the traditional accept–reject tradition associated with sign-identified SVARs. He demonstrates how embedding an elliptical slice sampling within a Gibbs sampler approach can deliver dramatic gains in speed and turn previously infeasible applications into feasible ones. A tractable example will illustrate the power of elliptical slice sampling for sign-identified SVARs. Finally, Juan will demonstrate the algorithm’s utility by applying it to a well-known small-SVAR model of the oil market featuring a tight identified set, and to a large-SVAR model with more than 100 sign restrictions.

Analysts and researchers increasingly face demands to accelerate the estimation of large-scale macroeconomic models such as DSGE and BVARs, and to efficiently manage multiple model runs. This session addresses these challenges by exploring strategies for using parallel computing to reduce estimation times and streamline workflows for a range of models, including those based on Dynare. Key topics include the application of parallel computing in optimization and estimation, orchestrating simultaneous runs, and systematically tracking the inputs and outputs of each run—enabling users to revisit and reproduce past experiments as needed. Attendees will gain practical insights and step-by-step guidance for implementing these solutions.