Central Banks Notes

BEAR Toolbox for Estimating Economic Relationships

Conference presentation and 5.2 release

On October 12, 2023, Alistair Dieppe from the European Central Bank (ECB) presented the BEAR Toolbox 5.2 at the MathWorks Finance Conference. The BEAR Toolbox is a robust Bayesian (Panel) VAR toolbox, predicated on MATLAB, designed to cater to forecasting and policy analysis needs. With its user-friendly interface, the toolbox accommodates seasoned analysts and non-technical users.

Significant new features of BEAR 5.2:

  • Overhauled infrastructure that facilitates multiple simultaneous estimations
  • Inputs have been transitioned to objects to simplify adjustments and enhance model tracking
  • Updated interface with examples to help users acclimate to the toolbox
  • Source control integrated through GitHub, which enables community feedback and collaboration

Applications encompass various VAR model estimations and identification types. The toolbox is proficient in estimating various VAR models, including Bayesian VARs, OLS, Panel VARs, stochastic volatility, time-varying parameters, and more. Also, the toolbox offers diverse identification types, such as sign/magnitude restrictions, correlation restrictions, and FEVD restrictions. Forecast evaluation is well-rounded with both classical and Bayesian-specific criteria at disposal.

BEAR Toolbox for Estimating Economic Relationships (Conference Presentation and 5.2 Release)

Upcoming models and improvements (BEAR 6 plans)

  • New models adept at handling large shocks and structural changes, along with the integration of updated approaches
  • Infrastructure with modernized code, a new data management system, an improved interface, and optimizations to increase speed

Other noteworthy topics include advancements in various VAR methodologies

  • Factor augmented VARs within the toolbox facilitate better representation of economic concepts
  • Bayesian Proxy SVARs enable shock identification devoid of theoretical structure imposition
  • Trend-Cycle Bayesian VARs target enhanced forecasting performance
  • Equilibrium VARs aid in trend-cycle decompositions for historical analysis
  • Priors for the long run are developed to discipline long-run predictions using economic theory
  • Comprehensive forecasting evaluation procedures to assess VAR forecasting performance

BEAR Toolbox 5.2 is openly accessible on GitHub, with additional resources available on the ECB website.

See Dieppe’s presentation from the MathWorks Finance Conference.