Assesing the use of Structural VAR models for policy recommendations during periods of high volatiltiy

Resumen

Vector autoregressive models are widely used in central banks around the world to assist in producing recommendations to economic policy makers. Some authors argue that whether their use with this purpose is susceptible to the Lucas Critique is a practical matter and depends on the magnitude and persistence of the intervention. This fact is particularly relevant during economic crises, given that the magnitude of the interventions needed to achieve the policy-maker’s goals might render the recommendations of these models meaningless.
In this context, a flexible framework is proposed to formalize the process of designing policy interventions with Structural Vector Autoregressions as a constrained optimization problem. In it, the modeler should set a loss function that reflects aspects of the model that she wants to be preserved during the intervention, while the restrictions reflect objectives of the intervention and other types of considerations on the future evolution of the variables. It is additionally suggested that the adequacy of the designed intervention scenarios be subsequently evaluated by fitting ARMA models to the associated structural shock sequence, and testing for structural breaks with these auxiliary models.
This methodology is illustrated with a simple monetary policy model estimated with data from Guatemala, with which interventions are designed for the period of the COVID-19 pandemic, and its adequacy is then evaluated. It is concluded that, about half of the time, the designed policy intervention may produce biased estimations in its effects, so one should be extra careful when extracting policy recommendations from this class of models during periods of high volatility.

La presentación de este documento se puede ver en este enlace.