New VAR model improves real-time fiscal monitoring for Italy

The paper introduces a Bayesian Mixed-Frequency Vector Autoregression (VAR) model designed for real-time fiscal monitoring, specifically to nowcast the government deficit-to-GDP ratio.

This model uniquely integrates both monthly cash and quarterly accrual fiscal indicators, alongside other high-frequency macroeconomic and financial variables, real GDP, and the GDP deflator.

Applied to Italy, the model generates timely monthly density nowcasts of the annual deficit ratio.

Crucially, these nowcasts are shown to be as accurate, and often more accurate, than those produced by the European Commission, a significant finding given the model's parsimonious statistical basis compared to the Commission's extensive multi-model approach incorporating expert judgment and nearly two hundred variables.

Nowcasting the current-year deficit ratio is complex, involving predictions for expenditures, revenues, and nominal GDP, compounded by significant delays and revisions in quarterly data.

This model addresses these by leveraging timely monthly cash data, filtering noise from differing accounting procedures.

Focusing on Italy, a relevant case due to its high debt-to-GDP ratio, the model also provides an economic narrative for changes in predictions and facilitates robust scenario analysis.

For example, it compares deficit dynamics under a monetary policy tightening versus a typical recession, revealing a more muted fiscal response in the latter due to stabilizing monetary policy.