AI Meets Fiscal Policy: Mapping Government Spending Actions Across 64 Countries
IMF Working Papers, March 6, 2026
Source details
- Canonical URL
- AI Meets Fiscal Policy: Mapping Government Spending Actions Across 64 Countries
Other formats
Bibliographic details
- Authors: Shuvam Das, Davide Furceri, Nikhil Patel, Adrian Peralta-Alva
- Published: March 6, 2026
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9798229038799.001
Methodology and Data Construction
- Built the first global quarterly narrative database of discretionary government spending actions by applying a fixed GPT–4.1 prompt to Economist Intelligence Unit (EIU) Country Reports.
- Identifies exogenous spending shocks—expansions and contractions—for an unbalanced panel of 64 countries from 1952:Q1 to 2023:Q4.
- Uses country-by-country VARs that treat the narrative indicator as an internal instrument to derive comparable cumulative government spending multipliers.
Validation and Robustness Checks
- Database validated by:
- Replicating expert narrative coding in Romer and Romer (2019).
- Showing that the identified shocks predict subsequent movements in measured government spending.
- Establishing alignment with action-based consolidation measures in Adler et al. (2024).
Main Findings: Fiscal Multipliers
- Derived the first set of comparable cumulative government spending multipliers across countries.
- Median multiplier is 0.7 at horizons up to two years.
- Substantial heterogeneity in multipliers across countries and over time.
Heterogeneity and Conditional Patterns
- Pooled estimates imply:
- Larger multipliers in less open economies.
- Larger multipliers under fixed exchange-rate regimes.
- Larger multipliers in downturns.
- Multipliers are:
- Smaller when uncertainty is high.
- Larger when political support is stronger.
Content in this bundle
- Working Paper