IMF Working Paper AI Meets Fiscal Policy: Mapping Government Spending Actions Across 64 Countries introduces the first global quarterly narrative database of discretionary government spending actions, utilising a fixed GPT-4.1 model to analyse fiscal shocks across 64 countries from 1952 to 2023.
By applying AI-assisted narrative methods to Economist Intelligence Unit (EIU) reports, the study distinguishes between "endogenous" actions (reactive to current economic conditions) and "exogenous" shocks (driven by long-term objectives or inherited imbalances).
The findings reveal a median global government spending multiplier of approximately 0.7 at a two-year horizon, though this varies significantly based on trade openness, exchange-rate regimes, and political stability. Crucially, the research identifies that high political support amplifies fiscal multipliers, while election proximity and economic uncertainty weaken the pass-through effect of government spending.
Key Pillars of the AI-Fiscal Mapping Framework
GPT-4.1 Narrative Classification: Using a standardized, non-fine-tuned AI prompt to classify fiscal actions based on stated motivations in historical country reports.
Exogenous Shock Identification: Isolating spending changes that are unrelated to contemporaneous macroeconomic cycles to measure their "pure" impact on output.
Structural Heterogeneity Analysis: Categorizing multipliers based on country-specific characteristics such as labor market flexibility, informality, and public debt levels.
State-Dependent Multipliers: Identifying that fiscal transmission is stronger during economic downturns and when monetary policy is constrained at the zero lower bound.
Political Environment Indexing: Formulating indicators for majority cohesion and policy certainty to determine how the political landscape influences fiscal success.
Long-Horizon Impact Tracking: Measuring the evolution of multipliers at horizons up to two years to provide a high-fidelity view of policy lag.
What is the "Fiscal Multiplier"? A fiscal multiplier is a ratio that measures the change in national income (GDP) resulting from a change in government spending. A multiplier of 0.7 implies that for every $1 the government spends, the economy grows by $0.70. This mechanical relationship is a functional prerequisite for budget planning; however, as the IMF paper demonstrates, this ratio is not fixed. It acts as a primary mechanic that is "dampened" in open economies where spending leaks into imports, or "amplified" in environments with strong political implementation and low fiscal uncertainty.
Deep Dive: India’s Fiscal Multiplier Analysis
India is classified as an emerging market economy in the database, with quarterly macroeconomic and fiscal series data starting from 1997 Q2.
Emerging Market Multipliers: The document highlights that Emerging Market and Developing Economies (EMDEs), including India, exhibit slightly larger median fiscal multipliers than advanced economies.
Output Effects: For the EMDE group, the median multiplier is estimated at 0.80 at one year and 0.81 at two years.
Dispersion Factors: While India benefits from these higher median effects, the study notes a wider dispersion in output outcomes across EMDEs compared to advanced nations, likely due to varying levels of informality and labor market flexibility.
Political Implementation: Consistent with the global findings, the impact of India’s fiscal actions is mechanically linked to the level of political support and the proximity of elections, which can weaken the actual implementation of sanctioned spending.
Policy Relevance: India’s Macro-Fiscal Strategy
Operationalizing Counter-Cyclical Spending: The finding that multipliers are higher during downturns acts as a primary mechanic for the Ministry of Finance to justify increased infrastructure spending during periods of low private demand.
Internalizing Exchange-Rate Regimes: As a country with a managed exchange rate, India can leverage the paper’s insight that multipliers are larger under less flexible regimes compared to pure floats.
Bypassing Election-Year Lags: The evidence that election proximity weakens spending pass-through provides a functional shield for NITI Aayog to advocate for the front-loading of critical projects early in the political cycle.
Link to AI-Driven Research: Utilizing GPT-4.1 for fiscal mapping serves as a prerequisite for Indian economic researchers to adopt transparent, replicable AI tools for domestic policy evaluation.
Follow the Full Release Here: AI Meets Fiscal Policy: Mapping Government Spending Actions Across 64 Countries
