How AI Is Transforming Fiscal Policy Research Across 64 Countries, IMF Study Finds

An IMF study uses artificial intelligence to analyze decades of economic reports and build a global dataset of government spending actions across 64 countries. The research finds that government spending generally boosts economic output, but the strength of its impact varies depending on economic structure, uncertainty, and political conditions.

Artificial intelligence is beginning to reshape how economists study public policy. A new study by researchers from the International Monetary Fund’s Fiscal Affairs Department and Research Department shows how AI can help analyze decades of government spending decisions across the world. In the study “AI Meets Fiscal Policy: Mapping Government Spending Actions Across 64 Countries,” IMF economists Shuvam Das, Davide Furceri, Nikhil Patel and Adrian Peralta-Alva use artificial intelligence to build one of the most comprehensive datasets on fiscal policy ever assembled. Their work offers fresh insights into how government spending affects economic growth and why the results vary across countries.

A New Way to Study Fiscal Policy

Economists have long debated the extent to which government spending affects economic activity. This relationship is known as the fiscal multiplier, the extent to which economic output increases when government spending increases. The issue is important for policymakers because governments often rely on spending programs during economic crises to support growth and employment.

One major challenge in studying fiscal policy is separating cause from effect. Governments usually increase spending when the economy slows and cut spending when finances are tight. Because of this, it can be difficult to determine whether government spending itself drives economic growth or simply responds to economic conditions.

Traditionally, researchers have solved this problem by manually studying historical documents to identify spending decisions that were made for reasons unrelated to the current economic situation. This method, known as narrative analysis, is slow and requires reading thousands of reports and policy documents. As a result, earlier research usually focused on only a few countries.

AI Reads Thousands of Economic Reports

The IMF study uses artificial intelligence to dramatically expand this approach. The researchers applied a large language model to thousands of Economist Intelligence Unit country reports, which describe political and economic developments in many countries over long periods of time.

The AI system reads these reports and identifies references to government spending decisions. It then classifies each decision based on the reasons given in the reports. Spending actions linked to long-term goals, such as infrastructure investment or structural reforms, are treated as independent policy choices. Measures taken to respond directly to economic conditions, such as fighting recession or inflation, are excluded because they are not considered independent policy shocks.

Using this method, the researchers created a quarterly record of discretionary government spending actions across 64 countries from 1952 to 2023. The dataset includes more than sixteen thousand country-quarter observations and identifies over eight thousand clear spending actions.

What the Data Reveals

The dataset shows that expansionary and contractionary fiscal actions occur with similar frequency overall. However, patterns differ across country groups. Advanced economies record more spending cuts, reflecting fiscal consolidation policies and budget rules. Low-income countries show more expansionary actions, often linked to infrastructure and development spending.

The researchers then used this dataset to estimate fiscal multipliers for each country. Their results suggest that government spending typically increases economic output, but not on a one-to-one basis. On average, the multiplier is about 0.7 over one to two years. This means that a one-unit increase in government spending raises economic output by roughly 0.7 units.

However, the strength of this effect varies widely across countries. Some economies experience strong growth effects from government spending, while others show much smaller responses.

Why Fiscal Policy Works Better in Some Economies

The study also identifies several factors that influence how effective government spending is. Countries that are less open to international trade tend to have larger fiscal multipliers. In highly open economies, a portion of government spending leaks abroad through imports, reducing the impact on domestic production.

Exchange rate regimes also matter. Fiscal policy tends to have stronger effects in countries with fixed exchange rates than in those with floating currencies. When exchange rates are fixed, monetary policy has less ability to offset government spending changes, allowing fiscal policy to play a larger role in driving economic activity.

Other structural characteristics also shape fiscal outcomes. Economies with large informal sectors tend to have smaller fiscal multipliers, while countries with more flexible labor markets often experience slightly stronger responses to government spending.

Economic Conditions and Political Support Matter

The effectiveness of government spending also depends on the broader economic and political environment. The study finds that fiscal stimulus tends to be more powerful when economic growth is weak. During slowdowns, government spending can help support demand and boost output more effectively than during economic booms.

Uncertainty also plays an important role. When businesses and households face high economic or policy uncertainty, they may delay investment and consumption decisions. In such situations, the impact of government spending becomes weaker.

Political factors matter as well. Fiscal measures implemented when governments have strong political support are more likely to translate into real spending and stronger economic effects. Spending announcements made near elections, by contrast, sometimes fail to produce meaningful increases in actual expenditure.

Overall, the study demonstrates how artificial intelligence can open new possibilities for economic research. By allowing computers to analyze thousands of policy reports, economists can now study fiscal policy across many countries and decades of history. The findings offer a clearer picture of how government spending affects economic growth and show that its impact depends heavily on the structure of the economy, economic conditions and political context.