## wpiea2024209-print-pdf

## Source details

**Canonical URL:** [wpiea2024209-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024209-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024209-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024209-print-pdf.pdf.json)

---

### News-Based Fiscal Policy Uncertainty Index — Construction
- Data source and coverage:
  - News articles from Dow Jones Factiva, covering more than 30,000 news sources.
  - Sample restricted to printed articles published by 43 major English-language news outlets in the US, Canada, and the UK.
  - Dataset spans from January 1977 to July 2024, encompassing over 47 million articles.
  - Inclusion of international news outlets (e.g., “ABC News: World News Saturday”, International BBC network, Reuters) to improve coverage for small emerging markets and low-income countries.
- Article selection logic:
  - Keywords built around three layers corresponding to “fiscal”, “policy” and “uncertainty”:
    - One fiscal topic term in tax, government expenditure, public debt, or others.
    - Policy authority identifier (e.g., “congress”, “parliament”, “minister”).
    - An uncertainty identifier (e.g., “uncertainty”, “ambiguous”, “unpredictable”).
  - Country-level FPU: overlay above conditions with a country identifier and restrict to articles where a country name appears in the title, snippet, or main paragraph identified by Factiva.
  - Fiscal perimeter defined by the government budget constraint:
    - B_{t+1} = R_t B_t + G_{t+1} − T_{t+1}
      - where B_t denotes end-of-period total outstanding public debt, R_t denotes gross interest rate on public debt, G_t denotes government spending and T_t denotes tax revenues.
  - Exact queries for country-specific FPU are reported in Appendix D (per source).
- Search keyword categories (truncation * denotes any possible ending):
  - Category I. Policy authority: government or parliament or congress* or regulat* or policy or policies or legislat* or minister or ministry or national assembly
  - Category II. Uncertainty: uncertain* or ambiguous or dubious or precarious or unpredictable or undecided or undertermined or unresolved or unsettled or risk or doubt or doubtful or volatilit*
  - Category III. Fiscal Topics:
    - Tax: taxation or taxes or taxed or tax
    - Government expenditure: government spending or federal spending or public spending or government expenditure or federal expenditure or public expenditure or defense spending or military spending or pension reform* or pension expenditure or healthcare expenditure or medical care expenditure or social securities expenditure or social safety net* or public investment* or social protection or subsidies or subsidy or entitlement spending or social security
    - Public debt: federal debt or government debt or national debt or public debt or debt ceiling or sovereign debt or debt burden or debt repay* or debt sustainability or government borrowing or sovereign borrowing or debt consolidation or sovereign risk or sovereign default or debt default or debt restructure or foreign debt or external debt or government bond or sovereign bond or sovereign yield
    - Others: public sector or public finance or sublementary budet or government deficit or national deficit or federal deficit or fiscal policy or fiscal or budget deficit* or budget gap* or national budget or government budget or fiscal stimulus or balanced budget or balance the budget
  - General fiscal search keyword is the union of tax, government expenditure, public debt and other keywords.

### Validation and Robustness of the Index
- Validation exercises to address false positives:
  - Human audits:
    - Manual review of articles from selected dates with particularly high index values to confirm narratives corroborating high index values.
  - Time series and alternative-method comparisons:
    - Compare movements with existing economic policy uncertainty indicators.
    - Construct two alternative fiscal policy uncertainty indicators for selected countries:
      - (i) A stochastic volatility approach as in Fernández-Villaverde et al. (2015) and Jurado, Ludvigson, and Ng (2015).
      - (ii) A textual analysis using LLM on Financial Times articles.
  - Outcomes of validation presented in subsequent subsections (per source).
- LLM-based textual validation:
  - LLM textual analysis on Financial Times articles from January 1985 to July 2023 using an open AI source with prompts to assess whether a fiscal policy mentioned increases uncertainty.
  - Global LLM-based GFPU (“LLM-based GFPU”) shows a remarkable similarity to the news-based GFPU.
  - US-specific LLM vs. news-based FPU: correlation of 0.78 for the US.
  - Subcategory findings (US):
    - Government spending-related policy uncertainty: clear upward trajectory with spikes during COVID-19 pandemic.
    - Public debt-related policy uncertainty: clear spikes corresponding to US debt ceiling episodes.
  - Explanation: Baker, Bloom, and Davis (2016) US indicator includes only tax and government expenditure-related terms; the news-based FPU also captures government debt-related uncertainty.

### Country-Level and Global FPU Indexes, Stylized Facts
- Country-level examples and coverage:
  - Figure 1 (per source) plots seven country-specific FPU indexes: United States, United Kingdom, France, Italy, Japan, Greece, Argentina with identified narratives for key months/events.
  - Appendix G contains additional country narratives (e.g., Brazil, Thailand, Indonesia, Hungary).
  - 189 country-level FPU indexes constructed (per source).
- Global FPU (GFPU) construction:
  - Baseline GFPU: global keyword search across newspapers for all countries each month.
  - Two alternative GFPU bottom-up approaches:
    - (i) Principal component analysis (PCA) across 189 country-level FPU indexes.
    - (ii) GDP-weighted average of the 189 country-level FPUs.
  - Strong correlations exceeding 0.9 between baseline GFPU and these alternatives validate baseline GFPU.
- Stylized facts and observed behaviors:
  - FPU indexes show dramatic spikes during fiscal stress events (examples: US debt ceiling gridlocks, government shutdowns).
  - Fiscal policy uncertainty tends to escalate during elections when potential fiscal trajectories diverge (example: France: sharp increase in June and July 2024 coinciding with political turbulence around a snap election).
  - Certain fiscal stress events detected primarily by FPU (examples: UK mini-budget incident; several US debt-ceiling episodes).
    - UK: EPU rose most in 2016 due to Brexit; highest spike in FPU occurs in the 2022 mini-budget episode.
    - US: FPU and EPU behave differently in response to debt ceiling deadlocks.
  - Comovements between FPU and EPU can change over time:
    - Japan: FPU and EPU movements fairly consistent throughout sample.
    - Greece: correlation between FPU and EPU rose from 0.2 (pre-crisis) to 0.6 (post-crisis); similar pattern observed for Italy after European sovereign debt crisis.

### Comparisons and Correlations with Other Indicators
- Monthly correlations:
  - GFPU and global economic policy uncertainty (Baker, Bloom, and Davis (2016), PPP-adjusted and normalized): correlation of 0.45.
  - GFPU and CBOE VIX: correlation of 0.28.
- Quarterly comparisons and stochastic-volatility fiscal measures:
  - Quarterly aggregation used to compare with quarterly-frequency indicators.
  - Fiscal volatility measures:
    - Volatility of primary balance-to-GDP and changes in debt-to-GDP estimated using autoregressive model allowing for stochastic volatility (Fernández-Villaverde et al. (2015) approach).
    - Aggregation uses PPPGDP-weighted average of log of time-varying standard deviation of innovations to primary balance-to-GDP and changes in debt-to-GDP at country level.
  - Table 2 correlations between GFPU and quarterly indicators:
    - PB-SV / Debt-SV / Fiscal-WUI
      - Pre-COVID: 0.40 / 0.60 / 0.45
      - Full sample: 0.38 / 0.52 / 0.41
    - Pre-COVID sample runs from January 1977 to February 2020. The full sample runs from January 1977 to July 2024.
  - Interpretation:
    - Correlations with PB-SV and Debt-SV are around 0.4 and 0.6 respectively in pre-COVID sample; some reduction in full sample possibly because these measures reflect backward-looking, realized fiscal outcomes and not fiscal policy responses to shocks, particularly during COVID-19.
  - Comparison with fiscal components of the World Uncertainty Index (Ahir, Bloom, and Furceri (2022)) shows similar patterns; correlations robust at around 0.4 despite timing discrepancies from monthly-to-quarterly aggregation.

### VAR-Based Analysis — Setup and Baseline Identification
- Baseline VAR endogenous variables:
  - Y_t = [g f p u_t, s p500_t, i_t, i pUS_t, i p AEexUS_t, i pEM_t]
    - where g f p u_t is global FPU index; s p500_t is log of the S&P 500; i_t is US shadow interest rate; i pUS_t is log US industrial production; i p AEexUS_t is log industrial production for advanced economies excluding the US; i pEM_t is log industrial production for emerging market economies.
- Data frequency and sample:
  - Monthly data running from January 1987 to April 2023.
- Lag structure and COVID treatment:
  - Baseline includes three lags.
  - Dummy variables for the first five months of 2020 used to avoid extreme observations associated with the COVID-19 pandemic.
- Identification in baseline:
  - Cholesky decomposition ordering places the GFPU index as the most exogenous shock.

### Baseline VAR Results — Economic Impacts of a GFPU Shock
- Overall:
  - A one-standard deviation shock to the GFPU index has contractionary effects across all income groups, with statistical significance.
- Financial market effects (US):
  - At impact, US stock prices drop by 4 percent, reaching their trough 3 months after the shock.
  - US monetary policy reacts in an accommodative manner, indicated by a reduction of the shadow interest rates.
- Real activity effects (global and by income group):
  - The GFPU shock lowers industrial production by as much as 0.5 percent.
  - Peak impacts on real activity occur about five months following the shock.
  - The impact is persistent: the negative impact lingers 6 months and even 20 quarters after the shock for advanced economies excluding the US.
- Figure presentation conventions:
  - Sample: January 1987 to April 2023.
  - Solid line: Median; Dashed line: 68% intervals; Shaded area: 90% intervals.

### Alternative Identifications and Robustness Checks (VAR)
- Alternative Cholesky ordering:
  - Y_t = [i pUS_t, i p AEexUS_t, i pEM_t, g f p u_t, s p500_t, i_t].
  - Results similar to baseline in magnitude and persistence.
- Sign restrictions combined with narrative and FEVD restrictions:
  - Combined approach: sign restrictions + narrative restriction (Antolín-Díaz and Rubio-Ramírez, 2018) + FEVD restriction (Barsky and Sims, 2011; Volpicella, 2022).
  - Sign restriction summary:
    - Positive GFPU shock leads to contemporaneous increase in GFPU.
    - Responses of U.S. stock prices, U.S. industrial production (IP), U.S. interest rates, AE (ex-U.S.) IP, and EM IP are negative at least once within first six months.
  - Narrative restriction:
    - Shock is positive in the August 2011 US debt ceiling crisis (S&P downgrade from AAA to AA+).
  - Variance decomposition restriction:
    - GFPU shock causes largest forecast error variance decomposition of GFPU in the first forecast horizon.
    - Appendix H: alternative bound restriction imposing GFPU shock explain at least 50 percent of FEVD of GFPU in the first month.
- Results with sign restrictions (qualitatively similar but larger):
  - US stock price declines by more than 6 percent (trough three months after the shock).
  - US industrial production declines by about 1 percent.
  - Industrial production of advanced economies excluding the US and in emerging markets declines by approximately 1.4 percent, respectively.
  - Responses are larger and more persistent; largest real activity declines between 4 and 6 months after shock.
- No timing restrictions imposed; FEVD uses max share identification and alternative bounded restrictions considered.

### Extensions and Cross-Country Spillovers
- Robustness overview:
  - Additional robustness checks vary the sample depending on data availability.
- Excluding COVID-19 observations:
  - Baseline uses dummy variables for COVID; alternative uses pre-COVID sample (1987M1-2019M12).
    - IRFs using pre-COVID sample lead to similar results as baseline (Appendix Figure 14).
  - Estimating baseline Cholesky model on full sample without dummies:
    - Finding: Larger decline in industrial production of US and other country groups (Appendix Figure 15).
  - Sample/date references: January 1987 to April 2023.
- Controlling for global economic policy uncertainty (GEPU):
  - Include GEPU in VAR ranked second after GFPU to allow contemporaneous effect of GFPU on GEPU.
  - Finding: Increase in GFPU leads to an unambiguous increase in GEPU (Appendix Figure 16).
  - Controlling for GEPU yields quantitatively similar results as baseline.
- Different lag structures:
  - Baseline uses three lags; alternatives: two lags, four lags, six lags (Appendix Figures 17–19).
  - Finding: Results similar across lag specifications.
- Cross-country spillovers: Country FPU vs. Global FPU (Panel-VAR exercises):
  - Empirical approach: three layers of balanced Panel-VAR models covering 26 countries.
  - Model 1 — Country FPU: domestic FPU and domestic industrial production; exogenous: log US stock index, federal funds rate, US industrial production.
    - Response to one standard deviation country-level FPU shock (Model 1): domestic industrial production declines about 0.2 percent.
  - Model 2 — Global FPU: replace domestic FPU with GFPU.
    - Response: decline reaches as much as 0.3 percent, 12 months following the shock.
  - Model 3 — Global FPU with US feedback: GFPU first, then US variables endogenously, then country IP.
    - Response: domestic industrial production contracts by as much as 0.6 percent within four months; effects persistent with similar magnitude after one year.
  - Interpretation: cross-country amplification sizable when global trade and financial channels and US feedback are allowed; “agreed” uncertainty may generate large, protracted contractionary effects (reference Gambetti et al. (2023)).

### Financial Impacts of Fiscal Policy Uncertainty
- Extended VAR including sovereign bond spreads and excess bond premium:
  - Sample: December 1997 to April 2023 (due to EMBI availability).
- Key findings:
  - Tighter financial conditions after a GFPU increase remain robust.
  - Negative impact on industrial production larger and more persistent compared to baseline.
  - Advanced economy spreads:
    - After a one standard deviation increase of GFPU, spreads in advanced economies rise gradually, reaching peak of 12 basis points five months following the shock; impact dissipates after a year.
  - Emerging market economies (EMBI spread):
    - Impact larger and more persistent — EMBI rises substantially by 30 basis points at impact, further increasing to about 50 basis points before fading out gradually.
  - Excess bond premium: sharp contemporaneous increase, indicating deterioration of investor sentiment or risk appetite in corporate bond market.
  - Financial variables respond more pronouncedly for emerging markets relative to advanced economies.
- Additional evidence:
  - Increase in GFPU lowers stock prices and leads to a contemporaneous, short-lived increase in stock volatility, peaking in the second month following the shock.
- Global financial cycle exercise (Miranda-Agrippino and Rey (2020) VAR extended with GFPU):
  - Sample: 1990M2-2010M12.
  - Identification: include US monetary policy shocks directly and rank them first; place GFPU after US monetary policy shock series.
  - Finding: Surprise increase in GFPU generates co-movements of global financial variables:
    - At impact, global risky asset prices fall sharply; global risk aversion jumps; credit spreads widen.
    - Sharp contraction in global economic activity and deleveraging in US and European banks.
  - Robustness: Replacing GFPU with US fiscal policy uncertainty yields similar results; results robust after controlling for global EPU.
  - Implication: Global and US fiscal policy uncertainty can be important drivers of the global financial cycle, even after controlling for US monetary policy shocks.

### Data, Appendices, and Country-Specific Procedures (Example: Afghanistan entry and appendices)
- Data description (Table 4) — key variables and sources:
  - Global FPU index — Factiva, see Section 2
  - US Stock Price Index- S&P’s 500 Composite — Haver Analytics
  - US Federal Funds Effective Rate — Haver Analytics
  - US Shadow Short Rate — Haver Analytics
  - US industrial production — Fed Dallas’s Global Economic Indicators
  - Other countries’ industrial production — Haver Analytics
  - Countries’ primary balance-to-GDP (quarterly) — Haver Analytics
  - Countries’ debt-to-GDP (quarterly) — Haver Analytics
  - Advanced economies (ex-US) industrial production — Fed Dallas’s Global Economic Indicators
  - Emerging economies industrial production — Fed Dallas’s Global Economic Indicators
  - US VIX — Haver Analytics
  - US VXO — Haver Analytics
  - US Excess Bond Premium — Gilchrist and Zakrajšek (2012) and FEDS
  - AE (ex-US) Long term interest rate — Fed Dallas’s Global Economic Indicators
  - Emerging Market Bond Index (EMBI) Spread — J.P. Morgan
  - Note: For Section 5 global financial cycle analysis, dataset mirrors Miranda-Agrippino and Rey (2020) global VAR exercise.
- News sources and FPU construction:
  - Table 5 lists the 43 Factiva news sources used for FPU index (examples: Chicago Tribune Online, The Dallas Morning News, The San Francisco Chronicle, USA Today Online, Wall Street Journal, The Washington Post, ABC News programs, BBC Monitoring regional feeds, CBS News programs, NBC News programs, The Daily Telegraph, The Times, The Globe and Mail, The Guardian, The Economist, Reuters).
  - Appendix D provides Boolean search queries combining legislative/government terms, uncertainty terms, fiscal terms, and country name token hlp=(country name).
- Country-specific FPU procedure (LLM-assisted classification for Financial Times archive):
  - Question cascade for country-level classification (FT archive January 1985 to July 2023 as noted in LLM section):
    1. Is [country name] mentioned? Answer "Yes" or "No".
    2. If yes, is [country name]’s tax policy mentioned? Answer "Yes" or "No".
    3. If tax policy is mentioned, does the news reflect that tax policy or its effects are more uncertain? Answer "Yes, more uncertain" or "No".
    4. Provide explanation for Q3 in natural language.
  - Tax uncertainty measure = (number of articles on tax uncertainty from Q3) divided by (number of articles mentioning the country from Q1). Analogous for government expenditure and government debt.
  - Fiscal policy uncertainty index = (counts on tax uncertainty + government spending/expenditure uncertainty + government debt uncertainty) divided by number of articles mentioning the country (Q1).
  - Global FPU procedure excludes country name from prompts and uses last three questions for global measure.
- Appendix and figure coverage references:
  - Appendix A: list of countries.
  - Appendix C: full list of publishers.
  - Appendix D: exact queries for country-specific FPU.
  - Appendix E: number of publications by news outlets and geographical coverage.
  - Appendix F: specific prompt used for LLM textual analysis (Appendix F.1 shows the comparison for the UK).
  - Appendix G: narratives for a sample of selected emerging economies (examples: Brazil, Thailand, Indonesia, Hungary).
  - Appendix H: alternative FEVD bound restriction.

### Main Conclusions and Policy Implications
- Database and capture:
  - Novel news-based indicators of fiscal policy uncertainty for 189 countries capture severe economic downturns and coincide with fiscal stress events (e.g., US debt ceiling episodes, UK’s mini-budget episode in 2022) often missed by existing economic policy uncertainty indicators.
- Observed patterns:
  - In some countries, upward trend in fiscal policy uncertainty in recent years; speculative causes include high public debt with limited fiscal space, polarized political views, and emerging spending pressures.
- Empirical conclusions:
  - Fiscal policy uncertainty adversely affects real economic activity and deteriorates financial conditions.
  - Contractionary effects on real activity are similar across income groups.
  - Emerging market economies are more vulnerable, facing sharper rises in sovereign borrowing costs in response to increases in global fiscal policy uncertainty.
  - Substantial cross-country spillover effects exist; adverse effects amplified when global trade and financial channels are allowed.
  - Key fiscal policy uncertainty events that attract global attention could be important drivers of the global financial cycle, even after controlling for US monetary policy shocks.
- Suggested avenues for future research:
  - In-depth analysis of possible non-linear transmission of fiscal policy uncertainty shocks (e.g., high versus low fiscal policy uncertainty periods).
  - Investigate factors that weaken or strengthen adverse impact of fiscal policy uncertainty (e.g., country-level institutional factors such as rule of law and quality of governance; macroeconomic environment such as high versus low inflation periods, high versus low unemployment).
  - Examine whether global spillover effects of US monetary policy depend on the state of global fiscal policy uncertainty.

*Source: Section 2 and related sections, "News-Based Fiscal Policy Uncertainty Index" and extensions (from the provided content unit).*

### Section 2 describes the methodology used to construct the fiscal policy uncertainty indicator at

### News-Based Fiscal Policy Uncertainty Index

### 2.1 Construction of News-Based Fiscal Policy Uncertainty Index
- Data source and coverage:
  - News articles from Dow Jones Factiva, covering more than 30,000 news sources.
  - Sample restricted to printed articles published by 43 major English-language news outlets in the US, Canada, and the UK.
  - Dataset spans from January 1977 to July 2024, encompassing over 47 million articles.
  - Inclusion of international news outlets (e.g., “ABC News: World News Saturday”, International BBC network, Reuters) to improve coverage for small emerging markets and low-income countries and reduce overstatement when denominators (total number of articles per month) are small.
- Article selection logic:
  - Construct a list of keywords built around three layers corresponding to “fiscal”, “policy” and “uncertainty”:
    - One of the fiscal topic terms in tax, government expenditure, public debt, or others.
    - Policy authority identifier (like “congress”, “parliament”, “minister”).
    - An uncertainty identifier (like “uncertainty”, “ambiguous”, “unpredictable”).
  - For country-level FPU indexes, overlay the above conditions with a country identifier (e.g., “US”, “China”) and restrict to articles where a country name appears in the title, snippet, or main paragraph identified by Factiva.
  - The perimeter of fiscal policy corresponds to components of a typical government budget constraint:
    - B_{t+1} = R_t B_t + G_{t+1} − T_{t+1}
      - where B_t denotes the end-of-period total outstanding public debt, R_t denotes the gross interest rate on the public debt, G_t denotes the government spending and T_t denotes tax revenues.
  - Exact queries used for country-specific FPU are reported in Appendix D (per source).
- Table 1 (search keywords summarized from source):
  - Category I. Policy authority:
    - government or parliament or congress* or regulat* or policy or policies or legislat* or minister or ministry or national assembly
  - Category II. Uncertainty:
    - uncertain* or ambiguous or dubious or precarious or unpredictable or undecided or undertermined or unresolved or unsettled or risk or doubt or doubtful or volatilit*
  - Category III. Fiscal Topics:
    1. Tax:
       - taxation or taxes or taxed or tax
    2. Government expenditure:
       - government spending or federal spending or public spending or government expenditure or federal expenditure or public expenditure or defense spending or military spending or pension reform* or pension expenditure or healthcare expenditure or medical care expenditure or social securities expenditure or social safety net* or public investment* or social protection or subsidies or subsidy or entitlement spending or social security
    3. Public debt:
       - federal debt or government debt or national debt or public debt or debt ceiling or sovereign debt or debt burden or debt repay* or debt sustainability or government borrowing or sovereign borrowing or debt consolidation or sovereign risk or sovereign default or debt default or debt restructure or foreign debt or external debt or government bond or sovereign bond or sovereign yield
    4. Others:
       - public sector or public finance or sublementary budet or government deficit or national deficit or federal deficit or fiscal policy or fiscal or budget deficit* or budget gap* or national budget or government budget or fiscal stimulus or balanced budget or balance the budget
  - Note: The truncation character (*) indicates any possible ending for a term, e.g. “govern*” includes “government” or “governor”. The general fiscal search keyword is the union of tax, government expenditure, public debt and other keywords.

### 2.2 Validation of the Index
- Validation exercises to address false positives:
  - Human audits:
    - Manual review of newspaper articles from selected dates with particularly high index values to confirm narratives corroborating high index values.
  - Time series and alternative-method comparisons:
    - Compare movements of the news-based indexes with existing economic policy uncertainty indicators.
    - Construct two alternative fiscal policy uncertainty indicators for selected countries and compare them with the news-based indicators:
      - (i) A stochastic volatility approach as in Fernández-Villaverde et al. (2015) and Jurado, Ludvigson, and Ng (2015).
      - (ii) A textual analysis using LLM on Financial Times articles.
  - Outcomes of these validation processes are presented in subsequent subsections (per source).

### 2.3 Country-Level and Global Fiscal Policy Uncertainty Indexes
- Country-level examples and narratives:
  - Figure 1 in the source plots seven country-specific FPU indexes for the United States, the United Kingdom, France, Italy, Japan, Greece, and Argentina with identified narratives corresponding to key months/events.
  - Additional country narratives (e.g., Brazil, Thailand, Indonesia, Hungary) are reported in Appendix G (per source).
- GFPU (Global Fiscal Policy Uncertainty) construction and alternatives:
  - Baseline GFPU index constructed by conducting a global keyword search across newspapers for all countries for each month.
  - Two alternative GFPU indicators (bottom-up approaches):
    - (i) Principal component analysis (PCA) across 189 country-level FPU indexes.
    - (ii) GDP-weighted average of the 189 country-level FPUs.
  - Strong correlations exceeding 0.9 between baseline GFPU and these alternative indicators validate the baseline GFPU’s capture of key cross-country and systemically important-economy events.
- Stylized facts and observed behaviors:
  - FPU indexes show dramatic spikes during fiscal stress events (examples: US debt ceiling gridlocks, government shutdowns).
  - Fiscal policy uncertainty tends to escalate during elections when potential fiscal trajectories diverge.
    - Example: France shows a sharp increase in June and July 2024 coinciding with political turbulence around a snap election.
  - Some fiscal stress events are detected primarily by the FPU indicator (examples: UK mini-budget incident; several US debt-ceiling episodes).
    - UK: economic policy uncertainty (EPU) rose most in 2016 due to Brexit; highest spike in fiscal policy uncertainty occurs in the 2022 mini-budget episode.
    - US: FPU and EPU behave differently in response to debt ceiling deadlocks.
  - Comovements between FPU and EPU can change over time across countries:
    - Example: Japan — FPU and EPU movements fairly consistent throughout the sample.
    - Example: Greece — correlation between FPU and EPU rose from 0.2 (pre-crisis) to 0.6 (post-crisis), suggesting fiscal factors became more central to economy-wide policy uncertainty. Similar pattern observed for Italy after the European sovereign debt crisis.

### Comparisons and Correlations with Other Indicators
- Monthly correlations:
  - GFPU and global economic policy uncertainty (Baker, Bloom, and Davis (2016), PPP-adjusted and normalized): correlation of 0.45 (Figure 4, Panel 1).
  - GFPU and CBOE VIX: correlation of 0.28.
- Quarterly comparisons and stochastic-volatility fiscal measures:
  - Quarterly aggregation used to compare with indicators available at quarterly frequency.
  - Fiscal volatility measures:
    - Volatility of primary balance-to-GDP and changes in debt-to-GDP ratios estimated using an autoregressive model allowing for stochastic volatility (Fernández-Villaverde et al. (2015) approach).
    - Aggregation uses PPPGDP-weighted average of log of time-varying standard deviation of innovations to primary balance-to-GDP and changes in debt-to-GDP at country level.
  - Table 2: Correlations between GFPU and various quarterly indicators (exact values from source):
    - PB-SV / Debt-SV / Fiscal-WUI
      - Pre-COVID: 0.40 / 0.60 / 0.45
      - Full sample: 0.38 / 0.52 / 0.41
    - Note: Pre-COVID sample runs from January 1977 to February 2020. The full sample runs from January 1977 to July 2024.
  - Interpretation:
    - Correlations with PB-SV and Debt-SV are around 0.4 and 0.6 respectively in the pre-COVID sample; some reduction in correlations in the full sample possibly due to these measures reflecting backward-looking, realized fiscal outcomes and not accounting for fiscal policy responses to shocks (automatic stabilizers), particularly during COVID-19.
  - Comparison to fiscal components of the World Uncertainty Index (Ahir, Bloom, and Furceri (2022)) shows similar patterns; correlations robust at around 0.4 despite potential timing discrepancies from monthly-to-quarterly aggregation.

### LLM-based Textual Validation
- LLM textual analysis details and findings:
  - LLM-based textual analysis conducted on Financial Times articles from January 1985 to July 2023 using an open AI source with prompts designed to assess whether a specific fiscal policy mentioned in an article is likely to increase uncertainty.
  - Global LLM-based GFPU (“LLM-based GFPU”) shows a remarkable similarity to the news-based GFPU (per Figure 5 in source).
  - Country-specific LLM validation conducted for the UK and the US (Financial Times coverage skewed to UK and US).
  - US-specific LLM-based sub-indicators constructed for taxes, government spending, and public debt:
    - Correlation between news-based FPU and LLM-based FPU for the US: 0.78 (upper left panel in Figure 5 per source).
    - Subcategory patterns:
      - Government spending-related policy uncertainty: clear upward trajectory with spikes during COVID-19 pandemic.
      - Public debt-related policy uncertainty: clear spikes corresponding to US debt ceiling episodes.
    - Explanation: Differences between news-based FPU and Baker, Bloom, and Davis (2016) US indicator arise because the latter includes only tax and government expenditure-related terms, whereas the news-based FPU also captures government debt-related uncertainty.

### Additional methodological notes and robustness checks
- Country coverage: 189 country-level FPU indexes constructed (per source).
- Appendices referenced by source (A, C, D, E, F, G) contain:
  - Appendix A: list of countries.
  - Appendix C: full list of publishers.
  - Appendix D: exact queries for country-specific FPU.
  - Appendix E: number of publications by news outlets and geographical coverage.
  - Appendix F: specific prompt used for LLM textual analysis.
  - Appendix G: narratives for a sample of selected emerging economies.
- Robustness checks:
  - PCA and GDP-weighted averages as alternative GFPU measures; correlations exceeding 0.9 with baseline GFPU.
  - Comparison with PB-SV, Debt-SV, and Fiscal-WUI at quarterly frequency.
  - LLM-based validation using Financial Times articles.

*Source: Section 2, "News-Based Fiscal Policy Uncertainty Index" (from the provided content unit).*

### Appendix F.1 shows the comparison for the UK.

### Appendix F.1 shows the comparison for the UK.

### VAR-based analysis: setup and data
- Endogenous variables in baseline VAR:
  - Y_t = [g f p u_t, s p500_t, i_t, i pUS_t, i p AEexUS_t, i pEM_t]
  - where g f p u_t is the global FPU index; s p500_t is the log of the S&P 500; i_t is the US shadow interest rate; i pUS_t is log of US industrial production; i p AEexUS_t is log industrial production for advanced economies excluding the US; i pEM_t is log industrial production for emerging market economies.
- Data frequency and sample:
  - Monthly data running from January 1987 to April 2023.
- Lag structure and COVID treatment:
  - Baseline specification includes three lags.
  - Dummy variables for the first five months of 2020 are used to avoid extreme observations associated with the COVID-19 pandemic.
- Identification in baseline:
  - Cholesky decomposition ordering follows equation (2), placing the GFPU index as the most exogenous shock.

### Baseline results: economic impacts of a GFPU shock
- Overall:
  - A one-standard deviation shock to the GFPU index has contractionary effects across all income groups, with statistical significance.
- Financial market effects (US):
  - At impact, US stock prices drop by 4 percent, reaching their trough 3 months after the shock.
  - US monetary policy reacts in an accommodative manner, indicated by a reduction of the shadow interest rates.
- Real activity effects (global and by income group):
  - The GFPU shock lowers industrial production by as much as 0.5 percent.
  - Peak impacts on real activity occur about five months following the shock.
  - The impact is persistent: the negative impact lingers 6 months and even 20 quarters after the shock for advanced economies excluding the US.
- Figure notes:
  - Sample: January 1987 to April 2023.
  - Figure presents the economic impact of a shock to the global fiscal uncertainty index.
  - Solid line: Median; Dashed line: 68% intervals; Shaded area: 90% intervals.

### Alternative identifications and robustness checks
- Alternative Cholesky decomposition:
  - Alternative ordering: Y_t = [i pUS_t, i p AEexUS_t, i pEM_t, g f p u_t, s p500_t, i_t].
  - Results are similar to baseline in magnitude and persistence.
- Sign restrictions with narrative and FEVD restrictions:
  - Combined approach: sign restrictions + narrative restriction (Antolín-Díaz and Rubio-Ramírez, 2018) + FEVD restriction (Barsky and Sims, 2011; Volpicella, 2022).
  - Sign restriction summary:
    - A positive GFPU shock leads to a contemporaneous increase in GFPU.
    - Responses of U.S. stock prices, U.S. industrial production (IP), U.S. interest rates, AE (excluding U.S.) IP, and EM IP are negative at least once within the first six months following the shock.
  - Narrative restriction:
    - The shock is positive in the August 2011 US debt ceiling crisis.
    - Context: Standard & Poor’s downgraded the US long-term credit rating in August 2011 from AAA to AA+.
  - Variance decomposition restriction:
    - GFPU shock causes the largest forecast error variance decomposition of GFPU in the first forecast horizon.
    - Appendix H: alternative bound restriction imposing that GFPU shock explain at least 50 percent of FEVD of GFPU in the first month of the shock.
- Results with sign restrictions (qualitatively similar to baseline, with larger magnitudes and persistence):
  - US stock price declines by more than 6 percent (trough three months after the shock).
  - US industrial production declines by about 1 percent.
  - Industrial production of advanced economies excluding the US and in emerging markets declines by approximately 1.4 percent, respectively.
  - Responses of industrial production are larger and more persistent compared to the baseline, with the largest decline occurring between 4 and 6 months after the shock.
- Additional methodological notes:
  - No restrictions are imposed on the timing of the impact; the data reveals magnitude and timing profile.
  - The FEVD restriction uses max share identification and alternative bounded restrictions are considered.

*wpiea2024209-print-pdf - Appendix F.1 shows the comparison for the UK.*

### 3.3    Extensions

### 3.3    Extensions

### Robustness checks overview
- Provide additional robustness checks to substantiate the baseline findings.
- The sample varies in these exercises depending on the data availability.

### Excluding observations since the COVID-19 pandemic
- Baseline specification uses dummy variables to treat extreme observations due to the COVID pandemic.
- Alternative: compare baseline IRFs with those using only the pre-COVID episode sample (1987M1-2019M12).
  - Finding: IRFs using the pre-COVID sample lead to similar results as the baseline (Appendix Figure 14).
  - Reference: consistency with Lenza and Primiceri (2022).
- Additional exercise: estimate the baseline Cholesky model using the full sample without introducing the dummies.
  - Finding: Larger decline in industrial production of US as well as other country groups (Appendix Figure 15), indicating the need to take these abnormal variations into account.
- Sample/date references: January 1987 to April 2023.

### Controlling for global economic policy uncertainty (GEPU)
- Question addressed: To what extent does global fiscal policy uncertainty (GFPU) add new components beyond existing global economic policy uncertainty?
- Method: include global economic policy uncertainty (GEPU) into the VAR model; rank GEPU in the second position after GFPU to allow contemporaneous effect of GFPU on GEPU.
- Finding: An increase in GFPU leads to an unambiguous increase in the GEPU (Appendix Figure 16).
- Robustness: Controlling for GEPU yields quantitatively similar results as in the baseline.

### Different lag structures
- Baseline uses three lags.
- Alternative specifications evaluated: two lags, four lags, and six lags.
  - Results for two lags, four lags, and six lags are reported in Appendix Figures 17, 18, and 19, respectively.
  - Finding: Results are similar across different lag specifications.

### Cross-country spillover effects: Country-specific FPU vs. Global FPU
- Research question: Compare economic impact of country-specific fiscal policy uncertainty (FPU) with global fiscal policy uncertainty (GFPU).
- Rationale: Unclear a priori whether origin of FPU matters for contractionary effects; systemic countries may generate sizable global spillovers; GFPU can amplify domestic shocks via tighter global financial conditions.
- Empirical approach: estimate three layers of balanced Panel-VAR models (three models described).
  - Note: Analysis includes 26 countries (list provided in source).
- Model descriptions:
  - Model 1 - Country FPU: Panel-VARX with two endogenous variables per country: domestic fiscal policy uncertainty index and domestic industrial production; exogenous variables: log of the US stock index, the federal funds rate, and the US industrial production index.
  - Model 2 - Global FPU: Panel-VARX similar to Model 1 but replace domestic FPU with GFPU.
  - Model 3 - Global FPU and US feedback: Panel-VAR including GFPU first, then US variables (log of US stock, federal funds rate, US IP), then country-specific industrial production; allows US variables to react endogenously.
- Findings on industrial production (Figure 9):
  - Response to a one standard deviation increase in country-level FPU shock (Model 1): domestic industrial production declines about 0.2 percent.
  - Response under Model 2 (Global FPU, no US feedback): decline reaches as much as 0.3 percent, 12 months following the shock.
  - Response under Model 3 (Global FPU with US feedback): domestic industrial production contracts by as much as 0.6 percent within four months; effects are persistent with broadly similar magnitude even after one year.
  - The Panel-VAR result (Model 3) is in line with baseline models, corroborating estimates.
- Interpretation and mechanisms:
  - Exact amplification mechanisms beyond scope; possible explanation from Gambetti et al. (2023) on “agreed” uncertainty vs. “disagreed” uncertainty:
    - “Agreed” uncertainty (low disagreement among agents) generates large, protracted contractionary effects.
    - Events generating sudden increases in GFPU (e.g., US debt ceiling deadlocks) may correspond to “agreed” uncertainty.

### Financial impact of fiscal policy uncertainty
- Objective: explore financial impact of FPU and document co-movements of global financial variables in response to GFPU shock.
- Model extension: include sovereign bond spreads for advanced and emerging market economies and the excess bond premium to capture corporate credit risk.
- Sample: December 1997 to April 2023 (shorter than baseline due to EMBI spread availability).
- Key findings (Figure 10):
  - Tighter financial conditions following an increase in GFPU remain robust despite sample differences.
  - Negative impact on industrial production for different country groups increased in magnitude and persistence compared to baseline.
  - Advanced economy spreads: after a one standard deviation increase of GFPU, spreads in advanced economies rise gradually, reaching peak of 12 basis points five months following the shock; impact dissipates after a year.
  - Emerging market economies (EMBI spread): impact larger and more persistent — EMBI rises substantially by 30 basis points at impact, further increasing to about 50 basis points before fading out gradually.
  - Excess bond premium: sharp contemporaneous increase, indicating deterioration of investor sentiment or risk appetite in corporate bond market.
  - Financial variables respond more pronouncedly for emerging markets compared to advanced economies.
- Additional evidence:
  - An increase in GFPU lowers stock prices and leads to a contemporaneous, short-lived increase in stock volatility, peaking in the second month following the shock (Appendix Figure 20) — suggests spillover from fiscal uncertainty to financial uncertainty.
- Global financial cycle exercise:
  - Extend Miranda-Agrippino and Rey (2020) VAR by adding GFPU.
  - Sample used for this extension: 1990M2-2010M12.
  - Identification: include US monetary policy shocks directly and rank them first; place GFPU after US monetary policy shock series.
  - Findings (Figure 11): A surprise increase in GFPU generates co-movements of global financial variables:
    - At impact, global risky asset prices fall sharply; global risk aversion jumps; credit spreads widen.
    - Generates sharp contraction in global economic activity and deleveraging in US and European banks.
  - Robustness: Replacing GFPU with US fiscal policy uncertainty yields similar results; results robust after controlling for global EPU (Appendix Fig 21).
  - Implication: Global and US fiscal policy uncertainty, in addition to US monetary policy shocks, can generate the global financial cycle.

### Conclusion and policy implications (summary)
- Database: novel news-based indicators of fiscal policy uncertainty for 189 countries.
- Indicators capture severe economic downturns and coincide with fiscal stress events (e.g., US debt ceiling episodes, UK’s mini-budget episode in 2022) often missed by existing economic policy uncertainty indicators.
- Observed patterns:
  - In some countries, upward trend in fiscal policy uncertainty in recent years; speculative causes include high public debt with limited fiscal space, polarized political views, and emerging spending pressures.
- Main empirical conclusions:
  - Fiscal policy uncertainty adversely affects real economic activity and deteriorates financial conditions.
  - Contractionary effects on real activity are similar across income groups.
  - Emerging market economies are more vulnerable, facing sharper rises in sovereign borrowing costs in response to increases in global fiscal policy uncertainty.
  - Substantial cross-country spillover effects exist; adverse effects are amplified when global trade and financial channels are allowed.
  - Key fiscal policy uncertainty events that attract global attention could be important drivers of the global financial cycle, even after controlling for US monetary policy shocks.
- Suggested avenues for future research:
  - In-depth analysis of possible non-linear transmission of fiscal policy uncertainty shocks (e.g., high versus low fiscal policy uncertainty periods).
  - Investigate factors that weaken or strengthen the adverse impact of fiscal policy uncertainty (e.g., country-level institutional factors such as the rule of law and quality of governance; macroeconomic environment such as high versus low inflation periods, high versus low unemployment).
  - Examine whether global spillover effects of US monetary policy depend on the state of global fiscal policy uncertainty.

*Source: IMF working paper content (section 3.3 Extensions).*

### 1.  Afghanistan

### 1. Afghanistan

### Content coverage in this unit
- Afghanistan is listed as item 1 in the country index (countries numbered 1 through 189).
- The surrounding pages include appendices and methodological material used across countries in the chapter/report rather than country-specific analysis for Afghanistan.

### Key methodological and data components present in the unit
- Table 4. Data Description: Lists variables and their sources used in the analysis, including:
  - Global FPU index — Factiva, see Section 2
  - US Stock Price Index- S&P’s 500 Composite — Haver Analytics
  - US Federal Funds Effective Rate — Haver Analytics
  - US Shadow Short Rate — Haver Analytics
  - US industrial production — Fed Dallas’s Global Economic Indicators
  - Other countries’ industrial production — Haver Analytics
  - Countries’ primary balance-to-GDP (quarterly) — Haver Analytics
  - Countries’ debt-to-GDP (quarterly) — Haver Analytics
  - Advanced economies (ex-US) industrial production — Fed Dallas’s Global Economic Indicators
  - Emerging economies industrial production — Fed Dallas’s Global Economic Indicators
  - US VIX — Haver Analytics
  - US VXO — Haver Analytics
  - US Excess Bond Premium — Gilchrist and Zakrajšek (2012) and FEDS
  - AE (ex-US) Long term interest rate — Fed Dallas’s Global Economic Indicators
  - Emerging Market Bond Index (EMBI) Spread — J.P. Morgan
- Note: For analysis on the global financial cycle in Section 5, the same dataset as in Miranda-Agrippino and Rey (2020) for their global VAR exercise is used.

### News sources and FPU construction
- Table 5 lists the 43 Factiva news sources used in constructing the Fiscal Policy Uncertainty (FPU) index. Examples included in the unit:
  - Chicago Tribune Online, The Dallas Morning News, The San Francisco Chronicle, USA Today Online, Wall Street Journal, The Washington Post
  - ABC News (multiple programs), BBC Monitoring (regional feeds), CBS News (multiple programs), NBC News (multiple programs)
  - The Daily Telegraph, The Times, The Globe and Mail, The Guardian, The Economist, Reuters, and others
- Appendix D provides the Boolean search query used to identify relevant articles for country-level FPU (includes logical combinations of legislative/government terms, uncertainty terms, fiscal terms, and the country name token hlp=(country name)).

### Construction of the Fiscal Policy Uncertainty (FPU) indicator
- Country-specific FPU procedure:
  - Use an LLM to classify articles from the Financial Times archive (January 1985 to July 2023 for FT prompts as noted) with question cascade:
    1. Is [country name] mentioned? Answer "Yes" or "No".
    2. If yes, is [country name]’s tax policy mentioned? Answer "Yes" or "No".
    3. If tax policy is mentioned, does the news reflect that tax policy or its effects are more uncertain? Answer "Yes, more uncertain" or "No".
    4. Provide explanation for Q3 in natural language.
  - Tax uncertainty measure = (number of articles on tax uncertainty from Q3) divided by (number of articles mentioning the country from Q1). Analogous procedures for government expenditure and government debt.
  - Fiscal policy uncertainty index = (counts on tax uncertainty + government spending/expenditure uncertainty + government debt uncertainty) divided by number of articles mentioning the country (Q1).
- Global FPU procedure:
  - Exclude the country name from prompts; use only the last three questions to construct global fiscal policy uncertainty.

### Examples and narratives
- Appendix G shows country-level FPU examples with narratives for Brazil, Thailand, Indonesia, and Hungary, listing dated episodes and short descriptions (e.g., Brazil: July-October 2002: Fears of sovereign debt crisis; Thailand: July-September 2011: Thai flood, high living costs; Indonesia: January-February 1998: Asian Financial Crisis; Hungary: November 2011: Moody’s cut Hungary’s rating to junk).

### Empirical analyses and figures included
- Sign restrictions with narrative and bounded share: Figure 13 presents the economic impact of a shock to the global fiscal policy uncertainty index for January 1987 to April 2023; plots include Global FPU, US stock price, US interest rate, US IP, AE excluding-US IP, EM IP with medians and 68 percent credible intervals.
- Extensions:
  - Pre-COVID sample (January 1987 to December 2019): Figure 14 shows impacts with median, 68% intervals, and shaded 90% intervals.
  - Full sample without special treatment for COVID-related extreme observations (January 1987 to April 2023): Figure 15.
  - Controlling for Global Economic Policy Uncertainty (GEPU) (January 1987 to April 2023): Figure 16.
- Robustness checks with different lag structures (January 1987 to April 2023):
  - VAR with two lags: Figure 17.
  - VAR with four lags: Figure 18.
  - VAR with six lags: Figure 19.
- Financial markets analyses:
  - Impact of GFPU controlling for financial risks (VIX) with medians, 68% intervals, and shaded 90% intervals: Figure 20.
  - Impact of GFPU on global financial variables controlling for GEPU (February 1990 to December 2010) with responses to a 1 standard deviation increase in global fiscal policy uncertainty while controlling for GEPU; included variables: FedFF contracts, Global FPU, Global EPU, Global Financial Cycle, Global Risk Aversion, Industrial Production, PCE Deflator, 1Y Treasury Rate, BIS Real EER, Global Real Economic Activity Ex US, Global Domestic Credit, Global Inflows All Sectors, GZ Credit Spread, Leverage US Brokers & Dealers, Leverage EU Global Banks, Leverage US Banks, Leverage EU Banks: Figure 21. Solid line: Median; Dashed line: 68% intervals; Shaded area: 95% intervals.

### Notes on temporal coverage and interval reporting in figures
- Several figures explicitly state sample periods and interval conventions:
  - Figures using January 1987 to April 2023 sample state: Solid line is the median response and dashed lines are the 68 percent credible intervals.
  - Pre-COVID figure uses January 1987 to December 2019 and shows Solid line: Median; Dashed line: 68% intervals; Shaded area: 90% intervals.
  - Global VAR extension controlling for GEPU uses February 1990 to December 2010 and shows Solid line: Median; Dashed line: 68% intervals; Shaded area: 95% intervals.

*Italic: Source: wpiea2024209-print-pdf - 1.  Afghanistan (canonical PDF content as supplied).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024209-print-pdf.pdf_