## wpiea2021262-print-pdf

## Source details

**Canonical URL:** [wpiea2021262-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021262-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021262-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021262-print-pdf.pdf.json)

---

### Literature review: identification approaches and related strands
- Two main identification approaches for fiscal spending effects:
  - Structural Vector Autoregression (SVAR) techniques (Blanchard and Perotti 2002) using a standard Cholesky decomposition with government spending ordered first.
  - Narrative approach (Romer and Romer 2010, Devries et al. 2011) to identify size, timing, and objective of fiscal shocks.
- Key extensions and relevant methods:
  - Ramey (2011): timing and anticipation effects; introduced a narrative “news” variable to capture anticipated government spending shocks.
  - High-frequency event studies: Gurkaynak, Sack and Swanson (2004); Gertler and Karadi (2015); Auerbach and Gorodnichenko (2016).
  - Empirical COVID-19 fiscal work: Benmelech and Tzur-Ilan (2020); Gourinchas et al. (2021) — fiscal policy offset about 8% of the economic downturn at the global level.
- This paper’s contribution:
  - Follows the narrative approach with a daily database of fiscal policy measures (announcement and implementation dates).
  - Examines effects of daily fiscal policy announcements on high-frequency economic activity indicators and heterogeneity across countries and pandemic severity.

### Data: sources, construction, and classification (Section III)
- Main fiscal data source:
  - Yale’s COVID-19 Financial Response Tracker (CFRT), supplemented and cross-checked with the IMF Policy Tracker, the OECD Country Policy Tracker, and newspaper reports.
- Coverage and scaling:
  - Daily fiscal policy measures for 52 countries with industrial production and PMI manufacturing data, from January 1 to December 31, 2020.
  - All data converted to USD and scaled to a percent of a country’s 2019 GDP.
- Quality controls:
  - Cross-check discrepancies using announcement links and newspaper reports.
  - Removal of duplicates and entries reported as financing or concessional lending by the IMF or World Bank.
- Policy-tool grouping (8 sub-categories):
  - asset guarantees; asset purchases; capital injections; credit guarantees; grants; loans; payment forbearances; tax relief measures.
- Above-the-line vs below-the-line classification:
  - Above-the-line: grants, payment forbearances, tax relief interventions (reflected in fiscal balance and government debt).
  - Below-the-line: other interventions with no immediate/upfront effect on a government’s deficit.
- Demand-support vs emergency lifeline classification:
  - Demand-support: boost demand and disposable income (cash transfers, unemployment insurance, wage subsidies, reduction/deferral of social security or tax payments, paid sick leave, public investments, healthcare spending).
  - Emergency lifeline: sustained cashflow support/liquidity (loans, umbrella guarantees, equity injections, other liquidity support measures). Most emergency lifelines are below-the-line; demand-support measures mainly above-the-line.
- Construction of fiscal shocks:
  - Daily announcements purged of predictable components using high-frequency indicators (NO2 emissions, international and domestic flights, mobility indicators, bilateral exchange rates, stock market indices, and other daily financial variables).
  - Residual of the regression used as the fiscal shock. Multiple shocks created using different indicator combinations for robustness.

### Stylized facts on fiscal announcements (Section IV)
- Sample composition and magnitudes:
  - 52 countries: 27 advanced economies (AEs) and 25 emerging market and developing economies (EMDEs).
  - Largest announced intervention: about 35 percent of 2019 GDP (Japan).
  - Largest announced intervention in an EMDE: about 15 percent of GDP (Indonesia).
  - Mean fiscal stimulus response:
    - Advanced economies: about 12 percent of GDP.
    - EMDEs: about 4 percent of GDP.
- Timing and rollout:
  - Responses were most frequent and significant in March and April 2020.
  - Advanced economies responded faster: 10 advanced economies announced stimulus measures on March 16, 2020.
  - Responses continued into June 2020, then diminished owing to dwindling fiscal space and relaxation of containment measures toward the end of 2020Q2.
- Composition and evolution:
  - Emergency lifelines concentrated in March and April 2020, but continued throughout the year.
  - Demand-support measures picked up toward the end of the second quarter of 2020 as reopening began.
  - Tool usage patterns:
    - Lifeline measures (loans, credit guarantees) more frequently used in EMDEs.
    - Demand-support measures (grants, tax relief) prominent in AEs.
  - Magnitude by instrument:
    - AEs: credit guarantees, grants, tax relief measures, and loans were largest.
    - EMDEs: grants, payment forbearances and loans were the largest three instruments.
- Fiscal accounting and space:
  - About 70 percent of stimulus measures for both AEs and EMDEs were below-the-line.
  - Countries with higher debt-to-GDP levels relied more on below-the-line measures; countries with more fiscal space favored above-the-line measures.

### Empirical methodology and main results (Section V)
- Empirical approaches:
  - High-frequency dynamic responses estimated using Jordà (2005) local projection method for daily horizons h = 0,..,30 and weekly horizons up to h = 15.
  - Daily regressions for stock prices and bilateral exchange rates; weekly regressions for the OECD weekly economic activity tracker.
  - Baseline monthly regressions (unbalanced sample of 52 economies, January to December 2020):
    - ∆Y_i,t = α + u_i + β Shock_i,t + θ X_i,t + ε_i,t
    - Y: industrial production (preferred), unemployment, manufacturing PMI, OECD CLI, sovereign CDS spreads.
    - Fiscal shock at monthly frequency: aggregate of daily shocks (simple average in baseline).
  - Controls: containment measures, mobility, retail, COVID-19 infections and fatalities, monetary policy announcements and policy rate changes, country fixed effects, country time-trends.
  - Standard errors: Driscoll-Kraay (1998).
- High-frequency (daily/weekly) results:
  - Announcements led to:
    - An appreciation of bilateral exchange rates of about 5 percent 30 days after announcement.
    - Stock market cumulative gains of 40 percent logs following announcement.
  - Weekly frequency: announced fiscal measures significantly boosted economic activity as proxied by the OECD weekly tracker.
- Monthly-frequency results and fiscal multipliers (baseline):
  - A one percent of GDP fiscal shock led, on average, to a 0.25 percent increase in the growth rate of industrial production (year-on-year).
  - Back-of-the-envelope multiplier based on historical IPI-GDP relationship implies a fiscal multiplier of about 0.2.
  - Manufacturing PMI: 0.39 percent increase associated with a one percent of GDP fiscal shock.
  - OECD CLI: about 0.1 percent increase.
  - Fiscal measures had economically and statistically significant effects on unemployment and sovereign CDS spreads (decline in unemployment; compression of CDS spreads).
- Selected baseline coefficients (Table 1):
  - Industrial Production (y-o-y growth): 0.254*** (0.057)
  - PMI Manufacturing: 0.190** (0.074)
  - Composite Leading Indicator (y-o-y growth): 0.089* (0.043)
  - Unemployment Rate (change): -0.053** (0.017)
  - CDS Spreads: -0.0443*** (0.008)
  - Containment Measures Index examples: Industrial Production: -21.212*** (3.497); CDS Spreads: 10.261*** (3.293)
  - COVID-19 Cases example: Industrial Production: -0.928*** (0.194)

### Robustness checks and specification variations
- Additional controls and alternative specifications:
  - Controls extended to include monetary policy announcements and policy rate changes, COVID-19 fatalities and retail mobility, and expectations from monthly consensus GDP growth forecasts; results robust (Table 2 and Annex Tables A1-4).
  - Time fixed effects and lagged dependent variables included in alternative specifications (Table 3 and Table 4); results remain robust.
- Aggregation and weighting:
  - Baseline aggregation: daily fiscal shocks aggregated into monthly using simple average (equal weights).
  - Alternative weighting scheme tested: 45 percent weight for announcements in the first ten days of the month, 35 percent for the next 10 days, and 20 percent for the remainder; results similar to baseline (Table 5).
- Announcements vs implementations:
  - Implementation-date based shocks constructed; results broadly similar to announcement-based results (Table 6).
  - Typical implementation lag in database: about a 1-week delay between announcement and implementation for most measures.
- Selected robustness coefficient examples:
  - Robustness check additional controls (Industrial Production): 0.285*** (0.068); other specs: 0.251*** (0.055), 0.301*** (0.059), 0.333*** (0.075), 0.317*** (0.068)
  - Time fixed effects (Fiscal Shock coefficients range): 0.129* (0.066) to 0.188** (0.083)
  - Lagged dependent variable (Industrial Production): 0.240*** (0.065)
  - Implemented measures (Table 6, Industrial Production): 0.269*** (0.053); CDS Spreads: -1.006** (0.347)

### Heterogeneity: country characteristics and measure types
- Effectiveness by country and debt:
  - Fiscal shocks more effective in advanced economies.
  - Fiscal shocks more effective in countries with lower public debt pre-crisis; crowding out larger in high-debt countries.
  - Sovereign CDS spread response by debt level:
    - Narrowing of 0.15 bps in countries with lower public debt-to-GDP ratio.
    - Narrowing of 0.04 bps in countries with higher debt ratios.
- Measure-type heterogeneity:
  - Emergency lifelines (loans, credit guarantees, provisions to firms) had a larger impact than demand-support measures during the COVID-19 sample (Table 8).
  - Demand-support measures are less effective during lockdowns when supply constraints bind; more effective as containment eases.
  - Classification rule: emergency lifeline (demand-support) dummy = 1 when corresponding measures constitute more than 50 percent of a country’s fiscal stimulus in 2020; robustness checks use 60 percent, 65 percent, and 70 percent thresholds.
- Interaction with pandemic cycle and mobility:
  - Emergency lifelines more effective when containment measures were high and mobility low (providing cashflow and liquidity during constrained months) (Table 9; Table 10).
  - Demand-support measures stronger when containment measures were eased and mobility higher (result for mobility not statistically significant).
- Selected interaction coefficients and tests:
  - Fiscal Shock * Emergency Lifelines Dummy: 0.414** (0.140)
  - Fiscal Shock * Demand Support Dummy: -0.049 (0.097)
  - Fiscal Shock * Below-the-Line Measures Dummy: 0.420** (0.137)
  - Fiscal Shock * Above-the-Line Measures Dummy: -0.074 (0.090)
  - F-test Difference examples (Table 9): 3.27, 5.62*, 8.12*; (Table 10): 1.66, 3.63, 5.12*

### Key quantitative relationships and supporting tables
- GDP–Industrial Production relationship (Table A11):
  - GDP Growth (y-o-y) coefficient on Industrial Production (y-o-y): 0.990*** (0.113) without Time Fixed-Effects; 1.200*** (0.203) with Time Fixed-Effects.
  - Constant terms: -0.017*** (0.002) and -0.021*** (0.007).
  - Observations: 291; No. of countries: 60; R-squared: 0.258 and 0.303.
- Representative baseline and robustness statistics (preserved exact figures where reported in tables and text).

### Conclusions and policy recommendations
- Main conclusions:
  - Announced fiscal measures during the COVID-19 pandemic were, on average, effective and statistically robust in stimulating economic activity and improving confidence indicators across the 52-country sample in 2020.
  - Emergency lifelines (below-the-line measures such as loans, credit guarantees, equity injections) were more effective during periods of lockdown and supply-side shocks.
  - Demand-support measures (tax cuts, payment deferrals, cash transfers, unemployment insurance) were more effective as containment measures eased and domestic consumption opportunities expanded.
  - Fiscal policy effects are larger in advanced economies and in countries with lower public debt.
- Policy implications:
  - Emergency lifelines should not be withdrawn prematurely given their vital role during the COVID-19 crisis.
  - As supply constraints abate, shift to demand-support policies—including investment in digital and green infrastructure—which will be more effective and can replace emergency lifelines.
  - In countries with high debt, fiscal policy is less effective and monetary policy may need to play a larger role to support economic activity.

*Source: wpiea2021262-print-pdf*

### Section III describes the data and Section IV presents some stylized facts on the fiscal measures.

### wpiea2021262-print-pdf - Section III describes the data and Section IV presents some stylized facts on the fiscal measures.

### Literature review: identification approaches and related strands
- Two main identification approaches for fiscal spending effects:
  - Structural Vector Autoregression (SVAR) techniques (Blanchard and Perotti 2002), assuming government spending pre-determined within the quarter using a standard Cholesky decomposition with government spending ordered first.
  - Narrative approach (Romer and Romer 2010, Devries et al. 2011) to identify size, timing, and objective of fiscal shocks.
- Key contributions and extensions:
  - Ramey (2011): importance of timing and anticipation effects; introduced a narrative “news” variable from reports to capture anticipated government spending shocks.
  - Use of high-frequency event studies for policy analysis: Gurkaynak, Sack and Swanson (2004); Gertler and Karadi (2015); Auerbach and Gorodnichenko (2016).
  - Empirical work on fiscal policy during COVID-19: Benmelech and Tzur-Ilan (2020); Gourinchas et al. (2021) — fiscal policy offset about 8% of the economic downturn at the global level.
- This paper’s contribution:
  - Follows the narrative approach using a daily database of fiscal policy measures with announcement and implementation dates.
  - Examines effects of daily fiscal policy announcements on high-frequency economic activity indicators and how effects vary across countries and pandemic severity.

### Data (Section III): sources, construction, and classification
- Main data source:
  - Yale’s COVID-19 Financial Response Tracker (CFRT) (supplemented and cross-checked with the IMF Policy Tracker, the OECD Country Policy Tracker, and newspaper reports).
- Time frame and coverage:
  - Daily fiscal policy measures for 52 countries for which industrial production and PMI manufacturing data are available, from January 1 to December 31, 2020.
  - All data converted to USD and scaled to a percent of a country’s 2019 GDP.
- Measures to ensure data quality:
  - Cross-checking discrepancies among datasets with announcement links and newspaper reports to determine best assessment of date, size, and type.
  - Removal of duplicates and entries reported as financing or concessional lending by the IMF or World Bank.
- Grouping of policy tools into eight sub-categories (as read from narratives):
  - i) asset guarantees
  - ii) asset purchases
  - iii) capital injections
  - iv) credit guarantees
  - v) grants
  - vi) loans
  - vii) payment forbearances
  - viii) tax relief measures
- Above-the-line vs below-the-line classification (following the IMF classification of fiscal measures):
  - Above-the-line: grants, payment forbearances (on tax payments and pension contributions), and tax relief interventions (likely reflected in fiscal balance and government debt).
  - Below-the-line: all other interventions with no immediate or upfront effect on a government’s deficit.
- Demand-support vs emergency lifeline classification:
  - Demand-support: measures that boost demand and household or firm disposable income (cash transfers, unemployment insurance, wage subsidies, reduction/deferral of social security or tax payments, paid sick leave, public investments, healthcare spending).
  - Emergency lifeline: measures providing sustained cashflow support to households and firms (loans and umbrella guarantees, government provision of loans, equity injections, other liquidity support measures). Most emergency lifelines are below-the-line; demand-support measures mainly fall under above-the-line.
- Construction of fiscal shocks:
  - Daily announcements purged of predictable components using high-frequency indicators (NO2 emissions, international and domestic flights, mobility indicators, bilateral exchange rates, stock market indices, and other daily financial variables).
  - The residual of the regression is used as the fiscal shock. Several fiscal shocks created based on different combinations of high-frequency indicators for robustness.

### Stylized facts on fiscal announcements (Section IV)
- Sample composition and magnitudes:
  - 52 countries: 27 advanced economies (AEs) and 25 emerging market and developing economies (EMDEs).
  - Largest announced intervention: about 35 percent of 2019 GDP (Japan).
  - Largest announced intervention in an EMDE: about 15 percent of GDP (Indonesia).
  - Mean fiscal stimulus response:
    - Advanced economies: about 12 percent of GDP.
    - EMDEs: about 4 percent of GDP.
- Timing and rollout:
  - Responses were most frequent and significant in March and April 2020.
  - Advanced economies responded faster: 10 advanced economies announced stimulus measures on March 16, 2020.
  - Responses continued into June 2020, then diminished owing to dwindling fiscal space and relaxation of containment measures toward the end of 2020Q2, prompting more targeted measures.
- Composition and evolution:
  - Emergency lifelines rollout concentrated in March and April 2020, but continued throughout the year.
  - Demand-support measures picked up toward the end of the second quarter of 2020, coinciding with the beginning of reopening in many countries.
  - Across policy tools:
    - Lifeline measures: loans to households and businesses, and credit guarantees — more frequently used in EMDEs.
    - Demand-support measures: grants (including cash transfers, wage subsidies, healthcare spending) and tax relief measures.
  - Magnitude by instrument:
    - AEs: credit guarantees, grants (including cash transfers, wage subsidies, healthcare spending), tax relief measures (tax cuts, tax payment deferrals), and loans were largest.
    - EMDEs: grants, payment forbearances and loans were the largest three instruments.
- Fiscal accounting and space:
  - About 70 percent of stimulus measures for both AEs and EMDEs were below-the-line measures.
  - Countries with higher debt-to-GDP levels relied more on below-the-line measures; countries with more fiscal space favored above-the-line measures.

### Analysis (Section V): methodology and main empirical results
- Empirical approaches:
  - High-frequency dynamic responses estimated using Jordà (2005) local projection method for daily (h = 0,..,30) and weekly (h up to 15) horizons.
  - Daily regressions for stock prices and bilateral exchange rates; weekly regressions for the OECD weekly economic activity tracker.
  - Baseline monthly regressions (unbalanced sample of 52 economies, January to December 2020) for standard economic activity indicators using:
    - ∆Y_i,t = α + u_i + β Shock_i,t + θ X_i,t + ε_i,t
    - Y: industrial production (preferred indicator), unemployment, manufacturing PMI, OECD CLI, sovereign CDS spreads.
    - Fiscal shock at monthly frequency: aggregate of daily shocks (simple average in baseline).
  - Controls include: containment measures, mobility, retail, number of COVID-19 infections and fatalities, monetary policy announcements and policy rate changes, country fixed effects, and country time-trends.
  - Standard errors: Driscoll-Kraay (1998) to account for cross-sectional and time dependence.
- High-frequency (daily/weekly) results:
  - Fiscal stimulus announcements in 2020 led to:
    - An appreciation of bilateral exchange rates of about 5 percent 30 days after their announcement.
    - Stock market cumulative gains of 40 percent logs following their announcement.
  - At the weekly frequency, announced fiscal measures significantly boosted economic activity as proxied by the OECD weekly tracker.
- Monthly-frequency results and fiscal multipliers:
  - Baseline monthly regressions (Table 1 summary):
    - A one percent of GDP fiscal shock has led, on average, to a 0.25 percent increase in the growth rate of industrial production (year-on-year).
    - Back-of-the-envelope calculation based on the historical relationship between industrial production and GDP implies a fiscal multiplier of about 0.2.
    - Manufacturing PMI: 0.39 percent increase associated with a one percent of GDP fiscal shock.
    - OECD composite leading indicator (CLI): about 0.1 percent increase.
    - Fiscal measures had economically and statistically significant effects on unemployment and sovereign CDS spreads (results indicate a decline in unemployment and compression of CDS spreads following shocks).

*Source: wpiea2021262-print-pdf*

### 0.06 percentage point, and a narrowing of sovereign CDS spreads by about 0.05 bps. The reduction

### wpiea2021262-print-pdf - 0.06 percentage point, and a narrowing of sovereign CDS spreads by about 0.05 bps. The reduction

### Key findings on effectiveness of fiscal announcements
- Announced fiscal policy measures in response to the COVID-19 pandemic have, on average, been effective in stimulating economic activity as captured by high-frequency indicators: industrial production, manufacturing PMI, unemployment, confidence indicators, and sovereign CDS spreads.
- The reduction in CDS spreads following the announcement of fiscal measures suggests that the measures were seen by the markets as positive news for debt sustainability; the reduction in CDS spread is statistically significant only for advanced economies.
- Specific magnitudes and parameters:
  - Historically, a one percentage point change in GDP growth results in a 0.9 percent increase in industrial production (Table A11), based on the reduced equation with K = 0.9.
  - Regressions estimating the effect of the fiscal shock on industrial production use ∆IPI/IPI = β * ∆G/Y, where β = 0.25.
  - The multiplier effect translated to GDP is ∆Y/∆G = 0.2.
  - The weighting function parameter γ = 1.5.

### Robustness checks and methodology notes
- Controls extended to include monetary policy announcements and policy rate changes, COVID-19 fatalities and retail mobility, and expectations of economic activity based on monthly consensus GDP growth forecasts. Results reported in Table 2 and Annex Tables A1-4 are robust to these controls.
- Time fixed effects and lagged dependent variables were included in alternative specifications (Table 3 and Table 4) to address time-dependence; results remain robust.
- Baseline aggregation: daily fiscal shocks aggregated into monthly using simple average (equal weights to all days of the month).
- Alternative weighting scheme tested: 45 percent weight for announcements in the first ten days of the month, 35 percent for the next 10 days, and 20 percent for the remainder of the days of the month. Results (Table 5) are similar to, and not statistically different from, the baseline.
- Implementation-date based shocks were constructed to check announcements vs implementations; results (Table 6) are broadly similar to those obtained with announcements.
- Typical implementation lag: in the database, for most measures there is only about a 1-week delay between the announcement and implementation.

### Country characteristics affecting fiscal shock effectiveness
- Fiscal shocks have been more effective in advanced economies.
- Fiscal shocks are likely to be much more effective in countries with lower public debt levels before the crisis; crowding out effects on private investment and consumption are typically larger in countries with high debt.
- Evidence on sovereign CDS spreads by debt level:
  - Narrowing of 0.15 bps of sovereign CDS spreads following a fiscal shock in countries with lower public debt-to-GDP ratio.
  - Narrowing of 0.04 bps in countries with higher debt ratios.
- No evidence that effects vary with degree of trade openness in this sample; likely due to limited cross-country variation and generally anemic trade during the crisis.
- Sample coverage: announcements implemented across 52 countries throughout 2020.
- Correlation and classification notes:
  - The correlation between below-the-line measures and emergency lifelines is around 0.8.
  - Thresholds used to classify packages: emergency lifeline (demand-support) dummy equal to 1 if emergency lifelines (demand-support measures) make up more than 50 percent of the fiscal stimulus package in 2020; robustness also checked at 60 percent (Annex Table A9) and different thresholds (65, or 70 percent).

### Type of fiscal measures and interaction with the pandemic cycle
- Model extensions distinguish packages by measure type: emergency lifelines, demand-support measures, below-the-line, and above-the-line.
- Classification rule: emergency lifeline (demand-support) dummy = 1 when emergency lifelines (demand-support measures) constitute more than 50 percent of the country’s fiscal stimulus in 2020.
- On average:
  - Emergency lifelines (loans, credit guarantees, provisions to firms, etc.) had a larger impact than demand-support measures during the COVID-19 sample (Table 8).
  - Demand-support measures (many classified as above-the-line) are likely less effective during lockdowns when supply constraints bind.
- Heterogeneity over the pandemic cycle:
  - Emergency lifelines were more effective when containment measures were high, providing cashflow and liquidity support during constrained months of economic activity (Table 9).
  - Demand-support measures were much stronger when containment measures were being eased, allowing more opportunities for increased domestic consumption.
  - Similar pattern when using mobility measures: lifelines more effective when mobility is low; demand-support coefficient higher with high mobility (result not statistically significant) (Table 10).
- Results robust to alternative thresholds for differentiating emergency lifelines and demand-support measures (Table A10 and Annex Tables A5-8; alternative thresholds such as 60 percent, 65 percent, or 70 percent reported as robustness checks).

### Conclusions and policy recommendations
- Countries launched historically unprecedented fiscal support measures during the COVID-19 pandemic; a novel daily database of announcements for 52 countries in 2020 was assembled to quantify effects.
- Main conclusions:
  - Announced fiscal measures were, on average, effective and statistically robust in stimulating economic activity and in improving confidence indicators.
  - Emergency lifelines (below-the-line measures such as loans, credit guarantees, equity injections) were more effective during periods of lockdown and supply-side shocks.
  - Demand-support measures (tax cuts, payment deferrals, cash transfers, unemployment insurance) were more effective as containment measures eased and domestic consumption opportunities expanded.
  - Fiscal policy effects are larger in advanced economies and in countries with lower public debt.
- Policy implications:
  - Emergency lifelines should not be withdrawn prematurely as they have been vital to support the economy during the COVID-19 crisis.
  - As supply constraints abate, demand-support policies—including investment in digital and green infrastructure—will be more effective and can replace emergency lifelines.
  - In countries with high debt, fiscal policy is less effective and monetary policy might have to play a larger role to support economic activity.

*Source: wpiea2021262-print-pdf (excerpt).*

### References

### wpiea2021262-print-pdf - References

### Key bibliographic sources
- Major referenced works on fiscal multipliers, fiscal consolidations, and COVID-19 fiscal/monetary responses, including:
  - Alesina, A., Favero, C. and Giavazzi, F., 2014. The output effect of fiscal consolidation plans. Journal of International Economics, 96, pp.S19-S42.
  - Auerbach AJ, Gorodnichenko Y (2013) Fiscal multipliers in recession and expansion.
  - Benmelech Efraim, and Nitzan Tzur-Ilan. 2020. “The Determinants of Fiscal and Monetary Policies During the Covid-19 Crisis”. NBER Working Paper 27461.
  - Ilzetzki, E., Mendoza, E.G. and Végh, C.A., 2013. How big (small?) are fiscal multipliers?. Journal of monetary economics, 60(2), pp.239-254.
  - Ramey, Valerie A. 2011. "Identifying Government Spending Shocks: It's all in the Timing," The Quarterly Journal of Economics, vol. 126(1), pages 1-50.
  - IMF. 2020. 'The Great Lockdown: Dissecting the Economic Effects'. World Economic Outlook, Chapter 2.
- Technical and methodological references for econometric approaches:
  - Driscoll, J.C. and Kraay, A.C., 1998. Consistent covariance matrix estimation with spatially dependent panel data. Review of Economics and Statistics, 80(4), pp.549-560.
  - Jordà, Òscar. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” The American Economic Review, vol. 95, no. 1, pp. 161–182.
  - Granger, Clive W. J. & Terasvirta, Timo, 1993. "Modelling Non-Linear Economic Relationships."

### Empirical results and substantive findings (selected table highlights)
- Baseline effect of announced fiscal measures on economic activity (Table 1; sample of 52 countries, daily data Jan 1, 2020–Dec 30, 2020):
  - Fiscal Shock coefficients:
    - Industrial Production (y-o-y growth): 0.254*** (0.057)
    - PMI Manufacturing: 0.190** (0.074)
    - Composite Leading Indicator (y-o-y growth): 0.089* (0.043)
    - Unemployment Rate (change): -0.053** (0.017)
    - CDS Spreads: -0.0443*** (0.008)
  - Containment Measures Index coefficients:
    - Industrial Production: -21.212*** (3.497)
    - PMI Manufacturing: -19.780*** (2.388)
    - CLI: -8.538*** (1.479)
    - Unemployment Rate: -0.685 (0.592)
    - CDS Spreads: 10.261*** (3.293)
  - COVID-19 Cases coefficients:
    - Industrial Production: -0.928*** (0.194)
    - PMI Manufacturing: -0.050 (0.197)
    - CLI: -0.185** (0.067)
    - Unemployment Rate: 0.073 (0.094)
    - CDS Spreads: -0.586* (0.294)
  - Observations and sample sizes:
    - Observations: 480, 382, 396, 292, 600 (respectively)
    - Number of countries: 40, 32, 33, 30, 50 (respectively)
  - R-squared values: 0.565, 0.648, 0.602, 0.0219, 0.388 (respectively)

- Robustness and specification checks (selected coefficients preserved exactly):
  - Robustness check: additional controls (Table 2)
    - Fiscal Shock (Industrial Production): 0.285*** (0.068); other specifications show 0.251*** (0.055), 0.301*** (0.059), 0.333*** (0.075), 0.317*** (0.068)
    - Containment Measures Index examples: -11.296* (6.055), -21.929*** (3.468), -21.540*** (6.292), -8.877*** (1.078), -10.698*** (1.148)
  - Time fixed effects (Table 3)
    - Fiscal Shock coefficients range: 0.129* (0.066) to 0.188** (0.083) across outcomes with Time FE Yes.
  - Lagged dependent variable (Table 4)
    - Fiscal Shock (Industrial Production): 0.240*** (0.065)
    - Lagged Dep. Variable example: 0.173** (0.063) for Industrial Production; -0.684*** (0.085) for Unemployment Rate
  - Alternative weights aggregation (Table 5)
    - Fiscal Shock (Industrial Production): 0.223*** (0.062), 0.216*** (0.062), 0.267*** (0.059), 0.286*** (0.068), 0.286*** (0.068)

- Implemented measures (Table 6)
  - Fiscal Shock (implemented measures):
    - Industrial Production: 0.269*** (0.053)
    - PMI Manufacturing: 0.419** (0.147)
    - CLI: 0.094** (0.041)
    - Unemployment Rate: -0.104*** (0.019)
    - CDS Spreads: -1.006** (0.347)
  - Observations: 480, 382, 396, 292, 600; R-squared: 0.565, 0.653, 0.602, 0.0250, 0.388; Number of countries: 40, 32, 33, 30, 50

- Heterogeneity by country and measure type (selected results)
  - Effect by country characteristics (Table 7)
    - High State * Fiscal Shock (Industrial Production): 0.264*** (0.056)
    - Low State * Fiscal Shock: 0.166 (0.394)
    - Public Debt to GDP interaction: Low State * Fiscal Shock = 0.777*** (0.099)
  - By type of measure (Table 8)
    - Fiscal Shock: 0.254*** (0.057)
    - Fiscal Shock * Emergency Lifelines Dummy: 0.414** (0.140)
    - Fiscal Shock * Demand Support Dummy: -0.049 (0.097)
    - Fiscal Shock * Below-the-Line Measures Dummy: 0.420** (0.137)
    - Fiscal Shock * Above-the-Line Measures Dummy: -0.074 (0.090)
    - Share of emergency lifeline (below-the-line) measures: 65% and 68%
  - Interaction with containment measures (Table 9) and mobility (Table 10) show statistically significant regime differences; F-test Difference examples: 3.27, 5.62*, 8.12* (Table 9); 1.66, 3.63, 5.12* (Table 10)
  - By type for specific outcomes (selected):
    - PMI Manufacturing (Table A5): Fiscal Shock 0.190** (0.074); Fiscal Shock * Demand Support Dummy 0.197** (0.065)
    - CLI (Table A6): Fiscal Shock 0.089* (0.043); Fiscal Shock * Demand Support Dummy 0.072*** (0.018)
    - Unemployment (Table A7): Fiscal Shock -0.053** (0.017); Fiscal Shock * Demand Support Dummy 0.059*** (0.013)
    - CDS spreads (Table A8): Fiscal Shock -0.044*** (0.009); Fiscal Shock * Above-the-Line Measures Dummy -0.060** (0.021)

- Relationship between GDP and Industrial Production (Table A11)
  - GDP Growth (y-o-y) coefficient on Industrial Production (y-o-y):
    - 0.990*** (0.113) without Time Fixed-Effects
    - 1.200*** (0.203) with Time Fixed-Effects
  - Constant terms: -0.017*** (0.002) and -0.021*** (0.007)
  - Observations: 291; No. of countries: 60; R-squared: 0.258 and 0.303

### Figures (descriptions of empirical visuals)
- Figure 1–6: Cross-country summaries of fiscal stimulus packages in 2020, including mean and maximum package size scaled as percent of 2019 GDP across 52 countries (27 advanced economies and 27 EMDEs), daily announced fiscal stimulus measures over 2020, and breakdowns by type (Emergency Lifeline vs Demand Support), policy tool frequency, and maximum tool size (percent of 2019 GDP).
- Figure 7: Impulse response functions (sample of 42 countries, daily Jan 1, 2020–Dec 31, 2020) showing effect of fiscal announcements on bilateral exchange rates and stock market prices; confidence bands at 90 and 95 percent; model specification and notation provided.
- Figure 8: Impulse response functions (sample of 46 countries, weekly Jan 1, 2020–Dec 31, 2020) showing effect of fiscal announcements on the OECD weekly tracker; confidence bands at 90 and 95 percent; model specification and notation provided.

### Data Annex: data sources, construction, and classification rules
- Main fiscal policy data source: Yale’s COVID-19 Financial Response Tracker (CFRT); supplemented and cross-checked with IMF Policy Tracker, OECD Country Policy Tracker, and newspaper reports.
- Country selection: refined to 52 countries with available industrial production and PMI manufacturing data (Table D1 lists countries).
- Institutional role: fiscal database uses policy actions proposed by fiscal authorities; actions not clearly fiscal were omitted.
- Policy tools grouped into 8 sub-categories: asset purchases; asset guarantees; capital injections; tax relief measures; grants; payment forbearances; loans; credit guarantees.
- Policy instruments identified per measure (e.g., “cash transfers”, “healthcare spending”, “sector support”).
- Classification by type of measure:
  - Demand-support measures: spending-side (wage subsidies, targeted transfers, unemployment benefits, paid sick leave, support to parents) and revenue-side (sectoral support, reduction of social security contributions, tax relief, deferrals of tax/social security payments).
  - Emergency lifelines: liquidity measures (loans to firms/households, umbrella guarantees, government loans, equity injections).
- Classification by accounting principle:
  - Above-the-line: reflected in fiscal balance and government debt in short term (grants, tax relief, payment forbearance).
  - Below-the-line: no upfront impact on fiscal deficit but increase debt/liabilities later (loans, capital injections, asset purchases, guarantees, credit guarantees).
- Monetary policy measures: monthly key policy rates for 38 countries from Haver Analytics; Yale CFRT used for monetary authority announcements (daily frequency Jan 1–Dec 30, 2020 for 52 countries).
- High-frequency indicators and other data:
  - NO2 emissions from WAQI (ppb), Google retail mobility, OxCGRT Stringency Index (normalized 0–1), Johns Hopkins COVID-19 cases, FlightRadar24 flights, bilateral exchange rates and stock indices from Bloomberg, OECD weekly activity tracker (46 countries).
  - Monthly indicators: Industrial Production Indices (40 countries), Manufacturing PMIs (32 countries), Composite Leading Indicator (CLI) from OECD (33 countries), Sovereign CDS spreads (50 countries), Unemployment rates (30 countries), Consensus Forecasts (85 countries).
- Country characteristics for interactions: income levels (AE vs. EMDE), public debt, trade openness drawn from IMF World Economic Outlook Database as of end-2019.

*Italic: Source — wpiea2021262-print-pdf (References, Figures, Tables, Data Annex).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021262-print-pdf.pdf_