## _wp0950

## Source details

**Canonical URL:** [_wp0950](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0950.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0950.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0950.pdf.json)

---

### I. Introduction and summary
- Context: examines cyclical behavior of fiscal and monetary policy in the G7 since 1980; addresses the view that fiscal policy provides "too little stimulus too late" and may be asymmetric over the cycle.
- Objectives:
  - Use quarterly data and an event study to analyze frequency, timing, instrument variation, and comparison with monetary policy (Section II).
  - Use a VAR to quantify strength and speed of fiscal and monetary responses to a growth shock, document withdrawal of fiscal stimulus during recoveries, and compare symmetry (Section III).
  - Assess timeliness and temporariness of countercyclical tax legislation for the United States using Romer and Romer (2007) data (Section IV).

### II. Event-study analysis — data and methodology
- Downturn definition:
  - Periods when either the growth rate is negative or the output gap is “unusually” negative (defined as one standard deviation below zero).
- Policy indicators:
  - Monetary policy: change in the short-term nominal interest rate (quarter-over-quarter, unannualized).
  - Fiscal policy: Primary fiscal balance; Cyclically-adjusted primary balance (discretionary); Cyclical primary balance (automatic stabilizers).
- Discretionary change thresholds:
  - Fiscal: exceed 0.25 percent of GDP per quarter.
  - Nominal short-term interest rates: exceed 25 basis points in one quarter.
- Data sources and coverage:
  - Quarterly data 1980Q1–2007Q4 from OECD Economic Outlook and IMF IFS.
  - OECD cyclical adjustment uses output gap estimates and constructed elasticities.
  - All series seasonally adjusted.

### II.B Results — frequency and timing during downturns (G7 averages and cross-country patterns)
- Sample-wide summary:
  - Downturns occurred in about 26 percent of the sample.
  - Discretionary fiscal stimulus arrived in 21 percent of all downturn quarters (G7).
  - Interest-rate easing arrived in 51 percent of downturn quarters (G7).
- Average arrival lags (quarter of first easing after start of downturn, G7):
  - Cyclically-adjusted primary balance: 2.4
  - Cyclical primary balance: 0.0
  - Cyclically-adjusted revenue: 2.3
  - Cyclically-adjusted current spending: 1.8
  - Capital spending: 3.8
  - Nominal policy interest rate: 0.8
- Arrival lag comparison:
  - Discretionary fiscal stimulus arrived, on average, 2.4 quarters after onset of a downturn, and 1.6 quarters after interest-rate easing.
- Automatic stabilizers:
  - Cyclical primary balance eased in almost all downturns in the quarter of the downturn itself (96 percent of downturn quarters for the G7).
- Cross-country shares of downturn quarters with discretionary fiscal easing (cyclically-adjusted primary balance):
  - G7: 21; Canada: 34; Germany: 3; France: 10; Italy: 19; Japan: 17; UK: 39; US: 26
- Cross-country shares of downturn quarters with nominal policy interest rate easing:
  - G7: 51; Canada: 72; Germany: 44; France: 48; Italy: 53; Japan: 22; UK: 70; US: 48
- Capital spending easing (share of downturn quarters):
  - G7: 13; Canada: 7; Germany: 17; France: 28; Italy: 8; Japan: 22; UK: 13; US: 21
- Arrival lag for capital spending (quarter of first easing):
  - G7: 3.8; Canada: 5.9; Germany: 1.3; France: 8.5; Italy: 0.9; Japan: 1.9; UK: 2.6; US: 5.4
- Regional pattern:
  - Discretionary fiscal easing more frequent in Anglo-Saxon countries (Canada, the United Kingdom, and the United States): Canada 34 percent, UK 39 percent, US 26 percent.
  - Average of 12 percent in the other G7 countries.
  - Difference attributable to more frequent easing in both revenues and current spending in Anglo-Saxon countries.
- Memorandum items:
  - Number of downturn quarters (G7): 204
  - Total number of quarters in sample: 784
  - Proportion of time spent in downturn (percent, G7): 26

### II.B Results — upturns (comparison)
- Upturns occurred in 74 percent of the sample.
- Fiscal and monetary behavior in upturns:
  - Fiscal policy displays similar cyclical behavior as during downturns.
  - Monetary policy responds noticeably less frequently in upturns (only in 27 percent of upturns) with a longer delay (2.2 quarters) than during downturns.
  - Monetary policy provided much quicker and more frequent easing during downturns than fiscal policy, but during upturns its performance is similar to that of fiscal policy.
- Quarter of first tightening after start of upturn (G7):
  - Cyclically-adjusted primary balance: 2.1
  - Cyclical primary balance: 0.5
  - Cyclically-adjusted revenue: 1.9
  - Cyclically-adjusted current spending: 1.9
  - Capital spending: 2.9
  - Nominal policy interest rate: 2.2
- Share of upturn quarters with tightening (G7):
  - Cyclically-adjusted primary balance: 24
  - Cyclical primary balance: 68
  - Cyclically-adjusted revenue: 23
  - Cyclically-adjusted current spending: 14
  - Capital spending: 18
  - Nominal policy interest rate: 27
- Memorandum items:
  - Number of upturn quarters (G7): 580
  - Total number of quarters in sample: 784
  - Proportion of time spent in upturn (percent, G7): 74

### III. VAR analysis — methodology
- VAR specification:
  - Quarterly, country-level, two lags.
  - Variables and ordering:
    1. Actual real GDP growth minus potential real GDP growth
    2. Inflation (GDP deflator)
    3. Changes in the nominal interest rate
    4. Changes in the primary cyclically-adjusted fiscal balance
    5. Changes in the automatic (cyclical) fiscal balance
- Focus: response of fiscal policy variables to a growth shock (not the response of growth to fiscal policy shocks).
- Robustness checks:
  - Changing ordering of fiscal and monetary variables has no effect on results as long as they are ordered after growth and inflation.
  - Including change in public debt-to-GDP ratio does not change qualitative results.

### III.B Baseline VAR results — key findings
- Impulse response benchmark: responses to a one percentage fall of growth below potential.
- Sample periods shown:
  - Entire sample: 1980Q1–2007Q4
  - Early sample: 1980Q1–1991Q4
  - Late sample: 1992Q1–2007Q4
- Main results:
  - Discretionary fiscal policy in G7 countries has on average responded weakly to downturns.
  - Discretionary fiscal easing is much weaker and slower to arrive than automatic stabilizers and interest-rate easing.
- Early vs. late sample:
  - Early sample: discretionary fiscal policy is pro-cyclical on impact and provides a cumulative pro-cyclical contraction of 0.05 percentage points of trend GDP over four quarters.
  - Late sample: discretionary policy provides no stimulus on impact but provides a cumulative stimulus of 0.2 percentage points over four quarters.
- Composition of fiscal responses:
  - Early sample procyclicality driven by mildly pro-cyclical revenue increases, small counter-cyclical current spending increases, and large pro-cyclical capital spending cuts.
  - Late sample improvement due to cuts in revenues, larger increases in current spending, and much smaller cuts in capital spending.
- Cross-country group differences (first four quarters, late sample):
  - Anglo-Saxon countries: discretionary fiscal policy eased on impact and provided a total stimulus of 0.8 percentage points of trend GDP over the first four quarters.
  - Continental European countries: discretionary fiscal policy was pro-cyclical, with a cumulative tightening of 0.4 percentage points of GDP during the first four quarters.
  - Japan: provided a pro-cyclical impulse on impact, but discretionary policy easing in the subsequent three quarters gave cumulative easing of 0.3 percentage points of trend GDP.
- Monetary policy response:
  - Anglo-Saxon and continental European countries: monetary policy provided timely and strongly countercyclical stimulus, with interest rates falling on impact and a cumulative easing of around 100 basis points after 4 quarters.
  - Japan: almost no nominal interest rate response due to the zero lower bound.
- Robustness to cyclical indicator:
  - Responses to a fall in growth below potential, a fall in growth, and a fall in the output gap were considered; qualitative results are robust.

### Responses of Monetary and Fiscal Policy (figures summary)
- Monetary policy:
  - Consistently timely and strongly countercyclical in downturns across measures.
  - Cumulative four-quarter nominal interest rate responses to a one percentage-point fall in growth below potential shown for 1980-2007, 1980-1991, and 1992-2007 samples.
- Fiscal policy (conventional indicators):
  - C-A Revenue, C-A Primary Current Spending (CAX), C-A Primary Capital Spending (CAK), Cyclical Primary Balance (CPB), and C-A Primary Balance (CAPB) show heterogeneous responses across G7 members and samples.
  - The cyclically adjusted primary balance is calculated from its components.

### Asymmetry in Policy Responses
- Concern: policymakers may ease in downturns and not tighten sufficiently in upturns.
- Empirical findings (cumulative four-quarter responses, 1992Q1–2007Q4):
  - Expansionary bias in both discretionary fiscal policy and monetary policy.
  - Monetary policy: easing during downturns has been more pronounced than tightening during upturns.
  - Discretionary fiscal policy: tendency to provide a substantial stimulus in downturns that is not offset by tightening in upturns.
  - Cross-region behavior:
    - European countries are symmetrically pro-cyclical.
    - Anglo-Saxon countries display strongly asymmetric behavior with large easing in both spending and revenues during downturns.
  - Automatic stabilizers respond symmetrically: easing in downturns is almost exactly offset by tightening during upturns.
- Robustness: results hold when changing ordering of downturns and upturns in the VAR, alternative orderings for fiscal and monetary variables, and including a time trend.

### Policy in Real Time and Growth Estimate Reliability
- Preliminary estimates and revisions:
  - Preliminary growth estimates display a strong negative relationship with subsequent revisions.
  - Of all preliminary estimates indicating negative quarter-over-quarter growth, 39 percent were subsequently revised to positive growth.
  - 30 percent of quarters that ultimately had negative growth showed positive growth in preliminary estimates.
- Forecast efficiency and bias:
  - Strong evidence of a bias towards pessimism in preliminary growth estimates.
  - Average bias for G7 countries estimated at 0.34 percentage points (unannualized) and highly statistically significant.
  - Example: an initial quarterly growth estimate of -0.6 percent is, on average, revised upwards by 0.6 percentage points to zero.
- Errors in identifying negative growth (p (Type-One Error)):
  - G7 38.9; Canada 40.7; France 25.0; Germany 40.0; Italy 42.9; Japan 46.4; U.K. 46.9; U.S. 30.0
- p (Type-Two Error):
  - G7 29.6; Canada 27.3; France 40.0; Germany 30.0; Italy 23.8; Japan 31.8; U.K. 21.2; U.S. 33.3
- Revisions, Full Sample regression (Revision = α + βPrelim + u) — selected coefficients and observations:
  - G7: α 0.341***; β -0.427***; Observations 891
  - Canada: α 0.428***; β -0.373***; Observations 160
  - France: α 0.122*; β -0.16; Observations 69
  - Germany: α 0.238***; β -0.462***; Observations 100
  - Italy: α 0.299***; β -0.610***; Observations 110
  - Japan: α 0.278***; β -0.411***; Observations 114
  - U.K.: α 0.442***; β -0.520***; Observations 160
  - U.S.: α 0.148**; β -0.11; Observations 160
- Revisions, 1995Q1-2004Q4 — selected coefficients and observations:
  - G7: α 0.262***; β -0.421***; Observations 280
  - Canada: α 0.145**; β -0.02; Observations 40
  - France: α 0.147**; β -0.20; Observations 40
  - Germany: α 0.240**; β -0.791***; Observations 40
  - Italy: α 0.169*; β -0.36; Observations 40
  - Japan: α 0.130; β -0.327**; Observations 40
  - U.K.: α 0.377***; β -0.408***; Observations 40
  - U.S.: α 0.10; β -0.186*; Observations 40
- Notes on significance and estimation:
  - Robust t statistics in brackets; ***, **, * denote significance at 1 percent, 5 percent, 10 percent respectively; individual regressions use Newey-West (1987) standard errors; G7 panel uses fixed effects with robust standard errors.
- Policy responses to perceived (erroneously estimated) downturns:
  - Augmented VAR with preliminary estimation errors shows:
    - In the G7 on average, interest rates were cut by around 50 basis points in response to a one percentage point fall in perceived growth relative to final revised growth.
    - Discretionary fiscal balance-to-potential GDP ratio falls by around 0.3 percentage points in response to erroneously perceived downturns, driven by cuts in revenues and increases in current spending.
    - Anglo-Saxon countries drive the interest-rate easing response and provide the largest fiscal stimulus; discretionary fiscal policy is also eased in continental Europe and Japan.
  - Implication: timely and sizable stimulus in Anglo-Saxon countries has often responded to erroneously perceived downturns, raising concerns because fiscal decisions are harder to reverse and errors can have long-lived debt consequences.

### IV. Case Study: U.S. Tax Cuts Timeliness and Temporariness
- Data source and classification:
  - Romer and Romer (2007a) database of legislated tax changes 1945–2007; classification by motivation: (i) exogenous long-run; (ii) exogenous spending-driven; (iii) endogenous “deficit-driven;” (iv) endogenous countercyclical.
  - Romer and Romer single out 5 tax changes motivated explicitly by countercyclical objectives; Annex adds the 2008 Stimulus Package.
- Timeliness:
  - Four out of five cyclically-motivated tax cuts occurred within one quarter of a downturn.
  - 2002 stimulus arrived three quarters after the downturn.
  - Average implementation lag (signed to effective) approximately one quarter.
  - Table 4 selected timeliness lags:
    - 1970Q1 Tax Reform: Lag 1.2 quarters.
    - 1975Q2 Tax Reduction: Lag 3.6 quarters.
    - 1977Q3 Tax Reduction and Simplification: Lag 1.0 quarter.
    - 2001Q3 Economic Growth and Tax Relief Reconciliation: Lag 1.7 quarters.
    - 2002Q2 Job Creation and Worker Assistance: Lag 1.1 quarters.
    - 2008Q2 Economic Stimulus: Lag 1.1 quarters.
  - Mean timeliness measure shown as 1.6.
- Temporariness and size:
  - Using Romer-and-Romer and governmental documents:
    - 84 percent of the average stimulus provided by tax changes was scheduled to be temporary, mostly due to rebates (which account for 52 percentage points of the total).
    - Only 73 percent of tax changes turned out to be temporary because some non-rebate tax changes were subsequently prolonged.
    - Average planned duration of tax changes: 2.5 quarters (rebates average duration 1 quarter; nonrebates average duration 4.6 quarters).
- Table 5 selected U.S. legislated tax-change entries (sizes and lags):
  - 1975 Tax Reduction Act: Size -58.1 $ bn, -3.6 % GDP; signed 3/29/1975; effective 1975Q2; lag 1 quarter; description includes Rebate.
  - 2001 Economic Growth and Tax Relief Reconciliation Act: Size -171.0 $ bn, -1.7 % GDP; signed 6/7/2001; effective 2001Q3; lag 1 quarter.
  - 2002 Job Creation and Worker Assistance Act: Size -110.7 $ bn, -1.1 % GDP; signed 3/9/2002; effective 2002Q2; lag 1 quarter.
  - 2008 Stimulus Act: Size -152.0 $ bn, -1.1 % GDP; signed 2/13/2008; effective 2008Q2; lag 1 quarter.
  - Average reported lag in Table 5: 0.9 quarters.
- Case study conclusion:
  - Countercyclical tax cuts in the U.S. were often timely (many within one quarter) and largely intended to be temporary, but actual temporariness was somewhat lower than planned.

### V. Overall conclusions
- Monetary policy:
  - Reliable countercyclical tool in the G7: timely and strongly countercyclical in downturns.
- Fiscal policy:
  - More nuanced:
    - Discretionary fiscal actions mostly delayed and pro-cyclical in continental European countries and Japan.
    - Discretionary fiscal actions generally countercyclical and more timely in Anglo-Saxon countries.
- Implementation caveats and risks:
  - Policy actions exhibit asymmetry over the cycle (easing bias in downturns not fully offset in upturns).
  - Policy actions respond to measurement errors of the cycle; preliminary growth estimates are biased toward pessimism, leading to potential erroneous discretionary responses.
  - Fiscal policy errors are particularly costly because fiscal decisions are less easily reversed and errors can have long-lived implications for public debt.

*Source: IMF staff estimates, OECD; based on quarterly data 1980Q1–2007Q4 as summarized in the provided chapter.*

### 1. How Often and Quickly did Fiscal Stimulus Arriva During Downturns?.............................5

### 1. How Often and Quickly did Fiscal Stimulus Arriva During Downturns?

### I. Introduction and summary
- Context: The paper examines cyclical behavior of fiscal and monetary policy in the G7 since 1980, addressing conventional wisdom that fiscal policy provides "too little stimulus too late" and may be asymmetric over the cycle.
- Objectives:
  - Use quarterly data and an event study to analyze how often and how quickly fiscal policy responds to a downturn, variation across fiscal instruments, and comparison with monetary policy (Section II).
  - Use a vector auto-regression (VAR) to quantify the strength and speed of fiscal and monetary responses to a growth shock, document withdrawal of fiscal stimulus during recoveries, and compare policy symmetry (Section III).
  - Assess timeliness and temporariness of countercyclical tax legislation for the United States using Romer and Romer (2007) data (Section IV).

### II. Event-study analysis — data and methodology
- Downturn definition:
  - Periods when either the growth rate is negative or the output gap is “unusually” negative (defined as one standard deviation below zero).
- Policy indicators:
  - Monetary policy: change in the short-term nominal interest rate (quarter-over-quarter, unannualized).
  - Fiscal policy:
    - Primary fiscal balance (total general government revenues minus expenditure net of interest).
    - Cyclically-adjusted primary balance (used to identify discretionary fiscal policy).
    - Cyclical primary balance (automatic stabilizers).
  - Discretionary changes defined as “large” changes:
    - Fiscal: exceed 0.25 percent of GDP per quarter.
    - Nominal short-term interest rates: exceed 25 basis points in one quarter.
- Data sources and coverage:
  - Quarterly data 1980Q1–2007Q4 from OECD Economic Outlook and IMF IFS.
  - OECD cyclical adjustment uses output gap estimates and constructed elasticities.
  - All series seasonally adjusted.

### II.B Results — frequency and timing during downturns (G7 averages and cross-country patterns)
- Sample-wide summary:
  - Downturns occurred in about 26 percent of the sample.
  - For the G7 as a whole, discretionary fiscal stimulus arrived in 21 percent of all downturn quarters, less than half as frequently as interest-rate easing (51 percent).
  - Average arrival lags (quarter of first easing after start of downturn, G7):
    - Cyclically-adjusted primary balance: 2.4
    - Cyclical primary balance: 0.0
    - Cyclically-adjusted revenue: 2.3
    - Cyclically-adjusted current spending: 1.8
    - Capital spending: 3.8
    - Nominal policy interest rate: 0.8
  - Arrival lags relative comparison:
    - Discretionary fiscal stimulus arrived, on average, 2.4 quarters after the onset of a downturn, and 1.6 quarters after interest-rate easing.
  - Automatic stabilizers:
    - Cyclical primary balance eased in almost all downturns in the quarter of the downturn itself (96 percent of downturn quarters for the G7).
- Cross-country differences (selected figures from Table 1):
  - Share of downturn quarters with discretionary fiscal easing (cyclically-adjusted primary balance):
    - G7: 21
    - Canada: 34
    - Germany: 3
    - France: 10
    - Italy: 19
    - Japan: 17
    - UK: 39
    - US: 26
  - Share of downturn quarters with nominal policy interest rate easing:
    - G7: 51
    - Canada: 72
    - Germany: 44
    - France: 48
    - Italy: 53
    - Japan: 22
    - UK: 70
    - US: 48
  - Capital spending response:
    - Share of downturn quarters with easing (capital spending): G7: 13; notable country values include Canada: 7; Germany: 17; France: 28; Italy: 8; Japan: 22; UK: 13; US: 21.
  - Arrival lag for capital spending (quarter of first easing):
    - G7: 3.8; country examples: Canada: 5.9; Germany: 1.3; France: 8.5; Italy: 0.9; Japan: 1.9; UK: 2.6; US: 5.4
- Regional pattern:
  - Discretionary fiscal easing was more frequent in “Anglo-Saxon” countries (Canada, the United Kingdom, and the United States):
    - Occurred in 34, 39, and 26 percent of downturn quarters in Canada, the UK, and the US, respectively.
    - Average of 12 percent in the other G7 countries.
  - This difference is attributable to more frequent easing in both revenues and current spending in Anglo-Saxon countries.
- Memorandum items (sample counts and proportions):
  - Number of downturn quarters (G7): 204
  - Total number of quarters in sample: 784
  - Proportion of time spent in downturn (percent, G7): 26

### II.B Results — upturns (comparison)
- Upturns occurred in 74 percent of the sample.
- Fiscal and monetary behavior in upturns:
  - Fiscal policy displays similar cyclical behavior as during downturns.
  - Monetary policy responds noticeably less frequently in upturns (only in 27 percent of upturns) and with a longer delay (2.2 quarters) than during downturns.
  - Monetary policy provided much quicker and more frequent easing during downturns than fiscal policy, but during upturns its performance is similar to that of fiscal policy.
- Selected upturn statistics (Table 2, G7 averages):
  - Quarter of first tightening after start of upturn (G7):
    - Cyclically-adjusted primary balance: 2.1
    - Cyclical primary balance: 0.5
    - Cyclically-adjusted revenue: 1.9
    - Cyclically-adjusted current spending: 1.9
    - Capital spending: 2.9
    - Nominal policy interest rate: 2.2
  - Share of upturn quarters with tightening (G7):
    - Cyclically-adjusted primary balance: 24
    - Cyclical primary balance: 68
    - Cyclically-adjusted revenue: 23
    - Cyclically-adjusted current spending: 14
    - Capital spending: 18
    - Nominal policy interest rate: 27
  - Memorandum items:
    - Number of upturn quarters (G7): 580
    - Total number of quarters in sample: 784
    - Proportion of time spent in upturn (percent, G7): 74

### III. VAR analysis — methodology
- VAR specification (quarterly, country-level, two lags):
  - Variables and ordering:
    1. Actual real GDP growth minus potential real GDP growth
    2. Inflation (GDP deflator)
    3. Changes in the nominal interest rate
    4. Changes in the primary cyclically-adjusted fiscal balance
    5. Changes in the automatic (cyclical) fiscal balance
- Focus: the response of fiscal policy variables to a growth shock (not the response of growth to fiscal policy shocks).
- Robustness: Changing ordering of fiscal and monetary variables has no effect on results as long as they are ordered after growth and inflation. Including change in public debt-to-GDP ratio does not change qualitative results.

### III.B Baseline VAR results — key findings
- Impulse response benchmark: responses to a one percentage fall of growth below potential.
- Samples reported:
  - Entire sample: 1980Q1–2007Q4 (solid line in figures).
  - Early sample: 1980Q1–1991Q4 (dashed line).
  - Late sample: 1992Q1–2007Q4 (dotted line).
- Main results (G7 averages and groups):
  - Discretionary fiscal policy in G7 countries has on average responded weakly to downturns.
  - Across all samples, discretionary fiscal easing is much weaker and slower to arrive than automatic stabilizers and interest-rate easing.
  - Comparison early vs. late sample:
    - Early sample: discretionary fiscal policy is pro-cyclical on impact and provides a cumulative pro-cyclical contraction of 0.05 percentage points of trend GDP over four quarters.
    - Late sample: discretionary policy provides no stimulus on impact but provides a cumulative stimulus of 0.2 percentage points over four quarters.
  - Composition of fiscal responses:
    - Early sample procyclicality driven by mildly pro-cyclical revenue increases, small counter-cyclical current spending increases, and large pro-cyclical capital spending cuts.
    - Late sample improvement attributable to cuts in revenues, larger increases in current spending, and much smaller cuts in capital spending.
  - Cross-country group differences (first four quarters, late sample):
    - Anglo-Saxon countries: discretionary fiscal policy eased on impact and provided a total stimulus of 0.8 percentage points of trend GDP over the first four quarters.
    - Continental European countries: discretionary fiscal policy was pro-cyclical, with a cumulative tightening of 0.4 percentage points of GDP during the first four quarters.
    - Japan: provided a pro-cyclical impulse on impact, but discretionary policy easing in the subsequent three quarters gave cumulative easing of 0.3 percentage points of trend GDP.
  - Monetary policy response:
    - Anglo-Saxon and continental European countries: monetary policy provided timely and strongly countercyclical stimulus, with interest rates falling on impact and a cumulative easing of around 100 basis points after 4 quarters.
    - Japan: almost no nominal interest rate response due to the zero lower bound.
- Robustness to cyclical indicator:
  - Responses to a fall in growth below potential, a fall in growth, and a fall in the output gap were considered.
  - Despite some differences (especially for responses to a fall in the output gap), the qualitative results are robust.

*Source: IMF staff estimates, OECD; based on quarterly data 1980Q1–2007Q4 as summarized in the provided chapter.*

### conclusions regarding the cyclical behavior of monetary and fiscal policy are robust to the

### _wp0950 - conclusions regarding the cyclical behavior of monetary and fiscal policy are robust to the choice of the cyclical indicator.

### Responses of Monetary and Fiscal Policy (summary of Figures 1–3)
- Monetary policy:
  - Has been consistently timely and strongly countercyclical in downturns across a range of measures.
  - Cumulative four-quarter nominal interest rate responses to a one percentage-point fall in growth below potential are shown for 1980-2007, 1980-1991, and 1992-2007 samples (see Figures 1–3 for detailed profiles by horizon and sample).
- Fiscal policy (conventional indicators):
  - C-A Revenue, C-A Primary Current Spending (CAX), C-A Primary Capital Spending (CAK), Cyclical Primary Balance (CPB), and C-A Primary Balance (CAPB) show heterogeneous responses across G7 members and samples (Figures 1–3).
  - The cyclically adjusted primary balance is calculated from its components.1

### Asymmetry in Policy Responses (Section C and Figure 4)
- Concern and definition:
  - Policymakers may respond asymmetrically, easing in downturns and not tightening sufficiently in upturns; downturns defined as quarters in which growth is negative or below trend with the output gap more than one standard deviation below zero.
- Empirical findings:
  - Cumulative four-quarter responses (1992Q1–2007Q4) indicate an expansionary bias in both discretionary fiscal policy and monetary policy.
  - Monetary policy: easing during downturns has been more pronounced than tightening during upturns.
  - Discretionary fiscal policy: tendency to provide a substantial stimulus in downturns that is not offset by tightening in upturns.
  - Cross-region behavior:
    - European countries are symmetrically pro-cyclical.
    - Anglo-Saxon countries display strongly asymmetric behavior with large easing in both spending and revenues during downturns.
  - Automatic stabilizers respond symmetrically: easing in downturns is almost exactly offset by tightening during upturns.
- Robustness:
  - Results robust to changing the ordering of downturns and upturns in the VAR, to alternative orderings for fiscal and monetary variables, and to including a time trend.

### Policy in Real Time and Growth Estimate Reliability (Section D, Figures 5–6, Table 3)
- Preliminary estimates and revisions:
  - Preliminary growth estimates display a strong negative relationship with subsequent revisions (Figure 5).
  - Of all preliminary estimates indicating negative quarter-over-quarter growth, 39 percent were subsequently revised to positive growth (Table 2 referenced in text).
  - 30 percent of quarters that ultimately had negative growth showed positive growth in preliminary estimates.
- Forecast efficiency and bias:
  - Standard forecast efficiency tests find strong evidence of a bias towards pessimism in preliminary growth estimates.
  - Average bias for G7 countries estimated at 0.34 percentage points (unannualized) and highly statistically significant.
  - Example: an initial quarterly growth estimate of -0.6 percent is, on average, revised upwards by 0.6 percentage points to zero.
- Panel and country-specific regression results (Table 3 highlights):
  - Errors in Identifying Negative Growth (p (Type-One Error)): G7 38.9; Canada 40.7; France 25.0; Germany 40.0; Italy 42.9; Japan 46.4; U.K. 46.9; U.S. 30.0
  - p (Type-Two Error): G7 29.6; Canada 27.3; France 40.0; Germany 30.0; Italy 23.8; Japan 31.8; U.K. 21.2; U.S. 33.3
  - Revisions, Full Sample regression (Revision = α + βPrelim + u):
    - G7: α 0.341***; β -0.427***; Observations 891
    - Canada: α 0.428***; β -0.373***; Observations 160
    - France: α 0.122*; β -0.16; Observations 69
    - Germany: α 0.238***; β -0.462***; Observations 100
    - Italy: α 0.299***; β -0.610***; Observations 110
    - Japan: α 0.278***; β -0.411***; Observations 114
    - U.K.: α 0.442***; β -0.520***; Observations 160
    - U.S.: α 0.148**; β -0.11; Observations 160
  - Revisions, 1995Q1-2004Q4:
    - G7: α 0.262***; β -0.421***; Observations 280
    - Canada: α 0.145**; β -0.02; Observations 40
    - France: α 0.147**; β -0.20; Observations 40
    - Germany: α 0.240**; β -0.791***; Observations 40
    - Italy: α 0.169*; β -0.36; Observations 40
    - Japan: α 0.130; β -0.327**; Observations 40
    - U.K.: α 0.377***; β -0.408***; Observations 40
    - U.S.: α 0.10; β -0.186*; Observations 40
  - Notes: Robust t statistics in brackets; ***, **, * denote significance at 1 percent, 5 percent, 10 percent respectively; individual regressions use Newey-West (1987) standard errors; G7 panel uses fixed effects with robust standard errors.
- Policy responses to perceived (erroneously estimated) downturns:
  - Augmented VAR with preliminary estimation errors shows:
    - In the G7 on average, interest rates were cut by around 50 basis points in response to a one percentage point fall in perceived growth relative to final revised growth.
    - Discretionary fiscal balance-to-potential GDP ratio falls by around 0.3 percentage points in response to erroneously perceived downturns, driven by cuts in revenues and increases in current spending.
    - Anglo-Saxon countries drive the interest-rate easing response and provide the largest fiscal stimulus; discretionary fiscal policy is also eased in continental Europe and Japan.
  - Implication: timely and sizable stimulus in Anglo-Saxon countries has often responded to erroneously perceived downturns, raising concerns because fiscal decisions are harder to reverse and errors can have long-lived debt consequences.

### Case Study: U.S. Tax Cuts Timeliness and Temporariness (Section IV, Tables 4–5)
- Data source and classification:
  - Romer and Romer (2007a) database of all legislated tax changes 1945–2007; classifies tax changes by motivation: (i) exogenous long-run; (ii) exogenous spending-driven; (iii) endogenous “deficit-driven;” (iv) endogenous countercyclical.
  - Romer and Romer single out 5 tax changes motivated explicitly by countercyclical objectives; Annex adds the 2008 Stimulus Package.
- Timeliness:
  - Four out of five cyclically-motivated tax cuts occurred within one quarter of a downturn.
  - 2002 stimulus arrived three quarters after the downturn.
  - Average implementation lag (signed to effective) approximately one quarter.
  - Table 4 summary (selected entries):
    - 1970Q1 Tax Reform: Timeliness Lag 1.2 quarters (planned/actual dates and permanence indicated in table).
    - 1975Q2 Tax Reduction: Lag 3.6 quarters.
    - 1977Q3 Tax Reduction and Simplification: Lag 1.0 quarter.
    - 2001Q3 Economic Growth and Tax Relief Reconciliation: Lag 1.7 quarters.
    - 2002Q2 Job Creation and Worker Assistance: Lag 1.1 quarters.
    - 2008Q2 Economic Stimulus: Lag 1.1 quarters.
  - Mean timeliness measure shown as 1.6 (table formatting indicates mean 1.6 across listed episodes).
- Temporariness and size:
  - Using Romer-and-Romer and governmental documents:
    - 84 percent of the average stimulus provided by tax changes was scheduled to be temporary, mostly due to rebates (which account for 52 percentage points of the total).
    - Only 73 percent of tax changes turned out to be temporary because some non-rebate tax changes were subsequently prolonged.
    - Average planned duration of tax changes: 2.5 quarters (rebates average duration 1 quarter; nonrebates average duration 4.6 quarters).
  - Table 5 (selected U.S. legislated tax-change entries show $ bn and % GDP values and signed/effective dates; average lag 0.9 quarters reported).
  - Examples from Table 5:
    - 1975 Tax Reduction Act: Size -58.1 $ bn, -3.6 % GDP (signed 3/29/1975, effective 1975Q2, lag 1 quarter; description: Rebate, tax credits, reductions in standard deduction and increase of ITC).
    - 2001 Economic Growth and Tax Relief Reconciliation Act: Size -171.0 $ bn, -1.7 % GDP (signed 6/7/2001, effective 2001Q3, lag 1 quarter).
    - 2002 Job Creation and Worker Assistance Act: Size -110.7 $ bn, -1.1 % GDP (signed 3/9/2002, effective 2002Q2, lag 1 quarter).
    - 2008 Stimulus Act: Size -152.0 $ bn, -1.1 % GDP (signed 2/13/2008, effective 2008Q2, lag 1 quarter).
- Conclusion from case study:
  - Countercyclical tax cuts in the U.S. were often timely (many within one quarter) and largely intended to be temporary, but actual temporariness was somewhat lower than planned.

### Overall Conclusions (Section V)
- Monetary policy:
  - Confirmed to be a reliable countercyclical tool in the G7: timely and strongly countercyclical in downturns.
- Fiscal policy:
  - More nuanced than common perception:
    - Discretionary fiscal actions have mostly been delayed and pro-cyclical in continental European countries and Japan.
    - Discretionary fiscal actions have generally been countercyclical and more timely in Anglo-Saxon countries.
- Implementation caveats and risks:
  - Policy actions exhibit asymmetry over the cycle (easing bias in downturns not fully offset in upturns).
  - Policy actions are responsive to measurement errors of the cycle; preliminary growth estimates are biased toward pessimism, leading to potential erroneous discretionary responses.
  - Fiscal policy errors are particularly costly because fiscal decisions are less easily reversed and errors can have long-lived implications for public debt.

*Italic: Source — IMF staff analysis and figures/tables as presented in _wp0950 (excerpts provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0950.pdf_