## 1. Methodology

## Source details

**Canonical URL:** [1. Methodology](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1152.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1152.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1152.pdf.json)

---

### Identification of fiscal shocks
- Identification challenge: two directions of causation — (i) government spending could affect output or (ii) output could affect government spending (through automatic stabilizers and implicit or explicit policy rules).
- Two main approaches discussed:
  - Structural vector autoregression approach (SVAR), as in Blanchard and Perotti (2002).
  - "Natural experiment" of large military buildups (Barro (1981); Ramey and Shapiro (1998)), including Ramey (2009) and Barro and Redlick (2009).
- SVAR assumption and critique:
  - Assumes fiscal policy requires some time (assumed to be at least one-quarter) to respond to news about the state of the economy; remaining correlation between unpredicted components of government spending and output is attributed to government spending impacting output.
  - Critique: identified shocks, while unpredicted by the econometrician, may have been known to private agents. The paper revisits predictability of SVAR residuals in section 2.1.
- Natural experiment critique:
  - Military buildups may be caused by wars, which have independent macroeconomic effects.
  - U.S. evidence dominated by WWII and the Korean War, limiting instrument variation.
- Rationale for SVAR in this paper:
  - Military buildup approach impractical for cross-country sample, especially developing countries.
  - Volatility of government consumption in developing countries may make it less predictable to private sector, mitigating the SVAR critique (see section 2.1).

### Estimation methodology
- Model system (following Blanchard and Perotti (2002)):
  - A Y_{n,t} = sum_{k=1}^K C_k Y_{n,t-k} + B u_{n,t}
  - Y_{n,t} is a vector including government expenditure variables (government consumption and/or investment), GDP, and other endogenous variables (current account, real exchange rate, policy interest rate).
  - C_k are lag matrices; B diagonal so u_{n,t} are orthogonal, i.i.d. shocks with E u_{n,t} = 0 and E(u_{n,t} u'_{n,t}) = identity. A allows for simultaneous effects.
  - Matrices A, B, C_k assumed invariant across time and countries.
- Estimation choices and robustness:
  - Primary estimation via panel OLS, yielding estimates of A^{-1} C_k.
  - Additional identification assumptions needed to estimate A and B.
  - Benchmark regressions are bivariate: Y_{n,t} = (g_{n,t}, y_{n,t})' with g_t = government consumption and y_t = output.
  - Identification: changes in government consumption require at least one quarter to respond to innovations in output — equivalent to Cholesky with g_t ordered before y_t (A = [[1,0],[a21,1]]).
  - Pooling across countries due to short country-specific samples (typical country sample ~ ten years, yielding ~ forty observations); pooled sample almost always exceeding one thousand observations.
  - Sample divided for comparisons: high-income versus developing; predetermined versus flexible exchange arrangements; open versus closed economies; among others.
  - Robustness:
    - Mean Group estimator of Pesaran and Smith (1995) yields similar results (power for inference reduced).
    - Results robust to an "international VAR" where endogenous variables of large countries (U.S. or all G7 excluding Japan) used as exogenous inputs.

### Fiscal multipliers: definitions
- Fiscal multiplier: change in real GDP or other output caused by a one-unit increase in a fiscal variable. Example given: "if a one dollar increase in government consumption in the United States caused a fifty cent increase in U.S. GDP, then the government consumption multiplier is0:5."
- Impact Multiplier:
  - Impact Multiplier = y0 / g0
  - Measures ratio of change in output to change in government expenditure at the time of the impulse.
- Cumulative Multiplier (at time T):
  - Cumulative Multiplier(T) = (Σt=0^T yt) / (Σt=0^T gt)
  - Captures cumulative change in output per unit of additional government expenditure from impulse to horizon.
- Long-Run Multiplier defined as cumulative multiplier as T → 1 (T!1 in source).
- Properties:
  - Multipliers may differ greatly across forecast horizons; both impact and cumulative multipliers reported.
  - Cumulative multipliers capture implementation lags and persistence.
- Empirical implementation notes:
  - Main approach uses a Panel VAR (OLS with fixed effects; robust to GLS).
  - K, number of lags: tests suggested optimal K varies from 1 to 8; for comparability K = 4 is used in reported results.
  - Reported results robust to alternative lags 1 to 8; likelihood ratio test rejects exclusion of lags 5 to 8 at the 95% confidence level.

### Data scope and key characteristics
- Dataset:
  - Novel quarterly dataset for 44 countries: 20 high-income and 24 developing.
  - Coverage varies across countries, spanning from 1960:1 to 2007:4.
  - Only data originally collected on a quarterly basis included (no interpolation from annual data).
  - Pooled data contains more than 2,500 observations.
  - Dataset end date: "we ended the dataset with the fourth quarter of 2007."
- De-seasonalization: SEATS algorithm (results similar using X-11 in earlier version).
- Detrending and stationarity:
  - Variables non-stationary except central bank interest rate and current account/GDP ratio.
  - Non-stationary variables used as deviations from quadratic trend (linear trend yielded similar results).
  - After detrending, unit roots rejected at the 99 percent confidence level (ADF and Im, Pesaran and Shin tests).
- Key descriptive statistics (exact values):
  - Government consumption to GDP during sample:
    - Low: 9.6 percent in El Salvador.
    - High: 27.4 percent in Sweden.
  - Averages:
    - High income countries: government consumption averages 20.8 percent of GDP.
    - Developing countries: government consumption averages 15.6 percent of GDP.
  - Autocorrelation (cyclical component, deviations from quadratic trend):
    - High income countries: 0.75.
    - Developing countries: 0.54.
  - Volatility (detrended government consumption relative to GDP):
    - In high-income countries, government consumption is less than twice as volatile as GDP.
    - In developing countries, government consumption is more volatile by a factor of 8.
- Evidence on foreseeability of innovations:
  - Vintage data from central banks (Bulgaria, Ecuador, Uruguay) examined.
  - Central bank preliminary estimate error correlates with VAR residuals; text concludes "VAR residuals are a fairly good measure of unexpected innovations in government consumption."
  - In developing countries, fiscal policy is sufficiently erratic that ex-post estimates are subject to significant revision; VAR residuals capture a significant portion of unanticipated shocks.

### Practical dataset construction notes
- Institutional drivers: ESA95 adoption in 1996 and Eurostat guidance (2006 Manual); IMF SDDS in 1996 encouraged quarterly reporting.
- Inclusion rules: data included only if originally quarterly and covering a large proportion of government expenditure category studied; data integrity confirmed via correspondence with agencies.
- Government level inconsistency: in most cases general government available; some central government only. Reported results unchanged when restricted to general government series.
- Price deflation: sometimes CPI used; correlation of series when deflated using different price indexes was close to1.

### Key empirical motivations (policy debate context)
- Contrasting views on fiscal multipliers:
  - Robert Barro argued peacetime fiscal multipliers were essentially zero.
  - Christina Romer used multipliers as high as 1.6 to estimate job gains from the $787 billion stimulus package approved by Congress in February 2009. The difference between Romer’s and Barro’s views amounts to a staggering 3.7 million jobs by the end of 2010.
- Emphasis on fiscal–monetary interactions as crucial determinant of fiscal policy effects on GDP.

### Selected main results (summary)
- Developing countries:
  - Response of output to increases in government consumption is negative on impact.
  - It is smaller by a statistically significant margin from both zero and the response estimated for high-income countries.
  - Output increases only with a lag (2 to 4 quarters); cumulative response not statistically different from zero.
  - Increases in government consumption are far more transient (dying out after approximately 6 quarters) versus highly persistent shocks in high-income countries.
- Exchange rate flexibility:
  - Degree of exchange rate flexibility is a critical determinant of fiscal multipliers.
  - Predetermined exchange rate regimes: long-run multipliers larger than one in some specifications.
  - Flexible exchange rate regimes: essentially zero multipliers.
  - Fiscal multiplier in predetermined exchange rates statistically different from zero and from flexible countries at any forecast horizon.
  - Main difference arises from degree of monetary accommodation to fiscal shocks.
- Openness to trade:
  - Closed economies have long-run multipliers around 1.3 to 1.4.
  - Relatively open economies have negative multipliers.
  - In economies with large trade-to-GDP ratios the multiplier is statistically different from zero and from the multiplier in open economies at any forecast horizon.
- High public debt:
  - When central government debt exceeded 60 percent of GDP, the fiscal multiplier was not statistically different from zero on impact and was negative (and statistically different from zero) in the long run.
  - The 60% of GDP threshold identified as critical above which fiscal stimulus may have a negative long-run impact on output.
- Government investment vs. consumption:
  - No general finding that multiplier on government investment is significantly higher than government consumption in most groups.
  - Exception: developing countries, where the multiplier on government investment is positive, close to 1 in the medium term, and statistically different from the multiplier on government consumption at forecast horizons of up to two years.
  - Point estimate of investment multiplier larger than consumption in high-income countries, but difference small and not statistically significant.

### Selected quantitative results (exact figures preserved)
- High-income vs developing (responses to a 1 percent shock to government consumption at time 0):
  - Impact response of output:
    - High-income countries: 0.08 percent (positive).
    - Developing countries: -0.03 percent (negative).
    - Both statistically significant from zero and from each other at the 99% confidence level.
  - Impact multipliers:
    - High-income countries: impact multiplier = 0.37.
    - Developing countries: impact multiplier = -0.21.
  - Cumulative multipliers (horizon 0 to 20 quarters; Monte Carlo 90% confidence intervals with 500 repetitions):
    - High-income countries: long-run cumulative multiplier = 0.80.
    - Developing countries: long-run cumulative multiplier = 0.18.
- Exchange rate regimes:
  - Predetermined exchange rates:
    - Impact multiplier = 0.09 (statistically significantly different from zero).
    - Long-run cumulative multiplier = 1.5.
  - Flexible exchange rates:
    - Impact multiplier: negative and statistically significant on impact.
    - Long-run multiplier: statistically indistinguishable from zero.
- Monetary accommodation (multivariate VAR; ordering government consumption before policy instrument):
  - Under predetermined exchange rates: short-term nominal interest rate declines by a cumulative 125 basis points in the two years following a government consumption shock of 1% of GDP.
  - Under flexible exchange rates: central banks increase the policy interest rate by an average of 60 basis points within the two years following a similar fiscal shock.
- Openness and size:
  - Trade openness (trade/GDP > 60%): impact response 0.11; long-run multiplier 1.4.
  - Alternative openness (tariff-based):
    - Open economies: impact -0.28; long run -0.75.
    - Closed economies: impact 0.02; long run 1.29.
  - Size: Large economies: impact multiplier 0.02; long-run 1.2. Small economies: impact -0.19; long-run -0.47.
- Financial fragility (high-debt episodes; debt > 60% of GDP for 3+ consecutive years):
  - Impact multiplier very close to zero.
  - Long-run multiplier estimated at -2.3.
  - Long-run multiplier remains negative when threshold set to 60 or 70 percent of GDP; becomes positive for debt-to-GDP ratios of 30 or 40 percent.
- Government investment:
  - Developing countries: impact multiplier of 0:6 (statistically significant).
  - Pure government investment (cross-dimension examples):
    - Predetermined exchange rates: impact 0.36; long run 1.42.
    - Flexible exchange rates: impact 0.46; long run 0.16.
    - High trade-to-output ratios: impact 0.51; long run -0.23.
    - Low trade-to-output ratios: impact 0.46; long run 0.70.
- Selected figure/model reported statistics (as presented in source):
  - Figure 12 (Cumulative government investment multiplier, High-income countries; includes indirect effects of government consumption): Long Run: 2.5; Impact: 0.38.
  - Figure 17 (Cumulative multiplier — high income and developing countries, Multivariate Regression): Long Run: 0.85; Impact: 0.39. Additional panel: Long Run: 0.53; Impact: -0.09.
  - Figure 20 (Cumulative multiplier: Highly indebted countries, Multivariate Regression): Long Run: -4.7; Impact: -0.16.

### Monetary accommodation and exchange rate regimes (detailed)
- Ordering government consumption before the central bank’s instrument follows assumption monetary authority can respond more rapidly than fiscal decision-makers.
- Results virtually unchanged if policy interest rate ranked lower in Cholesky ordering; reversing ordering weakens policy interest rate response.
- Strong evidence for a "monetary accommodation" channel:
  - Fixed exchange rates: short-term nominal interest rate declines by cumulative 125 basis points in two years after a 1% of GDP government consumption shock.
  - Flexible exchange rates: policy interest rate increases by average 60 basis points in two years after similar shock.
- Differences in monetary accommodation identified as main cause for differences in fiscal multipliers across exchange rate regimes.
- Current account response does not differ significantly across regimes; it deteriorates following a government consumption shock under both fixed and flexible regimes.

### Consumption versus investment responses
- Private investment: similar under fixed and flexible regimes — erratic, with investment declining significantly on impact and additional dips later.
- Private consumption:
  - Fixed exchange rates: consumption responds positively on impact and significantly.
  - Flexible exchange rates: response not statistically different from zero; point estimates show negative response short and long run.
- Interpretation: when central bank accommodates fiscal shock, private consumption responds positively; when it does not, consumption may decline.

### Multivariate regressions and robustness
- Expanded VAR (real government consumption, real GDP, current account/GDP, real effective exchange rate) with ordering government consumption before GDP; current account before real effective exchange rate.
- Point estimates for cumulative multipliers similar to bivariate regressions; error bands wider due to fewer observations.
- Negative multiplier in highly indebted economies robust and strengthened with additional controls.

### Policy implications and recommendations
- Monetary accommodation crucially determines expansionary effect of fiscal policy; fiscal multipliers larger when monetary policy accommodates.
- Fiscal stimulus likely ineffective or counter-productive in highly indebted countries (debt-to-GDP around or above 60 percent).
- In economies open to trade or with flexible exchange rates, fiscal expansions yield little or no significant output gains.
- Composition matters, especially in developing countries: government investment can have more positive short-run impact on GDP than government consumption.
- Emerging countries with procyclical fiscal policies should pursue a-cyclical fiscal policy (government consumption and tax rates not responding to the business cycle) to reduce probability of fiscal crises.

*Source: _wp1152 - 1. Methodology (section and introduction material) — IMF PDF content provided.*

### 1. Methodology .........................................................................................................

### 1. Methodology

### 1.1 Identification of Fiscal Shocks
- Identification problem: two directions of causation — (i) government spending could affect output or (ii) output could affect government spending (through automatic stabilizers and implicit or explicit policy rules).
- Two main approaches discussed:
  - Structural vector autoregression approach (SVAR), as in Blanchard and Perotti (2002).
  - "Natural experiment" of large military buildups (Barro (1981); Ramey and Shapiro (1998)), including Ramey (2009) and Barro and Redlick (2009).
- SVAR basic assumption: fiscal policy requires some time (assumed to be at least one-quarter) to respond to news about the state of the economy. After using a VAR to eliminate predictable responses, remaining correlation between unpredicted components of government spending and output is attributed to government spending impacting output.
- Critique of SVAR: identified shocks, while unpredicted by the econometrician, may have been known to private agents. The paper revisits predictability of SVAR residuals in section 2.1.
- Natural experiment approach critiques:
  - Military buildups may be caused by wars, which have independent macroeconomic effects.
  - In the United States, two buildups (WWII and the Korean War) dominate, limiting practical instrument variation.
- Rationale for using SVAR in this paper:
  - Military buildup approach is impractical for cross-country sample, especially developing countries where wars can have direct devastating macroeconomic effects and would confound estimates.
  - Volatility of government consumption in developing countries may make it less predictable to private sector, mitigating the standard SVAR critique (see section 2.1).

### 1.2 Estimation Methodology
- Estimating system (following Blanchard and Perotti (2002)):
  - AY_{n,t} = sum_{k=1}^K C_k Y_{n,t-k} + B u_{n,t}
  - Y_{n,t} is a vector of variables: government expenditure variables (government consumption and/or investment), GDP, and other endogenous variables (current account, real exchange rate, policy interest rate).
  - C_k are matrices of lag effects; B is diagonal so u_t is vector of orthogonal, i.i.d. shocks with E u_{n,t} = 0 and E(u_{n,t} u'_{n,t}) = identity.
  - A allows for simultaneous effects between endogenous variables.
  - Matrices A, B, C_k are assumed invariant across time and countries.
- Estimation choices and robustness:
  - Primary estimation via panel OLS, yielding estimates of A^{-1} C_k.
  - Additional identification assumptions needed to estimate A and B.
  - Benchmark regressions are bivariate: Y_{n,t} = (g_{n,t}, y_{n,t})' with g_t = government consumption and y_t = output.
  - Identification assumption: changes in government consumption require at least one quarter to respond to innovations in output. Equivalent to a Cholesky decomposition with g_t ordered before y_t (A takes form [[1,0],[a21,1]]).
  - The authors pool data across countries rather than estimate country-by-country due to short country-specific samples (typical country sample ~ ten years, yielding ~ forty observations).
  - Pooling produces a larger sample size — almost always exceeding one thousand observations.
  - Sample divided for comparisons across groups: high-income versus developing; predetermined versus flexible exchange arrangements; open versus closed economies; among others.
  - Results are also robust to alternative specifications:
    - Mean Group estimator of Pesaran and Smith (1995) yields similar results (though power for inference is significantly reduced).
    - Results robust to an "international VAR" specification where endogenous variables of large countries (either the U.S. or all G7 countries, excluding Japan) are used as exogenous inputs to other countries' estimating equations.
- Practical dataset construction (context relevant to methodology):
  - Novel quarterly dataset for 44 countries: 20 high-income and 24 developing.
  - Coverage varies across countries, spanning from as early as 1960:1 to as late as 2007:4.
  - Only data originally collected on a quarterly basis was included (no interpolation from annual data).

### Key empirical motivations and context (selected findings summarized in the methodology section)
- Policy debate context:
  - Disagreement about fiscal multipliers illustrated by contrasting views: Robert Barro argued peacetime fiscal multipliers were essentially zero; Christina Romer used multipliers as high as 1.6 to estimate job gains from the $787 billion stimulus package approved by Congress in February 2009. The difference between Romer’s and Barro’s views amounts to a staggering 3.7 million jobs by the end of 2010.
- Main results (as summarized in the Methodology/Introduction):
  1. In developing countries:
     - Response of output to increases in government consumption is negative on impact.
     - It is smaller by a statistically significant margin from both zero and the response estimated for high-income countries.
     - Output increases in response to a shock in government consumption only with a lag (of 2 to 4 quarters) and the cumulative response of output is not statistically different from zero.
     - Increases in government consumption are far more transient (dying out after approximately 6 quarters) versus highly persistent shocks in high-income countries.
  2. Exchange rate flexibility:
     - Degree of exchange rate flexibility is a critical determinant of fiscal multipliers.
     - Economies with predetermined exchange rate regimes have long-run multipliers that are larger than one in some specifications.
     - Economies with flexible exchange rate regimes have essentially zero multipliers.
     - The fiscal multiplier in countries with predetermined exchange rates is statistically different from zero and from the multiplier in flexible countries at any forecast horizon.
     - Main difference arises from degree of monetary accommodation to fiscal shocks; central bank response to fiscal shocks is crucial.
  3. Openness to trade:
     - Closed economies (due to trade barriers or larger internal markets) have long-run multipliers around 1.3 to 1.4.
     - Relatively open economies have negative multipliers.
     - In economies with large trade-to-GDP ratios the multiplier is statistically different from zero and from the multiplier in open economies at any forecast horizon.
     - The multiplier in open economies is negative and significantly lower than zero both on impact and in the long run.
  4. High public debt:
     - During episodes where outstanding debt of the central government exceeded 60 percent of GDP, the fiscal multiplier was not statistically different from zero on impact and was negative (and statistically different from zero) in the long run.
     - Experimentation indicates the 60% of GDP threshold is a critical value above which fiscal stimulus may have a negative long-run impact on output.
  5. Government investment vs. consumption:
     - No general finding that the multiplier on government investment is significantly higher than government consumption in most country groupings.
     - Exception: developing countries, where the multiplier on government investment is positive, close to 1 in the medium term, and statistically different from the multiplier on government consumption at forecast horizons of up to two years.
     - Point estimate of the investment multiplier is larger than consumption in high-income countries as well, but difference is small and not statistically significant.
- Policy implication noted in methodology/introduction:
  - Increasing trade integration and adoption of flexible exchange rate arrangements (particularly inflation-targeting regimes) cast doubt on the effectiveness of fiscal stimuli.
  - Fiscal stimuli likely to become weaker, and potentially yield negative multipliers, because a large number of countries now carry very high public debt ratios.
  - Findings highlight the importance of fiscal–monetary interactions as a crucial determinant of fiscal policy effects on GDP.

*Source: _wp1152 - 1. Methodology (section and introduction material) — IMF PDF content provided.*

### 1.3  Fiscal Multipliers: Denitions

### 1.3  Fiscal Multipliers: Definitions

### Definitions and measurement
- Fiscal multiplier: the change in real GDP or other measure of output caused by a one-unit increase in a fiscal variable.
  - Example in text: "if a one dollar increase in government consumption in the United States caused a fifty cent increase in U.S. GDP, then the government consumption multiplier is0:5."
- Impact Multiplier (defined):
  - Impact Multiplier = y0 / g0
  - Measures the ratio of the change in output to a change in government expenditure at the time in which the impulse to government expenditure occurs.
- Cumulative Multiplier (at time T) (defined):
  - Cumulative Multiplier(T) = (Σt=0^T yt) / (Σt=0^T gt)
  - Measures the cumulative change in output per unit of additional government expenditure from the time of the impulse to the reported horizon.
- Long-Run Multiplier:
  - Defined as the cumulative multiplier as T → 1 (T!1 in source).

### Properties emphasized
- Multipliers may differ greatly across forecast horizons; hence both impact and cumulative multipliers are reported.
- Cumulative multipliers capture implementation lags and persistence of responses over time.

### Empirical implementation notes (from source)
- Main empirical approach uses a Panel VAR (OLS regression with fixed effects; results robust to GLS allowing for different cross-sectional weights).
- K, the number of lags in system (1): tests suggested optimal K varies from 1 to 8; for comparability K = 4 is used in reported results.
- Reported results are robust to choosing alternative numbers of lags from 1 to 8; likelihood ratio test rejects exclusion of lags 5 to 8 at the 95% confidence level (implying reported results using only 4 lags may understate effects).

---

### Data scope and key data characteristics
- Sample coverage and construction:
  - Cross-section of 44 countries, of which 24 are developing countries (based on World Bank income classifications).
  - Pooled data contains more than 2,500 observations.
  - Dataset end date: "we ended the dataset with the fourth quarter of 2007."
- Quarterly government consumption: first broad catalogue of available quarterly data on government consumption across many countries.
- Institutional drivers of quarterly data availability:
  - ESA95 adoption in 1996 and Eurostat guidance (2006 Manual) leading to quarterly classification.
  - IMF Special Data Dissemination Standard (SDDS) in 1996 encouraging quarterly reporting for some subscribers.
- De-seasonalization: SEATS algorithm (results similar using X-11 in earlier version).
- Detrending and stationarity:
  - Variables non-stationary except central bank interest rate and current account/GDP ratio.
  - Non-stationary variables used as deviations from quadratic trend (linear trend yielded similar results).
  - After detrending, unit roots rejected at the 99 percent confidence level (ADF and Im, Pesaran and Shin tests).

### Key descriptive statistics (exact values as reported)
- Government consumption to GDP during sample:
  - Low: 9.6 percent in El Salvador.
  - High: 27.4 percent in Sweden.
- Averages:
  - High income countries: government consumption averages 20.8 percent of GDP.
  - Developing countries: government consumption averages 15.6 percent of GDP.
- Autocorrelation (cyclical component, deviations from quadratic trend):
  - High income countries: 0.75.
  - Developing countries: 0.54.
- Volatility (detrended government consumption relative to GDP):
  - In high-income countries, government consumption is less than twice as volatile as GDP.
  - In developing countries, government consumption is more volatile by a factor of 8.
- Note on interpolation risk: interpolating annual government expenditure data using quarterly tax revenue seasonal patterns can create spurious procyclical correlation between government expenditure and output.

### Evidence on whether innovations to government consumption are foreseen
- Vintage data from central banks (Bulgaria, Ecuador, Uruguay) examined.
- Central bank preliminary estimate error (difference between final published data and first published official estimate) correlates with VAR residuals.
  - Source text conclusion: "VAR residuals are a fairly good measure of unexpected innovations in government consumption."
- In developing countries, fiscal policy is sufficiently erratic that ex-post estimates are subject to significant revision; evidence is suggestive that VAR residuals capture a significant portion of unanticipated shocks to government consumption.

---

### 3  Results (selected empirical findings)

### High-income vs developing countries (Panel VAR bivariate: government consumption and GDP)
- Impulse-response summary (responses to a 1 percent shock to government consumption at time 0):
  - Impact response of output:
    - High-income countries: 0.08 percent (positive).
    - Developing countries: -0.03 percent (negative).
    - Both are statistically significant from zero and from each other at the 99% confidence level.
  - Persistence:
    - High-income countries: output response remains significantly positive for the 20 quarters shown.
    - Developing countries: output response becomes zero (statistically) after six quarters.
- Impact multipliers (computed from impulse responses):
  - High-income countries: impact multiplier = 0.37.
    - Interpretation from source: "an additional dollar of government spending will deliver only 37 cents of additional output in the quarter in which it is implemented."
  - Developing countries: impact multiplier = -0.21.
  - Difference between impact multipliers for the two groups is statistically significant at the 99 percent confidence level.
- Cumulative multipliers (horizon 0 to 20 quarters; Monte Carlo 90% confidence intervals with 500 repetitions):
  - High-income countries:
    - Impact: 0.37 (matches impact multiplier above).
    - Long-run cumulative multiplier: 0.80.
    - Interpretation: even in the long run output has risen less than the cumulative increase in government consumption (some crowding out at every horizon).
  - Developing countries:
    - Long-run cumulative multiplier: 0.18.
    - Interpretation: in the long run, over four fifths of the increase in government consumption is crowded out by other GDP components.

### Exchange rate regimes
- Sample partitioned into predetermined (fixed) exchange rate episodes and flexible exchange rate regimes (de facto classification of Ilzetzki, Reinhart, and Rogoff (2008) used).
- Cumulative multiplier findings by regime:
  - Predetermined exchange rates:
    - Impact multiplier = 0.09 (statistically significantly different from zero).
    - Long-run cumulative multiplier = 1.5.
  - Flexible exchange rates:
    - Impact multiplier: negative and statistically significant on impact.
    - Long-run multiplier: statistically indistinguishable from zero.
  - Difference between the two regimes is statistically significant at every forecast horizon.
- Interpretation (consistency with Mundell-Fleming intuition):
  - Under predetermined exchange rates, monetary accommodation can prevent currency appreciation and accommodate higher output, amplifying fiscal multipliers.
  - Under flexible exchange rates, appreciation and reduced net exports can offset fiscal expansion, rendering fiscal policy ineffective in the textbook case.
- Extended VAR (government consumption, central bank interest rate, GDP, current account/GDP ratio, real exchange rate):
  - Responses to a 1 percent shock to government consumption:
    - GDP increase is larger and more persistent when exchange rates are fixed than when flexible.
    - Loss of statistical significance arises when adding these variables because of reduced sample size (policy interest rate data available only for a subset).
  - Weak evidence for the traditional Mundell-Fleming channel in the extended VAR:
    - Real exchange rate appreciates on impact under flexible regimes (as expected) but depreciates under fixed regimes (source text).
    - Current account responses examined but not strongly conclusive in the figure described.

---

*Source: _wp1152 - 1.3  Fiscal Multipliers: Denitions (excerpt from the provided IMF chapter PDF).*

### section 3.6, where full results from multivariate VARs are presented. The ordering of the scal variable before

### _wp1152 - section 3.6, where full results from multivariate VARs are presented. The ordering of the scal variable before

### Monetary accommodation and exchange rate regimes
- Ordering government consumption before the central bank’s instrument follows from the assumption that the monetary authority can respond more rapidly to news than can scal decision-makers.
- Results are virtually unchaged if the policy interest rate is ranked lower in the Cholesky ordering; however, the response of the policy interest rates is signicantly weakened if the ordering of the scal and monetary variables is reversed.
- Strong evidence for a "monetary accommodation" channel:
  - Under predetermined (fixed) exchange rates, the short-term nominal interest rate declines by a cumulative 125 basis points in the two years following a government consumption shock of 1% of GDP.
  - Under flexible exchange rates, central banks increase the policy interest rate by an average of 60 basis points within the two years following a scal shock of similar magnitude.
- Differences in monetary accommodation are identified as the main cause for differences in the magnitude of scal multipliers across exchange rate regimes.
- The response of the current account does not differ significantly across exchange rate regimes; the current account deteriorates following a government consumption shock under both fixed and flexible regimes.

### Consumption versus investment responses
- Replacing GDP with private consumption and private investment shows:
  - Investment response is similar under predetermined and flexible exchange rate regimes: erratic, with investment declining by a statistically signicant margin on impact and additional dips later.
  - Private consumption response differs markedly:
    - Under fixed exchange rates, consumption responds positively on impact and by a statistically signicant margin to a government consumption shock.
    - Under flexible exchange rates, the response of private consumption is not statistically signicant from zero; point estimates show a negative response in both the short and long run.
- Interpretation: when the central bank accommodates the fiscal shock, private consumption responds positively; when it does not, consumption may decline.

### Openness to trade
- Countries classified as "open" if trade (imports plus exports) to GDP ratio exceeded 60 percent; "closed" otherwise. Using this criterion: 28 "open" countries and 16 "closed" countries.
- Cumulative responses:
  - For high trade-GDP ratio economies: impact response is 0.11 and the long-run multiplier is 1.4.
  - For low trade-GDP ratio economies: impact response is negative and the long-run response is negative and statistically signicant from zero.
  - The difference between the two categories is statistically signicant at every forecast horizon.
- Alternative openness definition (average tariffs greater than 4 percent vs less than 4 percent):
  - Open economies: multipliers of -0.28 on impact and -0.75 in the long run.
  - Closed economies: multipliers of 0.02 on impact and 1.29 in the long run.
- Size of economy (ten largest vs remaining):
  - Large economies: impact multiplier 0.02; long-run multiplier 1.2.
  - Small economies: impact multiplier -0.19; long-run multiplier -0.47.
  - Difference statistically significant on impact, but not at longer horizons.

### Financial fragility (debt burden)
- High-debt episodes defined where total central government debt exceeded 60 percent of GDP for any period of three or more consecutive years.
- Estimates during high debt burden:
  - Impact multiplier is very close to zero.
  - Long run multiplier estimated at -2.3.
- Robustness:
  - Long run multiplier remains negative when threshold set to 60 or 70 percent of GDP; becomes positive for debt-to-GDP ratios of 30 or 40 percent.
  - The 60 percent threshold identified as a meaningful cuto¤ above which fiscal stimulus appears ine¤ective.
- Interpretation: high sovereign debt may signal future fiscal tightening, offsetting short-term expansionary effects and possibly making fiscal stimulus counter-productive.

### Government investment
- System estimated with three endogenous variables gI (real government investment), g (real government consumption), and y (real GDP); government investment ordered before government consumption in the Cholesky decomposition.
- Simple bivariate regression (government investment and GDP) for high-income countries shows higher impact and long-run investment multipliers than for government consumption, but results may be misleading because government consumption responds strongly to government investment.
- "Pure" government investment multiplier (setting g_t = 0 in forecasts for gI_t and y_t):
  - High-income countries: estimates highly uncertain; point estimates similar to government consumption multipliers.
  - Developing countries: impact multiplier of 0:6 and statistically significant; can reject at the 95% condence level that government investment effect is no higher than government consumption in short horizons (first two years).
- Cross-dimension results for pure government investment multiplier:
  - Predetermined exchange rates: impact 0.36 and long run 1.42.
  - Flexible exchange rates: impact 0.46 and long run 0.16.
  - High trade-to-output ratios: impact 0.51 and long run -0.23.
  - Low trade-to-output ratios: impact 0.46 and long run 0.70.
  - Some differences statistically significant (e.g., trade openness at horizons 1 to 14 quarters), but smaller sample sizes reduce power.

### Multivariate regressions (four-variable VAR)
- Expanded VAR includes: real government consumption, real GDP, current account balance to GDP, and real effective exchange rate.
- Ordering: government consumption before GDP; current account balance before real effective exchange rate (following Kim and Roubini (2008)).
- Results:
  - Point estimates for cumulative multipliers similar to bivariate regressions.
  - Error bands wider due to loss of observations from additional variables.
  - Negative multiplier in highly indebted economies is robust and strengthened when additional controls are included.

### Key quantitative findings (selected)
- Short-term nominal interest rate cumulative decline: 125 basis points over two years after 1% of GDP government consumption shock under fixed exchange rates.
- Short-term nominal interest rate increase: average 60 basis points over two years after similar shock under flexible exchange rates.
- Trade openness (trade/GDP > 60%): impact response 0.11; long-run multiplier 1.4.
- Trade openness (tariff-based definition, open): impact -0.28; long run -0.75. Closed: impact 0.02; long run 1.29.
- Large vs small economies: impact multipliers 0.02 (large) vs -0.19 (small); long-run 1.2 (large) vs -0.47 (small).
- High-debt long-run multiplier: -2.3 (threshold 60 percent of GDP).
- Government investment in developing countries: impact multiplier 0:6 (statistically significant).
- Pure government investment (predetermined exchange rates): impact 0.36; long run 1.42. Flexible: impact 0.46; long run 0.16.

### Policy implications and recommendations
- Monetary accommodation crucially determines the expansionary effect of fiscal policy; fiscal multipliers are larger when monetary policy accommodates.
- Fiscal stimulus likely to be ineffective or counter-productive in highly indebted countries (debt-to-GDP around or above 60 percent).
- In economies open to trade or with flexible exchange rates, fiscal expansions yield little or no significant output gains.
- Composition matters, especially in developing countries: government investment can have a more positive short-run impact on GDP than government consumption.
- Emerging countries with procyclical fiscal policies should pursue a-cyclical fiscal policy (government consumption and tax rates not responding to the business cycle) to reduce the probability of fiscal crises.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1152.pdf*

### References

### _wp1152 - References

### References bibliography
- Comprehensive list of cited works on fiscal policy, government spending multipliers, and related econometric methods, including:
  - Barro, Robert J. (1981), “Output e¤ects of government purchases,” Journal of Political Economy 89, 1086-1121.
  - Barro, Robert J. (2009), “Government spending is no free lunch,” Wall Street Journal (January 22).
  - Barro, Robert, J. and Charles J. Redlick (2009), “Macroeconomic e¤ects from government purchases and taxes,” NBER Working Papers No. 15369.
  - Beetsma, Roel, Massimo Giuliodori, and Franc Klaassen (2008), “The e¤ects of public spending shocks on trade balances and budget decits in the European Union,” Journal of the European Economic Association 6(2-3).
  - Blanchard, Olivier and Roberto Perotti (2002), “An empirical characterization of the dynamic e¤ects of changes in government spending,” Quarterly Journal of Economics 117: 4, 1329-1368.
  - Coenen, Günter, Christopher Erceg, Charles Freedman, Davide Furceri, Michael Kumhof, René Lalonde, Douglas Laxton, Jesper Lindé, Annabelle Mourougane, Dirk Muir, Susanna Mursula, Carlos de Resende, John Roberts, Werner Roeger, Stephen Snudden, Mathias Trabandt and Jan in t Veld (2010), “E¤ects of scal stimulus in structural models,” IMF Working Paper 10/73.
  - Christiano, Lawrence, Martin Eichenbaum, and Sergio Rebelo (2009), “When is the government spending multiplier large?” (mimeo, Northwestern University).
  - Davig, Troy, and Eric Leeper (2009), “Monetary-Fiscal Policy Interactions and Fiscal Stimulus,” (mimeo, Indiana University).
  - Erceg, Christopher, and Jesper Lindé (2010), “Is there a scal free lunch in a liquidity trap?” (mimeo, Board of Governors of the Federal Reserve).
  - Fatás, Antonio and Ilian Mihov (2001), “The e¤ects of scal policy on consumption and employment: Theory and evidence,” CEPR Discussion Papers 2760.
  - Eurostat (2006), Manual on quarterly non-nancial accounts for general government (European Commission and Eurostat).
  - Gómez, Victor and Augstín Maravall (2000), “Seasonal adjustment and signal extraction in economic time series” in Daniel Peña, George C. Tiao, and Ruey S. Tsay, eds. A Course in Time Series Analysis.
  - Hall, Robert (2009), “By how much does GDP rise if the government buys more output?” Brookings Panel on Economic Activity (September).
  - Jonas D. M., and Ryan Peters (2009), “Using stock returns to identify government spending shocks,” Working Paper Series WP-09-03 (Federal Reserve Bank of Chicago).
  - Ilzetzki, Ethan, Carmen Reinhart and Kenneth Rogo¤ (2009), “Exchange rate arrangements entering the 21st century: Which anchor will hold?” (mimeo, University of Maryland and Harvard University).
  - Ilzetzki, Ethan and Carlos A. Végh (2008), “Procyclical scal policy in developing countries: Truth or ction?" NBER Working Paper No. 14191.
  - Im, Kyung So, M. Hashem Pesaran and Yongcheol Shin (2003), “Testing for unit roots in heterogeneous panels,” Journal of Econometrics, 115:1.
  - IMF (2007). The Special Data Dissemination Standard. Guide for Subscribers and users. International Monetary Fund.
  - Kaminsky, Graciela, Carmen Reinhart, and Carlos Vegh (2004), “When it rains, it pours: Procyclical capital ows and macroeconomic policies,” NBER Macroeconomics Annual.
  - Kelejian, H. H., 1974. “Random Parameters in a Simultaneous Equation Framework: Identication and Estimation.” Econometrica, 42:3.
  - Kim, Soyoung and Nouriel Roubini (2008), “Twin decit or twin divergence? Fiscal policy, current account, and real exchange rate in the U.S,” Journal of International Economics 74:2.
  - Nickell, Stephen (1981), “Biases in dynamic models with xed e¤ects,” Econometrica 49, 1417-26.
  - Mountford, Andrew and Harald Uhlig (2008), “What are the e¤ects of scal policy shocks?,” NBER Working Papers 14551.
  - Perotti, Roberto (2004a), “Estimating the e¤ects of scal policy in OECD countries” (mimeo, Bocconi University).
  - Perotti, Roberto (2004b), “Public investment: another (di¤erent) look” (mimeo, Bocconi University).
  - Roberto Perotti (2007), “In search of the transmission mechanism of scal policy,” NBER Working Paper No. 13143.
  - Ramey, Valerie A. (2009), “Identifying government spending shocks: Its all in the timing," NBER Working Papers No. 15464.
  - Ramey, Valerie A. and Matthew D. Shapiro (1998), “Costly capital reallocation and the e¤ects of government spending,” Carnegie-Rochester Conference Series on Public Policy, Elsevier, vol. 48(1), 145-194.
  - Romer, Christina, and Jared Bernstein (2009), “The job impact of the American recovery and reinvestment plan” (Council of Economic Advisers).
  - Sa, Filipa, Pascal Towbin and Tomasz Wieladek (2010), “Low interest rates and housing booms: Capital inows, monetary policy, and the e¤ects of nancial innovation,” Bank of England Working Paper.

### Data Appendix — dataset construction, coverage, and quality checks
- Main challenge:
  - Collecting and vetting data originally collected at a quarterly frequency across a large number of countries.
- Inclusion and coverage rules:
  - Data included only if originally collected at a quarterly frequency and covered a large proportion of the government expenditure category studied.
  - Data integrity confirmed directly through correspondence or conversation with national statistical agencies, central banks, or scal authorities.
  - While some time series extend to 2008 or 2009, data from these last two years were not used in the empirical analyses in this paper, as recent data may still be subject to significant revisions.
- Government level inconsistency:
  - In most cases, data for the general government was available; in some cases, only central government expenditures were available.
  - The exclusion of regional government consumption may bias results, depending on correlation with central government consumption and their effects on economic activity.
  - The authors included all countries regardless of the level of government for which data was available.
  - Reported results did not change when analysis was restricted to countries for which general government consumption data was available.
- Price deflation and deators:
  - In some cases data were deated directly by local statistical agencies; in other cases a CPI deator was used.
  - This creates inconsistencies across countries and raises the question of whether consumer prices are the appropriate deator for government purchases.
  - Where multiple deators were available (e.g. GDP deator, CPI, or a government consumption deator), the correlation of the series when deated using di¤erent price indexes was close to1.

### Data sources and variable definitions
- Gross Domestic Product:
  - Whenever possible, GDP data are taken from the same data source as the government expenditure data.
  - Elsewhere, IFS series 99B was used for developing countries and real GDP from OECD Statistics was used for high-income countries.
- Consumer Price Index:
  - IFS series 64.
- Current Account:
  - OECD Statistics, when available, and IFS series 75 elsewhere.
  - Current account to GDP ratio was created by dividing this series using nominal GDP from OECD statistics or from IFS series 99B converted to US dollars using the average interest rate for that year.
- Central Bank Discount Rate:
  - Central bank discount rates were taken from IFS series 60 whenever available.
  - Series 60P (repurchase agreement rate) was used for France.
  - Series 60A (rate on advances) was used for the Netherlands, and 60A for Romania (National Bank of Romania Structural Credit rate).
  - Central bank data was used to obtain the policy rates for Australia, El Salvador, the Eurozone, Estonia, Mexico, and the United Kingdom.
  - Series for Eurozone countries were created by splicing interbank rates of the national central banks until the later of 1998 or the date of Euro adoption, with the ECB’s deposit facility rate thereafter.
  - In Estonia, the average between the one-month Talibid and Talibor rates was used (bid and ask rates), as the overnight rate was used as a policy tool starting only in

*Source: _wp1152 - References (PDF chapter/section).*

### 2007. For Mexico, the 30-day interbank rate (TIIE) was used.

### _wp1152 - 2007. For Mexico, the 30-day interbank rate (TIIE) was used.

### Data definitions and series construction
- Debt to GDP: "Debt of the central government (external only for developing countries) as a proportion of GDP" from the OECD, Eurostat, the Joint External Debt Hub, and the Federal Reserve Bank of St. Louis and ONS for the United States and United Kingdom, respectively.
- Real Effective Exchange Rate: "A CPI-based real exchange rate was used. Where ever available, the narrow real exchange rate index of the Bank for International Settlements was used. Otherwise, the broad index was used. Elsewhere, IFS series RECZF was used."
- Trade to GDP: "The ratio of the sum of imports (IFS series 70) and exports (IFS series 71) to GDP (IFS series 99)."
- For Mexico: "the 30-day interbank rate (TIIE) was used."

### Tables and model selection
- Table 1: "Optimal Number of Lags Based on Specification Tests" (table referenced; specification criteria listed include Akaike, Schwartz, Hannan Quinn with cell entries such as 8, 8, 8, 8, 8, 4, 3, 2, 1 in the reproduced fragment).
- Table 2: "Summary Statistics on Quarterly Government Consumption Data" (table referenced).

### Figures and key reported multiplier statistics
- Figure 12: Cumulative government investment multiplier (High-income countries; includes indirect effects of government consumption)
  - Long Run: 2.5
  - Impact: 0.38
- Figure 13: Responses to a 1% government investment shock (Responses shown for government investment, government consumption, and GDP; 90% confidence intervals in dashed lines).
- Figure 14: Cumulative multiplier to a "pure" government investment shock — high-income and developing countries (High income top panel; developing countries lower panel; 90% confidence intervals in dashed lines).
- Figure 17: Cumulative multiplier — high income and developing countries (Multivariate Regression)
  - Long Run: 0.85
  - Impact: 0.39
  - Additional panel/reporting shows:
    - Long Run: 0.53
    - Impact: -0.09
- Figure 20: Cumulative multiplier: Highly indebted countries (Multivariate Regression)
  - Long Run: -4.7
  - Impact: -0.16
- Additional figures referenced (without numeric summaries in the extracted text) include:
  - Figure 1: Quarterly government consumption
  - Figure 2: Quarterly government consumption and commodity prices
  - Figure 3a & 3b: Central bank estimation errors and VAR residuals
  - Figure 4: Impulse responses in high-income countries (Responses of G and GDP to a 1% shock)
  - Figure 5: Impulse responses in developing countries (Responses of G and GDP to a 1% shock)
  - Figure 6: Cumulative multiplier — high income and developing countries
  - Figure 7: Cumulative multiplier — predetermined (fixed) and flexible (flex) exchange arrangements
  - Figures 8a–8c: Responses to a 1% shock to government consumption (Fixed / Flex; policy interest rate responses)
  - Figure 9: Responses of private investment and consumption to a 1% shock to government consumption (Fixed / Flex)
  - Figure 10a & 10b: Cumulative multiplier — open and closed economies; effect of total trade to GDP
  - Figure 11: Cumulative multiplier: Highly indebted countries
  - Figure 15–16: Cumulative multiplier to a "pure" government investment shock (fixed vs flex; open vs closed)
  - Figure 18–19: Cumulative multiplier — predetermined vs flexible exchange rates; open vs closed economies (Multivariate Regression)
  - Figure 21+ (referenced implicitly): additional robustness and subgroup analyses

### Country coverage, panels, and sample periods (selected entries from tables)
- Government consumption series (Table A1) — selected entries:
  - Argentina: 1993q1–2006q4, MECON, "Real General Government Consumption. Seasonally Adjusted and deflated by MECON"
  - Australia: 1960q1–2008q4, OECD, "General Government Consumption"
  - Mexico: 1991q1–2008q4, Finance Ministry, "Central Government Consumption. Calculated as the sum of wages and sallaries and acquisitions."
  - United Kingdom: 1960q1–2008q4, OECD, "Real General Government Consumption, Chain Valued, Seasonally Adjusted"
  - United States: 1960q1–2008q4, OECD, "Real General Government Consumption, Chain Valued, Seasonally Adjusted"
  - (Full table contains many additional country-level start/end dates, sources, and series descriptions.)
- Government investment series (Table A2) — selected entries:
  - Argentina: 1993q1–2006q4, INDEC, "Central government investment."
  - Australia: 1959q3–2009q3, OECD, "Real general government investment. Seasonally adjusted."
  - Mexico: 1991q1–2009q1, Secretaría de Hacienda y Crédito Público, "Central government investment."
  - United States: 1960q1–2009q3, OECD, "Real general government investment. Current prices. Seasonally adjusted. Millions of national currency."
- Episodes of de-facto fixed and flexible exchange rates (Table A3) — examples:
  - Argentina: Fixed episodes 1993q1:2001q4; 2003q2:2006q4
  - Australia: Fixed 1960q1:1982q4; Flexible 1983q1:2007q4
  - Mexico: Flexible episodes listed as 1991q1:2007q4
  - United Kingdom: Fixed 1960q1:1972q2; Flexible 1972q3:2007q4
  - United States: Fixed 1960q1:1971q3; Flexible 1971q3:2007q4
  - Total Observations: 15011119 (as reproduced in the extracted fragment)
- Open and closed economies (Table A4) — de-facto classification uses a ratio of exports+imports to GDP greater than or less than 60%
  - Open/Closed threshold explicitly stated: "De-facto: Economies with a ratio of exports+imports to GDP greater than or less than 60%"
  - Sample trade/GDP values (Open group shown with Trade/GDP):
    - Netherlands: 217.1%
    - Estonia: 154.7%
    - Czech Republic: 125.0%
    - Ireland: 118.0%
    - Slovenia: 118.4%
    - Malaysia: 132.5%
    - United States (Closed group example): 18.2%
  - Total Observations: 13661234 (as reproduced in the extracted fragment)
- Episodes of High Debt (Table A6) — definition: "Gross Debt of Central Government Exceeding 60% of GDP"
  - Examples:
    - Argentina: 2002–2006
    - Belgium: 1981–2007
    - Brazil: 1991–1993, 1999–2005
    - Italy: 1982–2007
    - Uruguay: 2002–2006
  - Total Observations: 1501

### Classification thresholds and supplementary rules
- Average tariff rates classification: "Average tariff rates greater than or smaller than 4%"
- Open vs Closed economies classification: "De-facto: Economies with a ratio of exports+imports to GDP greater than or less than 60%"

*Italic: Source — _wp1152 - 2007. For Mexico, the 30-day interbank rate (TIIE) was used.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1152.pdf_
