## spillovernote10

## Source details

**Canonical URL:** [spillovernote10](https://www.imf.org/-/media/files/publications/spillovernotes/spillovernote10.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/spillovernotes/spillovernote10.pdf.md)
- [Structured JSON version](/-/media/files/publications/spillovernotes/spillovernote10.pdf.json)

---

### Introduction — Motivation, scope, and identification
- Examines the impact of fiscal spending shocks from the United States on the US trade balance and real exchange rate.
- Motivation:
  - Spillovers of US fiscal policy on exchange rates and the trade balance remain controversial.
  - The United States is a very significant contributor to developments in global imbalances.
  - Survey data required for the identification procedure (Survey of Professional Forecasters, SPF) are available only for the United States and cover only government spending.
- Identification challenge highlighted: difficulty identifying fiscal shocks because of fast-moving variables like the exchange rate and the presence of fiscal foresight (anticipated fiscal actions).
- Identification strategy:
  - Uses Forni and Gambetti (2016) implementation exploiting SPF forecasts to extract preannounced/anticipated (“news” or “foresight”) fiscal spending shocks.
  - Fiscal news variable: cumulative SPF forecasts for government spending growth one year ahead (H = 4).
  - Two complementary methodologies:
    1. US-only VAR to estimate impacts of anticipated spending shocks on the US real exchange rate and trade balance.
    2. Panel VAR (PVAR) with trade-weighted linkages for a panel of advanced and emerging market economies, conditioning on recipient-country macro and policy variables.

### Literature context and theoretical channels
- Theoretical expectation (Mundell-Fleming and standard DSGE):
  - Expansionary fiscal policy should lead to nominal and real currency appreciation and a worsening of the trade balance via:
    - Demand (trade/expenditure-shifting) channel: fiscal expansion boosts domestic demand and foreign goods demand.
    - Competitiveness/expenditure-switching channel: inflationary pressures raise interest rates, attract capital inflows, and appreciate the currency, reducing competitiveness.
    - Home-bias of government spending: raises relative price of home-produced goods, causing real appreciation.
- Empirical controversies:
  - SVAR evidence sometimes finds real depreciation in response to government spending increases (the “depreciation puzzle”).
  - Narrative/historical approaches have more recently documented appreciation in response to expansionary spending shocks.
  - Findings on trade balance effects are mixed across studies.

### US VAR specification and data
- Sample: Quarterly US data from 1981:Q1 through 2016:Q4 (baseline VAR described also with fourth quarter of 1981 through third quarter of 2016 in the Government Surplus section).
- VAR variables, in this order:
  1. real federal government consumption expenditures and gross investment (FEDGOV)
  2. the fiscal “news” variable based on SPF forecasts
  3. real GDP
  4. private consumption
  5. the federal surplus divided by GDP
  6. net exports of goods and services divided by GDP
  7. the 10-Year Treasury Constant Maturity Rate
  8. the effective real exchange rate
- Identification: recursive Cholesky ordering placing slow-moving macro variables first and fast-moving financial variables last; spending and the spending “news” variable placed before GDP.

### Key empirical results — US responses to a 1 percent of GDP preannounced (“news”) spending shock
- FEDGOV (actual government spending):
  - Effect negligible in the first four to five quarters (anticipation), then spending rises, peaking after about three years at about 2 percent.
- Federal surplus:
  - Declines quite persistently by up to 1.5 percentage points of GDP at peak.
- 10-Year Treasury Constant Maturity Rate:
  - Increases immediately and very significantly by about 1.4 percentage points; effects highly significant for about 1.5 years.
- Output and private consumption:
  - Impact multiplier on output is 0.5 on impact; peak multiplier reaches 1.5 after one year.
  - Consumption is significantly and persistently crowded in.
- Real effective exchange rate (dollar):
  - Appreciates, peaking after about six quarters at about 7 percent.
- US trade balance (net exports of goods and services divided by GDP):
  - Worsens significantly by 0.65 percentage point of GDP after three years; effect is long lived.
- Interpretation:
  - Results support the twin-deficits hypothesis for the United States.
  - No “depreciation puzzle” is found when anticipation is captured via the fiscal news identification strategy.

### Panel (PVAR) overview and complementary findings
- PVAR composition:
  - Unbalanced panel of 30 US trading partners (23 advanced economies and 7 emerging market economies, representing about 80 percent of US imports) spanning fourth quarter of 1982 to the third quarter of 2016.
- Panel VAR baseline variables (in this order): trade-weighted US news shock; real GDP; fiscal balance as percent of GDP; net exports as share of GDP; long-term interest rates; real bilateral exchange rate.
- Main panel findings (impulse responses to a 1 percent of US GDP “news” shock):
  - Bilateral real exchange rate (recipient's perspective) depreciates by about 5 percent after five quarters.
  - Trade balance (as percent of recipient’s GDP) improves significantly and persistently, peaking at about 0.3 percentage point of recipient country’s GDP after two years.
- Magnitude interpretation:
  - Assuming an average trade balance as a share of a recipient’s GDP of 1.25 percent (historical average for sample), a peak improvement of 0.3 percentage point translates into an improvement of about 25 percent (≈0.3/1.25) in the recipient’s total trade balance.
  - Note: the calibrated fiscal shock (1 percent of US GDP) is very large by historical standards.

### Crisis-era differences and regional patterns
- Impact of the Global Financial Crisis:
  - Effects on exchange rates and possibly trade balances may have weakened after the global financial crisis (pre-2007 vs entire sample).
  - Magnitudes were significantly higher before the crisis for the real exchange rate and possibly for the trade balance.
  - Possible mechanism: post-crisis US interest rate at the zero lower bound dampened real exchange rate appreciation in response to expansionary fiscal shocks (weaker expenditure-switching channel).
  - Countervailing literature finding: fiscal multipliers on output might have been larger during the global financial crisis, implying a stronger direct demand effect on trade.
- Regional comparison:
  - Using regional trade-weighted aggregates (Europe, Asia, Latin America): expansionary US fiscal policy impacts trade with all regions significantly and quantitatively similarly.
  - For Europe, Asia, and Latin America, the decline in the trade balance is between 20 and 30 percent in response to the calibrated fiscal shock of 1 percent of US GDP.

### Robustness
- Results robust to:
  - Alternative orderings in the US VAR.
  - Using generalized impulse responses rather than orthogonalized Cholesky impulse-response functions.
  - Panel VAR specifications for groups of countries (advanced economies, Group of Twenty).
  - Excluding significant trade partners (China) or outliers (Ireland).
  - Inclusion of the short-term rate as an additional endogenous variable.
  - Time-varying trade weights with rolling windows of between 1 and 10 years and alternative definitions of weights (for example, country i’s imports from the United States as a share of country i’s total imports; US imports [exports] from [to] country i as a share of US total imports [exports]; similar measures for total trade volumes).

### Methodological note on the fiscal “news” variable
- Definition:
  - ηt+1 = Ft(1, H) = ∑h=1^H Et^Pt g t+h, with H = 4.
  - Et^Pt g t+h is the expectation at time t conditional on the information set Pt of government spending h quarters ahead.
  - The SPF reports forecasts for government spending for periods t, t+1, ..., t+4; forecasts are expressed as annualized percentage changes of forecasters’ mean response.
  - The cumulative four-quarter forecast best predicts government spending compared with shorter-horizon forecasts; an alternative robustness specification using differences in cumulative forecasts for three quarters ahead does not change results.
- Identification via Cholesky ordering:
  - First shock captures unanticipated (surprise) changes in government spending.
  - Second shock reflects anticipated (“news”) changes orthogonal to professional forecasts and contemporaneous macroeconomic shocks.

### Conclusions (summary)
- Preannounced government spending shocks in the United States:
  - Lead to a sizable real effective appreciation in the dollar and a worsening of the aggregate US trade balance.
  - Quantitatively: a 1 percent of GDP spending announcement would appreciate the dollar by about 5–7 percent over 1.5 years and worsen net exports by about 0.6 percentage point over 2–3 years.
- For main US trading partners:
  - US fiscal expansion results in an economically and statistically significant improvement in partner countries’ trade balances and a deterioration in bilateral exchange rates from the partners’ perspective.
  - Effects are consistent across regions.
- Post-global financial crisis weakening:
  - Spillovers on exchange rates and trade balances may have diminished following the global financial crisis, potentially because constrained monetary policy at the zero lower bound dampened exchange rate appreciation and thereby weakened the trade balance response.

### Annex 2 — Empirical approach for the panel (PVAR technical details)
- Baseline PVAR specification:
  - X_{i,t} = a_{i} + A(L) X_{i,t−1} + U_{i,t}, where a_{i} is country fixed effects and U_{i,t} is the error term.
  - X_{t} includes, in this order: trade-weighted fiscal news shocks extracted from the US VAR (US news shock); real GDP; fiscal balance as a percentage of GDP; net exports of goods and services as a share of GDP; long-term interest rates; real bilateral exchange rate.
- Variable treatments:
  - All variables enter in (logged) levels, except surplus, trade balance, and bond yield, which are expressed in percent.
- Lag structure and estimator:
  - Benchmark specification includes four lags.
  - PVAR methodology: least-squares dummy variable (LSDV) estimator as in Bun and Kiviet 2006.
- Sample and data coverage:
  - Unbalanced panel of the top 30 US trading partners (23 advanced economies and 7 emerging market economies), sample period fourth quarter of 1982 to third quarter of 2016.
- Construction of trade-weighted US news shock:
  - Baseline scaling: SPF news shock expressed as a 1 percent of US GDP shock.
  - Weighting: USNewsShock_{i,t} = w_{i,(t)}^{EXP} × USNewsShock_{t}^{US−VAR}, where w_{i,(t)}^{EXP} is country i’s exports to the United States as a share of its total exports.
- Data definitions and key variable constructions:
  - SPFNEWS: based on SPF forecasts; annualized percent change in mean responses for real federal government consumption expenditure and gross investment from the Federal Reserve Bank of Philadelphia.
  - FEDGOV, SUR, GDP, trade balance: from Federal Reserve Economic Data (FRED).
  - Trade balance in the PVAR: 100 × ((country i’s real exports to the United States) − (country i’s real imports from the United States))/(country i’s real GDP); nominal exports/imports deflated by partner country export/import deflators; uses average of US and partner reports; robust to using US-reported trade only.
  - Real bilateral exchange rate: (nominal exchange rate) × (GDP deflator in country i)/(US GDP deflator), normalized to be 100 in 2010; nominal rate defined as (US dollar/national currency).
- Annex Table 3.1 (variables and sources):
  - SUR: Fiscal balance, percent of GDP — Quarterly — GDS — G_B_GDP
  - STRATE: Short-term interest rate — Quarterly — GDS — FIDR
  - 10YBOND: Long-term interest rate — Quarterly — GDS — FIGB
  - RGDP: Real gross domestic product — Quarterly — GDS — NGDP_R
  - TB: Trade balance — Quarterly — DOTS
  - RER: Real bilateral exchange rate — Quarterly — DOTS
  - Note: DOTS = Direction of Trade Statistics; GDS = Global Data Source.

*Source: Spillovers from US Government Spending Shocks: Impact on External Positions — Introduction; Government Surplus; Annex 2 (spillovernote10).*

### Introduction

### Introduction

### Motivation and scope
- Examines the impact of fiscal spending shocks from the United States on the US trade balance and real exchange rate.
- Motivation:
  - Spillovers of US fiscal policy on exchange rates and the trade balance remain controversial.
  - The United States is a very significant contributor to developments in global imbalances.
  - Survey data required for the identification procedure (Survey of Professional Forecasters, SPF) are available only for the United States and cover only government spending.
- Challenge highlighted: identifying fiscal shocks is difficult because of fast-moving variables like the exchange rate and the presence of fiscal foresight (anticipated fiscal actions).

### Literature context and controversies
- Theoretical expectation (Mundell-Fleming and standard DSGE): expansionary fiscal policy should lead to nominal and real currency appreciation and a worsening of the trade balance via:
  - Demand (trade/expenditure-shifting) channel: fiscal expansion boosts domestic demand and foreign goods demand.
  - Competitiveness/expenditure-switching channel: inflationary pressures raise interest rates, attract capital inflows, and appreciate the currency, reducing competitiveness.
  - Home-bias of government spending: raises relative price of home-produced goods, causing real appreciation.
- Empirical controversies:
  - SVAR evidence for the United States sometimes finds real depreciation in response to government spending increases (the “depreciation puzzle”).
  - Narrative/historical approaches have more recently documented appreciation in response to expansionary spending shocks.
  - Findings on trade balance effects are mixed across studies, with some papers finding improvement and others finding deterioration.

### Identification strategy and data
- Uses a novel identification strategy based on Forni and Gambetti (2016), exploiting SPF forecasts of government spending to extract preannounced/anticipated (“news” or “foresight”) fiscal spending shocks.
  - The fiscal news variable is defined as the cumulative SPF forecasts for government spending growth one year ahead.
  - Anticipated shocks represent announcements about future changes in government spending that affect expectations immediately but whose effects on actual spending are delayed.
- Two complementary methodologies employed:
  1. VAR using US data only to estimate impacts of anticipated spending shocks on the US real exchange rate and trade balance.
  2. Panel VAR (PVAR) with trade-weighted linkages for a panel of advanced and emerging market economies, conditioning on recipient-country macro and policy variables (cyclical positions, monetary policy, domestic fiscal policy).

### VAR specification (US-only)
- Quarterly US data from 1981:Q1 through 2016:Q4.
- VAR variables, in this order:
  1. real federal government consumption expenditures and gross investment (FEDGOV)
  2. the fiscal “news” variable based on SPF forecasts
  3. real GDP
  4. private consumption
  5. the federal surplus divided by GDP
  6. net exports of goods and services divided by GDP
  7. the 10-Year Treasury Constant Maturity Rate
  8. the effective real exchange rate
- Identification: recursive ordering placing slow-moving macro variables first and fast-moving financial variables last; spending and the spending “news” variable are placed before GDP.

### Key empirical results (main findings)
- Aggregate US responses to a preannounced (“news”) spending shock of 1 percent of GDP:
  - Real federal government consumption expenditures and gross investment (FEDGOV):
    - Actual government spending increases significantly and very persistently.
    - Effect negligible in the first four to five quarters (anticipation), then spending rises, peaking after about three years at about 2 percent.
  - Federal surplus:
    - Declines quite persistently by up to 1.5 percentage points of GDP at peak (consistent with deficit financing).
  - 10-Year Treasury Constant Maturity Rate (nominal long-term interest rate):
    - Increases immediately and very significantly by about 1.4 percentage points; effects highly significant for about 1.5 years.
  - Output and private consumption:
    - Impact multiplier on output is 0.5 on impact; peak multiplier reaches 1.5 after one year.
    - Consumption is significantly and persistently crowded in.
  - Real effective exchange rate (dollar):
    - Appreciates, peaking after about six quarters at about 7 percent.
  - US trade balance (net exports of goods and services divided by GDP):
    - Worsens significantly by 0.65 percentage point of GDP after three years; effect is long lived.
- Interpretation:
  - Results support the twin-deficits hypothesis for the United States.
  - No “depreciation puzzle” is found when anticipation is captured via the fiscal news identification strategy.

### Additional findings and robustness
- Panel results (PVAR with trade-weighted linkages):
  - US fiscal stimulus is largely shared across all regions: for a panel of advanced and emerging market economies that constitute the main US trading partners, a US fiscal expansion results in an economically and statistically significant improvement in the trade balance and deterioration in the bilateral exchange rate for US partner countries.
- Crisis-era differences:
  - The response of exchange rates and trade balances to US fiscal shocks may have diminished following the global financial crisis, potentially because constrained monetary policy at the zero lower bound dampened exchange rate appreciation and therefore weakened the trade balance response.
- Robustness:
  - Findings hold for bilateral exchange rates and trade balances across a panel of advanced and emerging market economies.
  - Results are economically meaningful and very robust to method, specification, and regional grouping.

*Source: Spillovers from US Government Spending Shocks: Impact on External Positions — Introduction (spillovernote10 - Introduction).*

### 5. Government Surplus

### 5. Government Surplus

### Identified news shocks and historical episodes
- The VAR-extracted standardized “news” shocks exhibit spikes associated with large preannounced changes in government spending, including:
  - Strategic Defense Initiative (first quarter of 1983)
  - Emergency Deficit Control Act (episode marker in figure)
  - Fall of the Berlin Wall (fourth quarter of 1989)
  - Gulf war (third quarter of 1990)
  - Clinton's election (episode marker)
  - War in Afghanistan (first quarter of 2001)
  - Second Gulf War (first quarter of 2003)
  - 2008 fiscal stimulus (episode marker)
  - Obama (2009) fiscal stimulus (first quarter of 2009)
  - 2011 Budget Control Act (episode marker)
- Episode identification confirms that the identified shocks are consistent with interpretation as anticipated (preannounced) changes in government spending.

### Empirical framework and data coverage
- Baseline US VAR specification (quarterly data, fourth quarter of 1981 through third quarter of 2016) includes, in this order:
  - real federal government consumption expenditures and gross investment
  - fiscal news variable based on SPF forecasts (H = 4; cumulative expectation for subsequent four quarters)
  - real GDP
  - private consumption
  - federal surplus divided by GDP (expressed in percent)
  - net exports of goods and services divided by GDP (expressed in percent)
  - 10-year Treasury constant maturity rate (expressed in percent)
  - real effective exchange rate
- Panel VAR (PVAR) uses an unbalanced panel comprising 30 US trading partners (23 advanced economies and 7 emerging market economies, representing about 80 percent of US imports) spanning fourth quarter of 1982 to the third quarter of 2016.
- News shocks identification in the PVAR: news shocks are identified recursively as the first shocks in the PVAR.

### Key empirical findings — US VAR (aggregate)
- Preannounced increases in US government expenditures:
  - appreciate the dollar by about 5–7 percent over 1.5 years
  - worsen US net exports by about 0.6 percentage point over the course of 2–3 years

### Key empirical findings — Panel VAR (partner countries)
- Impulse responses to a 1 percent of US GDP “news”/anticipated fiscal spending shock show:
  - bilateral real exchange rate (recipient's perspective) depreciates by about 5 percent after five quarters
  - trade balance (as percent of recipient’s GDP) improves significantly and persistently, peaking at about 0.3 percentage point of a recipient country’s GDP after two years
- Magnitude interpretation:
  - Assuming an average trade balance as a share of a recipient’s GDP of 1.25 percent (historical average for sample), a peak improvement of 0.3 percentage point translates into an improvement of about 25 percent (≈0.3/1.25) in the recipient’s total trade balance
  - Note: the calibrated fiscal shock (1 percent of US GDP) is very large by historical standards

### Impact of the Global Financial Crisis
- Estimation over subsamples indicates that the impact of US fiscal expansion on exchange rates and possibly trade balances may have weakened after the global financial crisis (pre-2007 vs entire sample):
  - Magnitudes were significantly higher before the crisis for the real exchange rate and possibly for the trade balance
  - Possible mechanism: since the global financial crisis, the interest rate in the United States remained at a zero lower bound, dampening real exchange rate appreciation in response to expansionary fiscal shocks (weaker expenditure-switching channel)
  - Countervailing effect: prior literature suggests fiscal multipliers on output might have been larger during the global financial crisis, implying a stronger direct demand effect on trade

### Regional comparison
- Using regional trade-weighted aggregates (Europe, Asia, and Latin America) and region-specific estimates from the US VAR:
  - Expansionary US fiscal policy impacts trade with all regions significantly and in a quantitatively similar manner
  - For key trading partner regions Europe, Asia, and Latin America, the decline in the trade balance is between 20 and 30 percent in response to the calibrated fiscal shock of 1 percent of US GDP

### Robustness checks
- Results are robust to:
  - Alternative orderings of variables in the US VAR
  - Using generalized impulse responses rather than orthogonalized Cholesky impulse-response functions
  - Panel VAR specifications for groups of countries (advanced economies, Group of Twenty)
  - Excluding significant trade partners (China) or outliers (Ireland)
  - Inclusion of the short-term rate as an additional endogenous variable
  - Time-varying trade weights with rolling windows of between 1 and 10 years and alternative definitions of weights (for example, country i’s imports from the United States as a share of country i’s total imports, US imports [exports] from [to] country i as a share of US total imports [exports], and similar measures for total trade volumes)

### Methodological note on the fiscal “news” variable
- Fiscal news variable definition (Forni and Gambetti 2016 implementation):
  - ηt+1 = Ft(1, H) = ∑h=1^H Et^Pt g t+h, with H = 4
  - Et^Pt g t+h is the expectation at time t conditional on the information set Pt of government spending h quarters ahead
  - The SPF reports forecasts for government spending for periods t, t+1, ..., t+4; forecasts are expressed as annualized percentage changes of forecasters’ mean response
  - The cumulative four-quarter forecast best predicts government spending compared with shorter-horizon forecasts; alternative robustness specification using differences in cumulative forecasts for three quarters ahead does not change results
- Identification via Cholesky ordering:
  - First shock captures unanticipated (surprise) changes in government spending
  - Second shock reflects anticipated (“news”) changes orthogonal to professional forecasts and contemporaneous macroeconomic shocks

### Conclusions (summary)
- Preannounced government spending shocks in the United States:
  - lead to a sizable real effective appreciation in the dollar and a worsening of the aggregate US trade balance
  - quantitatively: a 1 percent of GDP spending announcement would appreciate the dollar by about 5–7 percent over 1.5 years and worsen net exports by about 0.6 percentage point over 2–3 years
- For main US trading partners (advanced and emerging market economies):
  - US fiscal expansion results in an economically and statistically significant improvement in partner countries’ trade balances and a deterioration in bilateral exchange rates from the partners’ perspective
  - Effects are consistent across regions
- Post-global financial crisis weakening:
  - Spillovers on exchange rates and trade balances may have diminished following the global financial crisis, potentially because constrained monetary policy at the zero lower bound dampened exchange rate appreciation and thereby weakened the trade balance response

*Source: Authors’ calculations.*

### Annex 2. Empirical Approach for the Panel

### Annex 2. Empirical Approach for the Panel

### Vector Autoregression (PVAR) specification
- Baseline PVAR specification:
  - X_{i,t} = a_{i} + A(L) X_{i,t−1} + U_{i,t}
  - a_{i} is country fixed effects.
  - U_{i,t} is the error term.
  - X_{t} includes, in this order: the trade-weighted fiscal news shocks extracted from the US VAR in the previous section (US news shock); real GDP; the fiscal balance as a percentage of GDP; net exports of goods and services as a share of GDP; the long-term interest rates; and the real bilateral exchange rate.
- Variable transformations and units:
  - All variables enter in (logged) levels, with the exception of the surplus, trade balance, and bond yield, which are expressed in percent.
- Lag structure and estimator:
  - The benchmark specification includes four lags.
  - The PVAR methodology used is the least-squares dummy variable (LSDV) estimator as in Bun and Kiviet 2006.
- Additional notes:
  - As typical in this literature, various robustness checks were performed with respect to the effects of changing the ordering, and orthogonalized impulse-response functions were also analyzed.

### Sample and panel composition
- Panel composition and coverage:
  - An unbalanced panel comprising the top 30 US trading partners (23 advanced economies and 7 emerging market economies representing about 80 percent of US imports).
  - Sample period: from the fourth quarter of 1982 to the third quarter of 2016.
- Data limitations:
  - The panel is unbalanced because of limitations in data for some recipient countries’ macroeconomic variables.
- Additional country and data source information:
  - See Annex 3 for additional information on the countries and data sources.

### Construction of the US news shock for the panel (trade-weighted)
- Baseline scaling:
  - The SPF news shock is expressed as a 1 percent of US GDP shock.
- Weighting scheme for heterogeneity in exposure:
  - USNewsShock_{i,t} = w_{i,(t)}^{EXP} × USNewsShock_{t}^{US−VAR}
  - w_{i,(t)}^{EXP} is the ratio of country i’s exports to the United States to its total exports.
  - The intercountry linkages capture country i’s exposure to the US fiscal shock from exports from country i to the United States as a share of country i’s total exports.
- Rationale:
  - The scheme captures the idea that the US fiscal stimulus would have a larger impact on a recipient’s economy the stronger is the recipient’s trade link with respect to the United States.
  - Theoretical models posit that the US fiscal stimulus would increase US imports from other countries through both the demand channel and the price competitiveness channel.
  - Preferred weighting uses country i’s exports to the United States as a share of its total exports, which corresponds to US imports from country i.

### Data definitions and variable construction (PVAR variables)
- Fiscal news variable (SPFNEWS):
  - Based on SPF forecasts; calculated using the annualized percent change in mean responses for the real federal government consumption expenditure and gross investment reported by the Federal Reserve Bank of Philadelphia.
- FEDGOV, SUR, GDP, trade balance sources:
  - Real federal consumption expenditures and gross investment (FEDGOV) is the federal surplus divided by GDP; federal government budget surplus (SUR), real GDP (GDP), and the trade balance are retrieved from Federal Reserve Economic Data (FRED) at the Federal Reserve Bank of St Louis.
- Trade balance (variable construction used in the PVAR):
  - For variables in the PVAR, the trade balance is calculated as 100 × ((country i’s real exports to the United States) – (country i’s real imports from the United States))/(country i’s real GDP).
  - Nominal exports/imports have been deflated by the partner country’s export/import deflators.
  - Exports and imports data use an average of those reported by the United States and by its partners.
  - Results are robust when using the trade variables reported by the United States only.
- Real bilateral exchange rate:
  - Calculated as (nominal exchange rate) × (GDP deflator in country i)/(US GDP deflator), normalized to be 100 in 2010.
  - The nominal exchange rate in the PVAR is defined as (US dollar/national currency).

### Annex Table 3.1 — Data definitions and sources for the Panel Vector Autoregression
- SUR: Fiscal balance, percent of GDP — Quarterly — GDS — G_B_GDP
- STRATE: Short-term interest rate — Quarterly — GDS — FIDR
- 10YBOND: Long-term interest rate — Quarterly — GDS — FIGB
- RGDP: Real gross domestic product — Quarterly — GDS — NGDP_R
- TB: Trade balance — Quarterly — DOTS
- RER: Real bilateral exchange rate — Quarterly — DOTS
- Note: DOTS = Direction of Trade Statistics; GDS = Global Data Source.

*Annex 2. Empirical Approach for the Panel — spillovernote10*

---


_Source: https://www.imf.org/-/media/files/publications/spillovernotes/spillovernote10.pdf_