## ch2annex - Annex Table 2.4.1 for a detailed list of countries included in the analysis). A country is classified as a net energy

## Source details

**Canonical URL:** [ch2annex - Annex Table 2.4.1 for a detailed list of countries included in the analysis). A country is classified as a net energy](https://www.imf.org/-/media/files/publications/esr/2024/english/ch2annex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/esr/2024/english/ch2annex.pdf.md)
- [Structured JSON version](/-/media/files/publications/esr/2024/english/ch2annex.pdf.json)

---

### Methodology: Time series and panel local projections
- Time series LP-IV specification (no country fixed effects) estimates cumulative change in oil production or global industrial production over the t−1 to t+ h horizon using structural shocks from Baumeister and Hamilton (2019).
- Rolling window time series LP-IV:
  - Model estimated at monthly frequency with a 36-month window over the sample period from January 1996 to May 2023.
  - Robustness checks use 42-month and 48-month windows.
- State-dependent panel local projections:
  - Binary-state specification separates sample by an indicator of predetermined country characteristics or policies; coefficient 훽 captures impacts of negative oil supply shock that increase real energy prices by 10 percent on impact when the state variable I = 1; 훽′ captures impacts when I = 0.
  - Continuous-state specification interacts energy price changes with the continuous state variable V; the term 훽 + 훽′ V captures the impact of negative oil supply shocks.
  - Impacts evaluated at the state variable’s 25th and 75th percentiles.

### COVID-19 adjustment and data treatment
- COVID-19 period treated as a separate regime with larger variance following Lenza and Primiceri (2022) simplified by Hamilton (2022).
- Observations during the COVID-19 period are down-weighted by inverse-variance weighting.
- Methodology noted as robust to alternative “decovidize” approaches such as Ng (2021).

### Country grouping and median net energy trade shares (sample period 1996:Q1 to 2023:Q2)
- Baseline grouping: countries classified as energy exporters (median net energy export share > 0) or importers (median net energy export share < 0).
- Energy Exporters (country — Median Net Energy Trade Share, Percent):
  - Saudi Arabia 58.4
  - Russia 37.7
  - Norway 36.5
  - Colombia 17.2
  - Australia 5.9
  - Canada 5.8
  - Indonesia 5.3
  - Mexico 2.7
  - Malaysia 2.6
  - Argentina 1.1
  - Denmark 0.1
- Energy Importers (country — Median Net Energy Trade Share, Percent):
  - Pakistan -14.4
  - India -13.4
  - South Africa -3.5
  - Poland -3.4
  - Brazil -3.2
  - Germany -3.0
  - Austria -2.9
  - Czech Republic -2.7
  - Peru -2.6
  - Romania -2.4
  - Hungary -2.4
  - Finland -2.4
  - Sweden -2.3
  - Belgium -2.2
  - Ireland -1.8
  - Switzerland -1.7
  - The Netherlands -1.4
  - United Kingdom -1.3
  - United States -5.2
  - Spain -4.8
  - New Zealand -4.5
  - France -4.4
  - Portugal -4.2
  - Italy -4.0
  - China -3.9
  - Japan -9.2
  - Türkiye -9.4
  - Greece -7.6
  - Korea -7.4
  - Chile -7.0
  - Philippines -5.9
  - Israel -5.5
  - Thailand -5.5
- Aggregate summary:
  - Mean 15.8 (Exporters) and -4.7 (Importers)
  - Median 5.8 (Exporters) and -3.9 (Importers)
- Note on grouping robustness:
  - Re-estimations using top 25th percentile (large net energy importers) and bottom 25th percentile (small importers) show larger (smaller) responses in current account, consumption, and output relative to the sample of all importers.
  - Impulse responses remain statistically significant for the sample of small importers.
  - State-dependent local projections using net energy import shares as the state variable produce impacts broadly proportional to importers’ net energy imports; comparable results obtained when splitting sample into small (bottom 25 percentile) and big importers (top 25 percentile).

### Shock types, calibration, and confidence intervals
- Shocks considered (from Baumeister and Hamilton (2019)): oil supply shocks, global activity shocks, oil consumption demand shocks, oil inventory demand shocks.
- General shock calibration used in impulse responses: effects of a shock that increases real energy price by 10 percent on impact.
- Confidence intervals reported for impulse responses: 68 and 90 percent.
- Global activity shocks and oil supply shocks: Main chapter focuses on oil supply and global activity shocks for importers and exporters.
- Oil consumption demand and oil inventory demand shocks: Impacts on oil production and global industrial production examined; impulse responses reported for real energy price, oil production, and global industrial production. Effects on energy exporters and importers (current account, real GDP, real private consumption) mirror those from global activity shocks but with approximately half the magnitude.

### Robustness checks on shock identification
- Alternative shock identification approaches used for robustness:
  - Baumeister and Hamilton (2019) (baseline) and Baumeister and Hamilton (2023) (BH’23).
  - Känzig (2021) oil supply news shock (leverages OPEC institutional features and high-frequency data).
  - Updated Kilian and Murphy (2014) methodology per Zhou (2020) (uses global industrial production index); identifies oil supply shock, aggregate demand shock, oil speculative demand shock.
- Findings from robustness checks:
  - Using Känzig (2021) oil supply news shock yields consistent impulse response dynamics across external and real variables (current account, net international investment position, real GDP, real consumption).
  - The updated Kilian and Murphy approach produces similar impulse responses to the BH’19 baseline for the oil supply and global activity shocks.
- Correlations between quarterly aggregated shock series (summary notes):
  - Sample runs from 1961:Q1 to 2023:Q2, except the BH’23 shock runs until 2020:Q1.
  - upd. KM’14 follows Zhou (2020) and uses the global industrial production index instead of the global real economic activity index.
  - All shocks normalized to be associated with an increase in the real energy price.

### Exporters' Real Output — identification and main impulse-response structure
- Shock identification: Kilian and Murphy (2014) approach and updates (Zhou 2020); robustness with Baumeister and Hamilton (2023) granular IV approach.
- Shock size: 10 percent increase in real energy price on impact.
- Confidence intervals: 68 and 90 percent.
- Outcomes assessed for exporters and importers include:
  - Real energy price (Percent change)
  - Global oil production (Percent change)
  - Global industrial production (Percent change)
  - Real Output (Percent change)
  - Real Consumption (Percent change)
  - Current Account (Percent of GDP)
  - NIIP (Percent change)
- Government debt scenarios evaluated at the 75th and 25th percentiles (labeled high and low government debt).
- External positions considered strong if IMF staff current account gap > -1.
- Exchange rate regimes classified with refined categories 1, 2, 12, 13 as "more flexible ER regime"; other categories as "less flexible ER regime".
- State-variable percentiles explicitly used for evaluation: 25th and 75th percentiles for continuous measures (e.g., government debt, inflation anchoring, energy dependence).

### Global Financial Conditions decomposition and transmission
- Global financial conditions proxied by Moody's seasoned BAA corporate bond yield relative to 10-year treasury constant maturity (BAA spread).
- Decomposition approach:
  - Regress BAA spread on US monetary policy shocks from Bu, Rogers and Wu (2021) updated by Ugazio and Xin (2024).
  - Fitted value = portion attributable to changes in US monetary policy.
  - Residual = portion that could reflect changes in global risk appetite.
- Key transmission distinctions:
  - When elevated BAA spreads arise from tighter US monetary policy: decline in importers’ consumption, investment, and output is more gradual (slower transmission).
  - When elevated BAA spreads arise from low global risk appetite: importers experience an immediate decline in real consumption, investment, and GDP following a negative oil supply shock that increases real energy price.
- Impulse responses shown for:
  - High vs Low BAA spread due to global risk
  - High vs Low BAA spread driven by US monetary policy
- Outcomes reported include Current Account (Percent of GDP), Real Output (Percent), and Private Inflows (Percent of lagged total liabilities).

### Country characteristics, state variables, and correlations
- State-dependent local projections use country characteristics and policy variables with low pairwise correlations, enabling multi-dimensional state analysis.
- State variables and thresholds:
  - Investment in energy-exporting countries: Foreign direct investment in Saudi Arabia, evaluated at Median (binary threshold).
  - Exchange rate regime: Ilzetzki, Reinhart, and Rogoff (2019) fine classification; Freely floating defined as categories 1, 2, 12, 13; other regimes 3–11.
  - External positions: IMF staff current account gap; "Strong if current account gap > -1" (binary).
  - Government debt: Government debt to GDP from IMF Global Data Source (GDS); evaluated at 25th and 75th percentiles (continuous).
  - Inflation Anchoring: Monetary policy credibility measures from Bems and others (2021); evaluated at 25th and 75th percentiles (continuous).
  - Energy dependence: Country's net energy trade balance as a share of GDP; evaluated at 25th and 75th percentiles (continuous).
- Correlations matrix excerpts:
  - Government debt with Energy dependence: -0.08
  - Government debt with Investment in energy-exporting countries: -0.13
  - Government debt with Inflation anchoring: -0.02
  - Energy dependence with Investment in energy-exporting countries: -0.02
  - Investment in energy-exporting countries with Inflation anchoring: -0.10
  - Exchange rate regime correlations with others: small values (e.g., with Investment in energy-exporting countries: 0.22)
  - External positions correlations: Government debt 0.01; Energy dependence 0.14; Investment in energy-exporting countries 0.36; Inflation anchoring 0.06; Exchange rate regime 0.01

### Heterogeneous policy effects and buffers
- Exchange rate flexibility:
  - More flexible ER regime vs Less flexible ER regime affect Current Account (Percent of GDP), Real Output (Percent), and Reserve Asset (Percent of GDP) responses to a 10 percent oil-supply-driven energy price increase.
- Government debt:
  - High government debt (75th percentile) vs Low government debt (25th percentile) evaluated for Current Account, Real Output, Reserve Asset, and Net Financial Flow (Percent of GDP).
- External positions:
  - Strong CA positions vs Weaker CA positions compared for Current Account, Real Output, and Net Financial Flow (Percent of GDP).
- Findings indicate policy and structural buffers (reserve assets, flexible ER regimes, stronger CA positions, lower government debt) modulate the adverse impact of oil supply shocks on output, current account, and financial flows.

### Oil price–US dollar relationship
- Rolling-window instrumental-variable local projections show the US dollar response to a 10 percent oil price increase driven by an oil supply shock.
- Analysis period for one set of estimates: 2020:M1 to 2023:M5.
- Figures compare baseline rolling regression to specifications controlling for the oil trade balance and for global risk aversion (residuals from regressing BAA spreads on US monetary policy shocks).
- Rolling-window panels shown for positive correlation periods in the 1970s, 1980s, 1990s, and 2020s; controlling for global risk aversion or oil trade balance shifts the estimated dollar response patterns.

*Source: IMF staff calculations and annex material (External Sector Report Chapter 2, online annexes and figures as provided).*

### Annex Table 2.4.1 for a detailed list of countries included in the analysis). A country is classified as a net energy

### ch2annex - Annex Table 2.4.1 for a detailed list of countries included in the analysis). A country is classified as a net energy

### Methodology: Time series and panel local projections
- Time series LP-IV specification (no country fixed effects) estimates cumulative change in oil production or global industrial production over the t−1 to t+ h horizon using structural shocks from Baumeister and Hamilton (2019).
- Rolling window time series LP-IV:
  - Model estimated at monthly frequency with a 36-month window over the sample period from January 1996 to May 2023.
  - Robustness checks use 42-month and 48-month windows.
- State-dependent panel local projections:
  - Binary-state specification separates sample by an indicator of predetermined country characteristics or policies; coefficient 훽 captures impacts of negative oil supply shock that increase real energy prices by 10 percent on impact when the state variable I = 1; 훽′ captures impacts when I = 0.
  - Continuous-state specification interacts energy price changes with the continuous state variable V; the term 훽 + 훽′ V captures the impact of negative oil supply shocks.
  - Impacts evaluated at the state variable’s 25th and 75th percentiles.

### COVID-19 adjustment and data treatment
- COVID-19 period treated as a separate regime with larger variance following Lenza and Primiceri (2022) simplified by Hamilton (2022).
- Observations during the COVID-19 period are down-weighted by inverse-variance weighting.
- Methodology noted as robust to alternative “decovidize” approaches such as Ng (2021).

### Country grouping and median net energy trade shares (sample period 1996:Q1 to 2023:Q2)
- Baseline grouping: countries classified as energy exporters (median net energy export share > 0) or importers (median net energy export share < 0).
- Summary statistics reported for sample countries:
  - Energy Exporters (country — Median Net Energy Trade Share, Percent):
    - Saudi Arabia 58.4
    - Russia 37.7
    - Norway 36.5
    - Colombia 17.2
    - Australia 5.9
    - Canada 5.8
    - Indonesia 5.3
    - Mexico 2.7
    - Malaysia 2.6
    - Argentina 1.1
    - Denmark 0.1
  - Energy Importers (country — Median Net Energy Trade Share, Percent):
    - Pakistan -14.4
    - India -13.4
    - South Africa -3.5
    - Poland -3.4
    - Brazil -3.2
    - Germany -3.0
    - Austria -2.9
    - Czech Republic -2.7
    - Peru -2.6
    - Romania -2.4
    - Hungary -2.4
    - Finland -2.4
    - Sweden -2.3
    - Belgium -2.2
    - Ireland -1.8
    - Switzerland -1.7
    - The Netherlands -1.4
    - United Kingdom -1.3
    - United States -5.2
    - Spain -4.8
    - New Zealand -4.5
    - France -4.4
    - Portugal -4.2
    - Italy -4.0
    - China -3.9
    - Japan -9.2
    - Türkiye -9.4
    - Greece -7.6
    - Korea -7.4
    - Chile -7.0
    - Philippines -5.9
    - Israel -5.5
    - Thailand -5.5
  - Aggregate summary:
    - Mean 15.8 (Exporters) and -4.7 (Importers)
    - Median 5.8 (Exporters) and -3.9 (Importers)
- Note on grouping robustness:
  - Re-estimations using top 25th percentile (large net energy importers) and bottom 25th percentile (small importers) show larger (smaller) responses in current account, consumption, and output relative to the sample of all importers.
  - Impulse responses remain statistically significant for the sample of small importers.
  - State-dependent local projections using net energy import shares as the state variable produce impacts broadly proportional to importers’ net energy imports; comparable results obtained when splitting sample into small (bottom 25 percentile) and big importers (top 25 percentile).

### Key empirical findings on shocks and impacts
- Shocks considered (from Baumeister and Hamilton (2019)): oil supply shocks, global activity shocks, oil consumption demand shocks, oil inventory demand shocks.
- General shock calibration used in impulse responses: effects of a shock that increases real energy price by 10 percent on impact.
- Global activity shocks and oil supply shocks:
  - Main chapter focuses on oil supply and global activity shocks for importers and exporters.
- Oil consumption demand and oil inventory demand shocks:
  - Impacts on oil production and global industrial production examined; impulse responses reported for real energy price, oil production, and global industrial production.
  - Effects on energy exporters and importers (current account, real GDP, real private consumption) mirror those from global activity shocks but with approximately half the magnitude.
- Confidence intervals reported for impulse responses: 68 and 90 percent.

### Robustness checks on shock identification
- Alternative shock identification approaches used for robustness:
  - Baumeister and Hamilton (2019) (baseline) and Baumeister and Hamilton (2023) (BH’23).
  - Känzig (2021) oil supply news shock (leverages OPEC institutional features and high-frequency data).
  - Updated Kilian and Murphy (2014) methodology per Zhou (2020) (uses global industrial production index); identifies oil supply shock, aggregate demand shock, oil speculative demand shock.
- Findings from robustness checks:
  - Using Känzig (2021) oil supply news shock yields consistent impulse response dynamics across external and real variables (current account, net international investment position, real GDP, real consumption).
  - The updated Kilian and Murphy approach produces similar impulse responses to the BH’19 baseline for the oil supply and global activity shocks.
- Correlations between quarterly aggregated shock series (summary notes):
  - Sample runs from 1961:Q1 to 2023:Q2, except the BH’23 shock runs until 2020:Q1.
  - upd. KM’14 follows Zhou (2020) and uses the global industrial production index instead of the global real economic activity index.
  - All shocks normalized to be associated with an increase in the real energy price.
  - (Correlations table reported in the source; alternative identification approaches show varying degrees of correlation across oil supply and global activity shock series.)

*Source: IMF staff calculations and annex material (External Sector Report Chapter 2, online annexes and figures as provided).*

### 3. Exporters' Real Output

### ch2annex - 3. Exporters' Real Output

### Identification and Shock Magnitudes
- Oil supply and aggregate demand shocks are identified through the Kilian and Murphy (2014) approach and its updates (Zhou 2020).
- Robustness checks also use shocks identified through Baumeister and Hamilton (2023) granular IV approach.
- All impulse responses report the effects of shocks that increase real energy price by 10 percent on impact.
- Confidence intervals reported: 68 and 90 percent.

### Main Impulse-Response Findings for Energy Exporters and Importers
- Exporters and importers are analyzed separately; panels depict exporters (panels 1–4) and importers (panels 5–8).
- Outcomes assessed include:
  - Real energy price (Percent change)
  - Global oil production (Percent change)
  - Global industrial production (Percent change)
  - Exporters' and importers' Real Output (Percent change)
  - Exporters' and importers' Real Consumption (Percent change)
  - Exporters' and importers' Current Account (Percent of GDP)
  - Exporters' and importers' NIIP (Percent change)
- Results are consistent across alternative identification strategies (Kilian and Murphy variants and Baumeister and Hamilton granular IV), with Baumeister and Hamilton (2023) confirming consistency with Baumeister and Hamilton (2019).

### Quantitative Characteristics of Responses (as reported)
- Shock size: 10 percent increase in real energy price on impact.
- Confidence intervals: 68 and 90 percent.
- Government debt scenarios evaluated at the 75th and 25th percentiles (labeled high and low government debt).
- External positions considered strong if IMF staff current account gap > -1.
- Exchange rate regimes classified with refined categories 1, 2, 12, 13 as "more flexible ER regime"; other categories as "less flexible ER regime".
- State-variable percentiles explicitly used for evaluation: 25th and 75th percentiles for continuous measures (e.g., government debt, inflation anchoring, energy dependence).

### Robustness Checks and Alternative Identification
- Kilian and Murphy (2014) approach (with Zhou 2020 updates) used to separate oil supply vs aggregate demand shocks.
- Granular instrumental variables approach (Baumeister and Hamilton 2023) used to identify oil demand and supply shocks exploiting idiosyncratic shocks of large firms/regions; produces impulse responses consistent with prior Baumeister and Hamilton results.
- Robustness figures examine effects on:
  - Real energy price, global oil production, global industrial production
  - Exporters' and importers' current accounts, real output, NIIP, and real consumption
- All robustness checks maintain the 10 percent shock specification and report 68 and 90 percent confidence intervals.

### Global Financial Conditions Decomposition and Transmission
- Global financial conditions proxied by Moody's seasoned BAA corporate bond yield relative to 10-year treasury constant maturity (BAA spread).
- Decomposition approach:
  - Regress BAA spread on US monetary policy shocks from Bu, Rogers and Wu (2021) updated by Ugazio and Xin (2024).
  - Fitted value = portion attributable to changes in US monetary policy.
  - Residual = portion that could reflect changes in global risk appetite.
- Empirical sample excludes the United States in some specifications to reduce US performance influence.
- Key distinction in transmission:
  - When elevated BAA spreads arise from tighter US monetary policy: decline in importers’ consumption, investment, and output is more gradual (slower transmission).
  - When elevated BAA spreads arise from low global risk appetite: importers experience an immediate decline in real consumption, investment, and GDP following a negative oil supply shock that increases real energy price.
- Impulse responses shown for:
  - High vs Low BAA spread due to global risk
  - High vs Low BAA spread driven by US monetary policy
- Outcomes reported include Current Account (Percent of GDP), Real Output (Percent), and Private Inflows (Percent of lagged total liabilities).

### Country Characteristics, Policy Variables, and Heterogeneity
- State-dependent local projections use country characteristics and policy variables with low pairwise correlations, enabling multi-dimensional state analysis.
- State variables and thresholds:
  - Investment in energy-exporting countries: Foreign direct investment in Saudi Arabia, evaluated at Median (binary threshold).
  - Exchange rate regime: Ilzetzki, Reinhart, and Rogoff (2019) fine classification; Freely floating defined as categories 1, 2, 12, 13; other regimes 3–11.
  - External positions: IMF staff current account gap; "Strong if current account gap > -1" (binary).
  - Government debt: Government debt to GDP from IMF Global Data Source (GDS); evaluated at 25th and 75th percentiles (continuous).
  - Inflation Anchoring: Monetary policy credibility measures from Bems and others (2021); evaluated at 25th and 75th percentiles (continuous).
  - Energy dependence: Country's net energy trade balance as a share of GDP; evaluated at 25th and 75th percentiles (continuous).
- Correlations matrix excerpts (preserved values):
  - Government debt with Energy dependence: -0.08
  - Government debt with Investment in energy-exporting countries: -0.13
  - Government debt with Inflation anchoring: -0.02
  - Energy dependence with Investment in energy-exporting countries: -0.02
  - Investment in energy-exporting countries with Inflation anchoring: -0.10
  - Exchange rate regime correlations with others: small values (e.g., with Investment in energy-exporting countries: 0.22)
  - External positions correlations: Government debt 0.01; Energy dependence 0.14; Investment in energy-exporting countries 0.36; Inflation anchoring 0.06; Exchange rate regime 0.01

### Heterogeneous Policy Effects from Country Characteristics
- Exchange rate flexibility:
  - More flexible ER regime vs Less flexible ER regime affect Current Account (Percent of GDP), Real Output (Percent), and Reserve Asset (Percent of GDP) responses to a 10 percent oil-supply-driven energy price increase.
- Government debt:
  - High government debt (75th percentile) vs Low government debt (25th percentile) evaluated for Current Account, Real Output, Reserve Asset, and Net Financial Flow (Percent of GDP).
- External positions:
  - Strong CA positions vs Weaker CA positions compared for Current Account, Real Output, and Net Financial Flow (Percent of GDP).
- Results indicate policy and structural buffers (reserve assets, flexible ER regimes, stronger CA positions, lower government debt) modulate the adverse impact of oil supply shocks on output, current account, and financial flows.

### Oil Price–US Dollar Relationship
- Rolling-window instrumental-variable local projections show the US dollar response to a 10 percent oil price increase driven by an oil supply shock.
- Analysis period for one set of estimates: 2020:M1 to 2023:M5.
- Figures compare baseline rolling regression to specifications controlling for the oil trade balance and for global risk aversion (residuals from regressing BAA spreads on US monetary policy shocks).
- Rolling-window panels shown for positive correlation periods in the 1970s, 1980s, 1990s, and 2020s; controlling for global risk aversion or oil trade balance shifts the estimated dollar response patterns.

*Sources: Kilian and Murphy (2014); Zhou (2020); Baumeister and Hamilton (2023); Juvenal and Petrella (2024); Bu, Rogers and Wu (2021) updated by Ugazio and Xin (2024); Baumeister and Hamilton (2019); IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/esr/2024/english/ch2annex.pdf_