## ch1annex - Section 1

## Source details

**Canonical URL:** [ch1annex - Section 1](https://www.imf.org/-/media/files/publications/weo/2022/april/english/ch1annex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/weo/2022/april/english/ch1annex.pdf.md)
- [Structured JSON version](/-/media/files/publications/weo/2022/april/english/ch1annex.pdf.json)

---

### A1. Estimating the price elasticity of global oil and gas investment
- Regression specification (log-differences of CAPEX for firm type j):
  - CAPEX_{j,t} = γ_{j,0} + ∑_{l=0}^{p} γ_{j,l} p_{t−l} + ∑_{k=0}^{3} γ_{j,3+k} Deflator_{t−k} + β X_t
  - CAPEX_{j,t} = global annual capital expenditures (log-differences) in upstream oil and gas for firm type j (national oil companies, private, public, total).
  - p_{t−l} = weighted average of oil and gas prices (log-differences) in year t−l.
  - Deflator_{t} = log-difference of the deflator for the oil and gas sector (expenditure weighted-average of equipment and structures deflators for the US oil and gas drilling sector (US NIPA)).
  - X_t includes controls (10-year US Treasury yield in logs, world real GDP growth in log-differences, etc.).
  - Regression estimated separately for each of the 4 firm types; sample 1971-2020.
  - Oil and gas price = weighted average of WTI benchmark (in $/TJ) and Henry Hub natural gas price (in $/TJ).
  - Data on upstream oil and gas expenditure are from Rystad.
- Annex Table 1.SF.1 — Key estimated coefficients (robust standard errors in parentheses):
  - Oil and Gas Price (t)
    - NOC: 0.243** (0.0872)
    - Public: 0.314*** (0.0457)
    - Private: 0.387*** (0.0659)
    - Sum: 0.296*** (0.0522)
  - Oil and Gas Price t-1
    - NOC: -0.0340 (0.0972)
    - Public: 0.259** (0.0745)
    - Private: 0.140 (0.0870)
    - Sum: 0.103 (0.0713)
  - Oil and Gas Price t-2
    - NOC: 0.0500 (0.0532)
    - Public: 0.134* (0.0605)
    - Private: 0.0497 (0.0692)
    - Sum: 0.0816 (0.0446)
  - US NIPA Deflator (t)
    - NOC: 1.232*** (0.267)
    - Public: 0.695*** (0.167)
    - Private: 1.092*** (0.212)
    - Sum: 1.029*** (0.177)
  - US NIPA Deflator t-1
    - NOC: -0.181 (0.220)
    - Public: -0.442* (0.192)
    - Private: -0.485* (0.222)
    - Sum: -0.343 (0.181)
  - Constant
    - NOC: -0.00968 (0.0181)
    - Public: 0.0247 (0.0178)
    - Private: -0.00981 (0.0185)
    - Sum: 0.00234 (0.0139)
  - Number of Observations: 48 for each column.
  - R^2: NOC 0.61; Public 0.75; Private 0.73; Sum 0.76.
  - Adjusted R^2: NOC 0.56; Public 0.72; Private 0.69; Sum 0.74.
- Price elasticity measures (cumulative effects of 1 percentage point change in oil and gas prices on CAPEX):
  - Price Elasticity 0
    - NOC: 0.243
    - Public: 0.314
    - Private: 0.387
    - Sum: 0.296
  - Price Elasticity 0+L1
    - NOC: 0.209
    - Public: 0.573
    - Private: 0.526
    - Sum: 0.399
  - Price Elasticity 0+L1+L2
    - NOC: 0.259
    - Public: 0.706
    - Private: 0.576
    - Sum: 0.481
- Model and inference:
  - Model: OLS for all firm types.
  - Significance levels reported: *** p <0.001, ** p <0.01, * p <0.05.

### A2. Firm-level analysis of oil and gas capital investment
- Difference-in-differences regression (log capex in firm i, group s, year t):
  - y_{ist} = a + λ D_s + (β_1 C_t + β_2 P_oil,t) D_s + γ X_{ist} + ε_{ist}
  - y_{ist} = log capex; D_s = 1 for oil and gas firms, 0 for other sectors.
  - X includes: log total assets, debt to equity ratio, asset turnover, Altman distance to default, region, industry and year fixed effects.
  - C_t represents either a post-2015 time dummy or climate change indices (energy transition awareness, sustainable investment awareness, share of sustainable funds net inflows on global GFCF, share of GHG covered by regulations, average global price of CO2).
  - P_oil,t = Brent oil prices interacted with treatment dummy.
  - Data: annual, 2012–2020 from Compustat (global coverage for publicly traded firms); estimation stops in 2019 to exclude pandemic effects; firms with total asset value below USD 49M are dropped.
  - Parallel trends: total capital investment in oil & gas and rest of economy followed similar declining trend through 2016; divergence afterwards.
- Main empirical findings:
  - Capex in oil/gas firms between 2016 and 2019 was 35% lower than in firms in the rest of the economy (column 1), net of oil prices and absent the regime change.
  - Energy transition awareness: every 1 percentage point increase in public awareness on the energy transition is accompanied by a 1% reduction in capex in oil and gas firms vis-à-vis non-energy firms (column 2).
  - Sustainable investment awareness: every 1 percentage point increase in public awareness on sustainable investments is accompanied by a 0.6% reduction in capex in oil and gas firms vis-à-vis non-energy firms (column 3).
  - None of the “hard” proxies (share of GHG covered by regulations, average global price of CO2, sustainable funds net inflows) show significant effects, signaling little influence of enacted climate policies and portfolio choices in these specifications.
  - Results robust to a broader definition of the oil and gas sector (column 7).
- Scenario analysis (based on specification in column 2):
  - Fixing energy transition awareness proxy at its 2014 level and tracing average investment:
    - Capex in oil and gas firms would have been 38% higher in 2020 had public awareness on energy transition not taken off.
    - Holding oil prices at 2014 levels: oil and gas capex would have been 67% higher on average in 2020.
    - Holding sustainable portfolio choices at 2014 levels: oil and gas capex would have been 29% higher on average in 2020.
  - Caveat: proxies include few data points and are highly correlated, making separation of individual effects difficult.
- Annex Table 1.SF.2 — Selected coefficients (standard errors in parentheses):
  - D_s
    - Column 1: -1.085** (0.553)
    - Column 2: -2.180*** (0.485)
    - Column 3: -2.521*** (0.430)
    - Column 4: -2.600*** (0.429)
    - Column 5: -2.402*** (0.832)
    - Column 6: -2.425*** (0.470)
    - Column 7: -2.205*** (0.416)
  - D_s C_t
    - Column 1: -0.351*** (0.079)
    - Column 2: -0.010** (0.004)
    - Column 3: -0.006* (0.003)
    - Column 4: -0.233 (0.202)
    - Column 5: -0.012 (0.023)
    - Column 6: -0.097 (0.063)
    - Column 7: -0.010*** (0.004)
  - D_s * P_oil,t
    - Column 1: 0.371*** (0.122)
    - Column 2: 0.625*** (0.103)
    - Column 3: 0.675*** (0.097)
    - Column 4: 0.688*** (0.097)
    - Column 5: 0.668*** (0.142)
    - Column 6: 0.669*** (0.100)
    - Column 7: 0.571*** (0.089)
  - Controls (examples)
    - Distance to Default: 0.003 (0.002) across columns.
    - Log Total Assets: 1.054*** (0.003) across columns (1.053*** in column 7).
    - Log Asset Turnover: 0.174*** (0.013) across columns (0.150*** in column 7).
    - Leverage: -0.164* (0.088) across columns (-0.162* (0.083) in column 7).
  - Number of Observations: Columns 1–6: 40,378; Column 7: 41,149.
  - Note: All specifications include region, industry and year fixed effects.
  - Significance levels: *** p <0.01; ** p <0.05; * p <0.1.

### A3. Data and Methodology for the Structural Oil Price Scenarios
- A3.1 Data (historical annual data used):
  - Global real GDP: 1840–2007 from Stuermer and Schwerhoff (2015) (building on Maddison (2010)), extended to 2021 using IMF World Economic Outlook growth rates.
  - Global oil production: IEA database, 1973–2021 (includes crude oil, natural gas liquids and feedstocks).
    - Conversion: EJ to BBL/d using conversion factor 23.88 (EJ to M toe) *7.33/365 (see IEA, 2021c, p. 352).
  - Annual real oil price: British Petroleum (2021) for 1973–2020; 2021 value derived by applying annual growth rate of IMF’s oil price in its Commodity Prices System (growth rate based on average between 2020 and 2021 up until November 17).
    - Price reference: Arabian Light posted at Ras Tanura until 1983; Brent thereafter.
    - Inflation adjustment: U.S. all urban consumers price index.
  - IEA NZE Scenario: total oil production would decline roughly 60 percent (scenario premise: limit global temperature increases to 1.5°C in 2050).
- A3.2 Econometric Model:
  - VAR specification with three endogenous variables y_t = (REA_t, ΔQ_t, P_t)′:
    - REA_t = log of global real GDP.
    - ΔQ_t = percentage change of global oil production.
    - P_t = log of the real price of crude oil.
  - Reduced-form VAR:
    - y_t = A_1 y_{t−1} + … + A_p y_{t−p} + Π D_t + u_t with p = 4 and D_t = constant.
  - Structural form:
    - B_0 y_t = B_1 y_{t−1} + … + B_p y_{t−p} + Γ D_t + ε_t.
    - u_t = B_0^{-1} ε_t; E(ε_t ε_t′) = I_n; Σ_u = B_0^{-1} B_0^{-1}′.
- A3.3 Identification:
  - Sign restrictions on impact effects (Annex Table 1.SF.3):
    - Aggregate Demand Shock: Global Real GDP + ; Global Oil Production + ; Real Oil Price +
    - Oil Supply Shock: Global Real GDP + ; Global Oil Production + ; Real Oil Price -
    - Oil Specific Demand Shock: Global Real GDP - ; Global Oil Production + ; Real Oil Price +
  - Interpretation of shocks:
    - First shock = aggregate demand shock (positive shock increases global economic activity, global oil production, and real price).
    - Second shock = oil supply shock (positive shock increases production, up global activity, lowers real price on impact).
    - Third shock = oil-specific demand shock (captures energy transition and precautionary demand; positive shock increases production and oil price, decreases global economic output on impact due to price increase).
  - Narrative sign restrictions further sharpen identification (e.g., aggregate commodity demand shock was main downward driver of crude oil price during the Great Recession in 2009).
- A3.4 Structural scenario analysis:
  - Objective: conditional forecast y_{T+1,T+h} over h = 9 years with T = 2021, attributing future paths to specific structural shocks (framework of Antolin-Diaz et al., 2021).
  - Procedure:
    - Take oil production scenario as given (pre-specify oil quantities; set global oil production equal to global oil consumption; assume no short-term inventory changes).
    - Constrain aggregate demand shock and oil supply shock to their unconditional distributions and leave oil-specific demand shock unrestricted; algorithm finds oil-specific demand shocks that generate the required oil production path and derives implied price path.
    - Alternative baseline: constrain aggregate demand shock and oil-specific demand shock to unconditional distributions; leave oil supply shock unspecified.
- A3.5 Estimation and inference:
  - Bayesian estimation approach (Gibbs sampler) following Waggoner and Zha (1999), Rubio-Ramirez et al. (2010), Antolin-Diaz et al. (2021).
  - Procedure specifics:
    - Random draw of structural parameters out of 25,000 potential draws that rely on actual data and a structural forecast.
    - Use structural parameters from the randomly picked draw to draw scenario paths of price series and real GDP consistent with the specified oil production path.
    - Subsequent 25,000 draws rely on original data plus data from the drawn structural scenario.
    - Priors: Minnesota-type prior with standard shrinkage parameters (Giannone et al., 2015), sum-of-coefficients prior (Doan et al., 1984), dummy-initial-observation prior (Sims, 1993).
    - Prior assumption: oil production growth is i.i.d.; log of real GDP and logs of price levels follow a random walk.
  - Identification via sign restrictions yields sets of admissible parameters rather than point estimates:
    - For each model obtain a set of 1,000 admissible draws (each draw = conditional forecast, future shocks, associated B_0^{-1} matrix satisfying restrictions).
    - These draws used for inference; report pointwise median and percentiles of impulse responses for set-identified SVAR models.

### Section 2 — Econometrics, energy, and historical data sources
- Econometrics and VAR methods cited:
  - Econometric Reviews, 3(1):1–100.
  - Faust, J. (1998). The robustness of identified VAR conclusions about money. In Carnegie-Rochester Conference Series on Public Policy, volume 49, pages 207–244.
  - Giannone, D., Lenza, M., and Primiceri, G. (2015).  Prior selection for vector autoregressions. The Review of Economics and Statistics, 97(2):436–451.
  - Rubio-Ramirez, J. F., Waggoner, D. F., and Zha, T. (2010). Structural vector autoregressions:  Theory of identification and algorithms for inference. The Review of Economic Studies, 77(2):665–696.
  - Sims, C.  A.  (1993).   A nine-variable  probabilistic  macroeconomic  forecasting  model,  in business cycles, indicators and forecasting. NBER Studies in Business Cycles, In J.H. Stock and M.W. Watson (eds.) University of Chicago, pages 179–212.
  - Uhlig, H. (2005).  What are the effects of monetary policy on output?  Results from an agnostic identification procedure. Journal of Monetary Economics, 52(2):381–419.
  - Waggoner, D. and Zha, T. (1999).  Conditional forecasts in dynamic multivariate models. The Review of Economics and Statistics, 81(4):639–651.
- Energy, resources, and transition analyses cited:
  - IEA (2021a). Net zero by 2050. A roadmap for the global energy sector. International Energy Agency. Paris, France.
  - IEA (2021b). The role of critical minerals in clean energy transitions. world energy outlook special report.  International Energy Agency. Paris, France.
  - IEA (2021c). World energy outlook 2021. International Energy Agency. Paris, France.
  - Rystad Energy, UCube Database January 2022.
  - Stuermer, M. and Schwerhoff, G. (2015). Non-renewable resources, extraction technology, and endogenous growth. Dallas Fed Working Papers 1506 (Updated version: August2020), Federal Reserve Bank of Dallas.
- Historical and data sources:
  - Maddison, A. (2010).Historical Statistics of the world economy:1-2008  AD.  http://www.ggdc.net/maddison/ (accessed on June 13, 2011).

*Source: ch1annex - Section 1 (IMF, April 2022).*

### Section 1

### ch1annex - Section 1

### A1. Estimating the price elasticity of global oil and gas investment
- Regression specification (log-differences of CAPEX for firm type j):
  - CAPEX_{j,t} = γ_{j,0} + ∑_{l=0}^{p} γ_{j,l} p_{t−l} + ∑_{k=0}^{3} γ_{j,3+k} Deflator_{t−k} + β X_t
  - CAPEX_{j,t} = global annual capital expenditures (log-differences) in upstream oil and gas for firm type j (national oil companies, private, public, total).
  - p_{t−l} = weighted average of oil and gas prices (log-differences) in year t−l.
  - Deflator_{t} = log-difference of the deflator for the oil and gas sector (expenditure weighted-average of equipment and structures deflators for the US oil and gas drilling sector (US NIPA)).
  - X_t includes controls (10-year US Treasury yield in logs, world real GDP growth in log-differences, etc.).
  - Regression estimated separately for each of the 4 firm types; sample 1971-2020.
  - Oil and gas price = weighted average of WTI benchmark (in $/TJ) and Henry Hub natural gas price (in $/TJ).
  - Data on upstream oil and gas expenditure are from Rystad.

- Annex Table 1.SF.1 — Key estimated coefficients (robust standard errors in parentheses):
  - Oil and Gas Price (t)
    - NOC: 0.243** (0.0872)
    - Public: 0.314*** (0.0457)
    - Private: 0.387*** (0.0659)
    - Sum: 0.296*** (0.0522)
  - Oil and Gas Price t-1
    - NOC: -0.0340 (0.0972)
    - Public: 0.259** (0.0745)
    - Private: 0.140 (0.0870)
    - Sum: 0.103 (0.0713)
  - Oil and Gas Price t-2
    - NOC: 0.0500 (0.0532)
    - Public: 0.134* (0.0605)
    - Private: 0.0497 (0.0692)
    - Sum: 0.0816 (0.0446)
  - US NIPA Deflator (t)
    - NOC: 1.232*** (0.267)
    - Public: 0.695*** (0.167)
    - Private: 1.092*** (0.212)
    - Sum: 1.029*** (0.177)
  - US NIPA Deflator t-1
    - NOC: -0.181 (0.220)
    - Public: -0.442* (0.192)
    - Private: -0.485* (0.222)
    - Sum: -0.343 (0.181)
  - Constant
    - NOC: -0.00968 (0.0181)
    - Public: 0.0247 (0.0178)
    - Private: -0.00981 (0.0185)
    - Sum: 0.00234 (0.0139)
  - Number of Observations: 48 for each column.
  - R^2: NOC 0.61; Public 0.75; Private 0.73; Sum 0.76.
  - Adjusted R^2: NOC 0.56; Public 0.72; Private 0.69; Sum 0.74.

- Price elasticity measures (cumulative effects of 1 percentage point change in oil and gas prices on CAPEX):
  - Price Elasticity 0
    - NOC: 0.243
    - Public: 0.314
    - Private: 0.387
    - Sum: 0.296
  - Price Elasticity 0+L1
    - NOC: 0.209
    - Public: 0.573
    - Private: 0.526
    - Sum: 0.399
  - Price Elasticity 0+L1+L2
    - NOC: 0.259
    - Public: 0.706
    - Private: 0.576
    - Sum: 0.481

- Model: OLS for all firm types.
- Significance levels reported: *** p <0.001, ** p <0.01, * p <0.05.

### A2. Firm-level analysis of oil and gas capital investment
- Difference-in-differences regression (log capex in firm i, group s, year t):
  - y_{ist} = a + λ D_s + (β_1 C_t + β_2 P_oil,t) D_s + γ X_{ist} + ε_{ist}
  - y_{ist} = log capex; D_s = 1 for oil and gas firms, 0 for other sectors.
  - X includes: log total assets, debt to equity ratio, asset turnover, Altman distance to default, region, industry and year fixed effects.
  - C_t represents either a post-2015 time dummy or climate change indices (energy transition awareness, sustainable investment awareness, share of sustainable funds net inflows on global GFCF, share of GHG covered by regulations, average global price of CO2).
  - P_oil,t = Brent oil prices interacted with treatment dummy.
  - Data: annual, 2012–2020 from Compustat (global coverage for publicly traded firms); estimation stops in 2019 to exclude pandemic effects; firms with total asset value below USD 49M are dropped.
  - Parallel trends: total capital investment in oil & gas and rest of economy followed similar declining trend through 2016; divergence afterwards.

- Main empirical findings:
  - Capex in oil/gas firms between 2016 and 2019 was 35% lower than in firms in the rest of the economy (column 1), net of oil prices and absent the regime change.
  - Energy transition awareness: every 1 percentage point increase in public awareness on the energy transition is accompanied by a 1% reduction in capex in oil and gas firms vis-à-vis non-energy firms (column 2).
  - Sustainable investment awareness: every 1 percentage point increase in public awareness on sustainable investments is accompanied by a 0.6% reduction in capex in oil and gas firms vis-à-vis non-energy firms (column 3).
  - None of the “hard” proxies (share of GHG covered by regulations, average global price of CO2, sustainable funds net inflows) show significant effects, signaling little influence of enacted climate policies and portfolio choices in these specifications.
  - Results robust to a broader definition of the oil and gas sector (column 7).

- Scenario analysis (based on specification in column 2):
  - Fixing energy transition awareness proxy at its 2014 level and tracing average investment:
    - Capex in oil and gas firms would have been 38% higher in 2020 had public awareness on energy transition not taken off.
    - Holding oil prices at 2014 levels: oil and gas capex would have been 67% higher on average in 2020.
    - Holding sustainable portfolio choices at 2014 levels: oil and gas capex would have been 29% higher on average in 2020.
  - Caveat: proxies include few data points and are highly correlated, making separation of individual effects difficult.

- Annex Table 1.SF.2 — Selected coefficients (standard errors in parentheses):
  - D_s
    - Column 1: -1.085** (0.553)
    - Column 2: -2.180*** (0.485)
    - Column 3: -2.521*** (0.430)
    - Column 4: -2.600*** (0.429)
    - Column 5: -2.402*** (0.832)
    - Column 6: -2.425*** (0.470)
    - Column 7: -2.205*** (0.416)
  - D_s C_t
    - Column 1: -0.351*** (0.079)
    - Column 2: -0.010** (0.004)
    - Column 3: -0.006* (0.003)
    - Column 4: -0.233 (0.202)
    - Column 5: -0.012 (0.023)
    - Column 6: -0.097 (0.063)
    - Column 7: -0.010*** (0.004)
  - D_s * P_oil,t
    - Column 1: 0.371*** (0.122)
    - Column 2: 0.625*** (0.103)
    - Column 3: 0.675*** (0.097)
    - Column 4: 0.688*** (0.097)
    - Column 5: 0.668*** (0.142)
    - Column 6: 0.669*** (0.100)
    - Column 7: 0.571*** (0.089)
  - Controls (examples)
    - Distance to Default: 0.003 (0.002) across columns.
    - Log Total Assets: 1.054*** (0.003) across columns (1.053*** in column 7).
    - Log Asset Turnover: 0.174*** (0.013) across columns (0.150*** in column 7).
    - Leverage: -0.164* (0.088) across columns (-0.162* (0.083) in column 7).
  - Number of Observations: Columns 1–6: 40,378; Column 7: 41,149.
  - Note: All specifications include region, industry and year fixed effects.
  - Significance levels: *** p <0.01; ** p <0.05; * p <0.1.

### A3. Data and Methodology for the Structural Oil Price Scenarios
A3.1 Data
- Historical annual data used:
  - Global real GDP: 1840–2007 from Stuermer and Schwerhoff (2015) (building on Maddison (2010)), extended to 2021 using IMF World Economic Outlook growth rates.
  - Global oil production: IEA database, 1973–2021 (includes crude oil, natural gas liquids and feedstocks).
    - Conversion: EJ to BBL/d using conversion factor 23.88 (EJ to M toe) *7.33/365 (see IEA, 2021c, p. 352).
  - Annual real oil price: British Petroleum (2021) for 1973–2020; 2021 value derived by applying annual growth rate of IMF’s oil price in its Commodity Prices System (growth rate based on average between 2020 and 2021 up until November 17).
    - Price reference: Arabian Light posted at Ras Tanura until 1983; Brent thereafter.
    - Inflation adjustment: U.S. all urban consumers price index.
  - IEA NZE Scenario: total oil production would decline roughly 60 percent (scenario premise: limit global temperature increases to 1.5°C in 2050).

A3.2 Econometric Model
- VAR specification with three endogenous variables y_t = (REA_t, ΔQ_t, P_t)′:
  - REA_t = log of global real GDP.
  - ΔQ_t = percentage change of global oil production.
  - P_t = log of the real price of crude oil.
- Reduced-form VAR:
  - y_t = A_1 y_{t−1} + … + A_p y_{t−p} + Π D_t + u_t with p = 4 and D_t = constant.
- Structural form:
  - B_0 y_t = B_1 y_{t−1} + … + B_p y_{t−p} + Γ D_t + ε_t.
  - u_t = B_0^{-1} ε_t; E(ε_t ε_t′) = I_n; Σ_u = B_0^{-1} B_0^{-1}′.

A3.3 Identification
- Sign restrictions on impact effects (Annex Table 1.SF.3):
  - Aggregate Demand Shock: Global Real GDP + ; Global Oil Production + ; Real Oil Price +
  - Oil Supply Shock: Global Real GDP + ; Global Oil Production + ; Real Oil Price -
  - Oil Specific Demand Shock: Global Real GDP - ; Global Oil Production + ; Real Oil Price +
- Interpretation of shocks:
  - First shock = aggregate demand shock (positive shock increases global economic activity, global oil production, and real price).
  - Second shock = oil supply shock (positive shock increases production, up global activity, lowers real price on impact).
  - Third shock = oil-specific demand shock (captures energy transition and precautionary demand; positive shock increases production and oil price, decreases global economic output on impact due to price increase).
- Narrative sign restrictions further sharpen identification (e.g., aggregate commodity demand shock was main downward driver of crude oil price during the Great Recession in 2009).

A3.4 Structural scenario analysis
- Objective: conditional forecast y_{T+1,T+h} over h = 9 years with T = 2021, attributing future paths to specific structural shocks (framework of Antolin-Diaz et al., 2021).
- Procedure:
  - Take oil production scenario as given (pre-specify oil quantities; set global oil production equal to global oil consumption; assume no short-term inventory changes).
  - Constrain aggregate demand shock and oil supply shock to their unconditional distributions and leave oil-specific demand shock unrestricted; algorithm finds oil-specific demand shocks that generate the required oil production path and derives implied price path.
  - Alternative baseline: constrain aggregate demand shock and oil-specific demand shock to unconditional distributions; leave oil supply shock unspecified.

A3.5 Estimation and inference
- Bayesian estimation approach (Gibbs sampler) following Waggoner and Zha (1999), Rubio-Ramirez et al. (2010), Antolin-Diaz et al. (2021).
- Procedure specifics:
  - Random draw of structural parameters out of 25,000 potential draws that rely on actual data and a structural forecast.
  - Use structural parameters from the randomly picked draw to draw scenario paths of price series and real GDP consistent with the specified oil production path.
  - Subsequent 25,000 draws rely on original data plus data from the drawn structural scenario.
  - Priors: Minnesota-type prior with standard shrinkage parameters (Giannone et al., 2015), sum-of-coefficients prior (Doan et al., 1984), dummy-initial-observation prior (Sims, 1993).
  - Prior assumption: oil production growth is i.i.d.; log of real GDP and logs of price levels follow a random walk.
- Identification via sign restrictions yields sets of admissible parameters rather than point estimates:
  - For each model obtain a set of 1,000 admissible draws (each draw = conditional forecast, future shocks, associated B_0^{-1} matrix satisfying restrictions).
  - These draws used for inference; report pointwise median and percentiles of impulse responses for set-identified SVAR models.

*Source: ch1annex - Section 1 (IMF, April 2022).*

### Section 2

### ch1annex - Section 2

### Econometrics and VAR methods
- Econometric Reviews, 3(1):1–100.
- Faust, J. (1998). The robustness of identified VAR conclusions about money. In Carnegie-Rochester Conference Series on Public Policy, volume 49, pages 207–244.
- Giannone, D., Lenza, M., and Primiceri, G. (2015).  Prior selection for vector autoregressions. The Review of Economics and Statistics, 97(2):436–451.
- Rubio-Ramirez, J. F., Waggoner, D. F., and Zha, T. (2010). Structural vector autoregressions:  Theory of identification and algorithms for inference. The Review of Economic Studies, 77(2):665–696
- Sims, C.  A.  (1993).   A nine-variable  probabilistic  macroeconomic  forecasting  model,  in business cycles, indicators and forecasting. NBER Studies in Business Cycles, In J.H. Stock and M.W. Watson (eds.) University of Chicago, pages 179–212.
- Uhlig, H. (2005).  What are the effects of monetary policy on output?  Results from an agnostic identification procedure. Journal of Monetary Economics, 52(2):381–419
- Waggoner, D. and Zha, T. (1999).  Conditional forecasts in dynamic multivariate models. The Review of Economics and Statistics, 81(4):639–651

### Energy, resources, and transition analyses
- IEA (2021a). Net zero by 2050. A roadmap for the global energy sector. International Energy Agency. Paris, France.
- IEA (2021b). The role of critical minerals in clean energy transitions. world energy outlook special report.  International Energy Agency. Paris, France.
- IEA (2021c). World energy outlook 2021. International Energy Agency. Paris, France.
- Rystad Energy, UCube Database January 2022.
- Stuermer, M. and Schwerhoff, G. (2015). Non-renewable resources, extraction technology, and endogenous growth. Dallas Fed Working Papers 1506 (Updated version: August2020), Federal Reserve Bank of Dallas.

### Historical and data sources
- Maddison, A. (2010).Historical Statistics of the world economy:1-2008  AD.  http://www.ggdc.net/maddison/ (accessed on June 13, 2011).

*Source: ch1annex - Section 2 (ch1annex - Section 2, https://www.imf.org/-/media/files/publications/weo/2022/april/english/ch1annex.pdf)*

---


_Source: https://www.imf.org/-/media/files/publications/weo/2022/april/english/ch1annex.pdf_