## _wp15209

## Source details

**Canonical URL:** [_wp15209](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15209.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15209.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15209.pdf.json)

---

### Introduction and motivation
- Objective: Provide empirical evidence of the effect of news shocks on the current account, saving, investment, GDP, consumption, and employment using plausibly exogenous variation in the timing of worldwide giant oil discoveries as an observable measure of news shocks about higher future output.
- Key empirical sample: 1970–2012 for up to 130 countries.
- Definition: A giant oil discovery is a discovery of an oil and/or gas field that contains at least a total of 500 million barrels of ultimately recoverable oil equivalent.
- Typical discovery-to-production lag: 4 to 6 years.
- Median value of constructed NPV as a percentage of country’s GDP: about 9 percent.
- Historical illustrative case: Norway — borrowing to build North Sea production facilities after discoveries in the late 1960s and early 1970s; saving rate declined; current account deficit approached minus 15 percent of GDP at its trough in 1977; current account improved after massive oil exports began.

### Theoretical framework and model structure
- Model extension:
  - Builds on Jaimovich and Rebelo (2008) small open economy model extended to two sectors where one sector is a resource (oil) sector.
  - Purpose: Generate macro predictions for news about oil discoveries and compare with aggregate TFP news shocks.
- Endowment economy (intuition):
  - Consumption obeys permanent income intuition; news shocks induce anticipatory consumption and saving responses.
  - Qualitative prediction: upon receiving news, current account should decline first (anticipation effect) and then increase as output materializes.
- Two-sector production model (key primitives preserved):
  - Preferences: utility specification nests KPR (σ = 1) and GHH (σ = 0); parameters include 0 < ߚ < 1, ߠ ≥ 1, ߰ ≥ 1, ߪ ≥ 0; GHH-like preference calibration emphasized.
  - Sector 1 (non-oil) production: y_{1,t} = A_{1,t} h_{1,t}^{α_h} k_{1,t-1}^{α_k}.
  - Sector 2 (oil) production: y_{2,t} = A_{2,t} h_{2,t}^{α_h2} k_{2,t-1}^{α_k2} R_{t-1}^{α_R}.
  - Capital accumulation with adjustment costs: sectoral k_{ζ,t} law with ϕ_ζ > 0 and no adjustment costs in steady state.
  - Labor adjustment costs (Ψ_h) impede intratemporal reallocation.
  - External interest rate with debt-elastic spread to ensure stationarity: r_t = r̄ + ξ [ (S̄ - S_{t-1}) / S̄ ] (text preserves original functional form).
  - Resource factor R_t specified as AR(2) to capture time-to-build and lag between news and production.
- Key analytical insights:
  - News about future oil raises expected future resource income and can induce immediate borrowing and consumption/investment responses that deteriorate the current account initially.
  - Realization of resource output reverses anticipation dynamics: saving and current account improve when oil income is received.
  - Sectoral adjustment costs and high capital share in the resource sector amplify investment responses and current account swings.
  - Two-sector setup implies hours can fall on news arrival (contrasting with one-sector models where hours rise).

### Calibration and baseline parameters (highlights from Table 1)
- Discount factor (ߚ): 0.9
- Steady-state labor parameter (߰): 0.0197
- Exponent on labor in utility (ߠ): 1.2
- Wealth effect parameter (ߛ): 0.01
- Investment adjustment cost parameter (ϕ_ζ): 0.1 in both sectors
- Labor adjustment cost parameter (Ψ_h): 0.5 in both sectors
- Capital depreciation (δ): 0.1 in both sectors
- Labor share in non-resource sector (α1): 0.64
- Labor share in resource sector (α2): 0.13
- Capital share in resource sector (αk): 0.49
- Elasticity of interest rate w.r.t. net foreign assets (߯): 0.00038
- Parameter in interest rate function set so steady-state tb/y = 0.02 (ܤ): -13
- Parameters governing exogenous depletion rate of reserves (ߩ1, ߩ2): 1.4, -0.5
- Relative price of oil (݌): 1.1 (set so steady-state oil share of GDP is 7%)
- TFP in both sectors (Ai): 1

### Empirical strategy and identification
- Identification approach:
  - Exploit exogenous timing and lags of giant oil discoveries as quasi-natural experiments to measure news shocks about future output without relying on VAR or parametric DSGE identification.
- Estimation method:
  - Dynamic panel distributed lag model with country and year fixed effects:
    - y_{i,t} = S(L) y_{i,t−1} + S(L) Φ_i + Σ_{j=0}^{J} β_j NPV_{i,t−j} + X_{i,t−1}' Γ + μ_{i,t}
  - Dependent variables include log real GDP, current account-GDP ratio, saving-GDP, investment-GDP, log real consumption, employment-population ratio.
  - Benchmark autoregressive order for dependent variable j = 1 and lag length J = 10; regressions in levels include country-specific quadratic trends where appropriate.
- Construction of NPV measure:
  - NPV_{i,t}: sum of gross oil revenue from an approximated production profile discounted by country-specific discount factors and valued at the oil price prevailing at the time of discovery, expressed as percentage of GDP.
  - Production profile: zero for first 5 years, then plateau q_p followed by exponential decline when depletion constraint binds; q_p and maximum depletion rate d_m estimated from field-level data.

### Data, coverage, and summary statistics
- Discovery dataset: Mike Horn (1970–2012) with 371 world discoveries; 64 countries with at least one giant discovery in sample period.
- Regional discovery counts (1970–2012 totals):
  - Middle East and North Africa: 97
  - Asia: 74
  - Western Hemisphere: 74
  - Commonwealth of Independent States and Mongolia: 51
  - Europe (including Central and Eastern Europe): 45
  - Africa: 38
  - World total: 371
- NPV distribution:
  - Median NPV: 9 percent of GDP (2.2 in logarithms).
  - Largest NPV estimated at 63 times the country’s GDP (textual statement preserved).
- Macroeconomic variable sample (selected statistics, 1980–2012 unless noted):
  - Ln(GDP): Obs 5537; Min -356.15; Median 579.65; Max 1477.83
  - CA/GDP: Obs 5408; Min -242.19; Median -3.2; Max 106.84
  - Saving/GDP (1980-2011): Obs 4711; Min -202.89; Median 18.19; Max 107.18
  - Investment/GDP (1980-2011): Obs 4956; Min -2.42; Median 21.13; Max 113.58
  - Employment rate (1991-2012): Obs 3630; Min 28.9; Median 57.4; Max 88.1

### Main empirical findings (dynamic responses)
- Current account, saving, investment:
  - Immediate aftermath: current account-GDP ratio decreases significantly; saving-GDP ratio declines for about five years.
  - Investment-GDP ratio rises starting one year after discovery, peaking around 5 years when production typically starts, and then returning to normal quickly.
  - Five years after discovery: average effect on current account turns positive and significant; peak effect around eight years after discovery, then gradual decline.
- GDP, consumption, employment:
  - GDP: slightly negative initially; robust positive effect about 5 years after discovery; peak about 7 to 8 years after discovery; then slowly returns to normal.
  - Consumption (log): little response at first; starts to rise three years after discovery; estimates are imprecise.
  - Employment-population ratio: declines immediately after discovery and remains below normal for several years; estimates small in magnitude but precisely estimated.
- Hypothesis tests on integrated responses (Table 3 p-values):
  - Current Account/GDP:
    - Horizons 0-4 test p-value: 0.00 (significant negative)
    - Horizons 5-11 test p-value: 0.00 (significant positive)
  - Saving/GDP:
    - Horizons 0-4 p-value: 0.05 (negative)
    - Horizons 5-11 p-value: 0.05 (positive)
  - Investment/GDP:
    - Horizons 0-4 p-value: 0.06 (positive, marginal)
    - Horizons 5-11 p-value: 0.58 (not significant)
  - GDP:
    - Horizons 0-4 p-value: 0.83 (not different from zero)
    - Horizons 5-11 p-value: 0.00 (positive)
  - Consumption (0-11): p-value: 0.13 (not statistically significant)
  - Employment-population ratio (0-11): p-value: 0.03 (negative)
- Comparison to theory:
  - Empirical and theoretical impulse responses broadly consistent in shape, magnitude, and timing after normalizing theoretical model so GDP peak matches empirical peak.
  - Notable empirical-theoretical differences: empirical positive current account and savings responses larger after production starts; saving response peaks later empirically; consumption response larger empirically but imprecise.

### Theoretical experiment results (simulation highlights)
- News shock with 5-year delay versus immediate availability:
  - News shock: current account turns sharply negative for five years, then becomes positive; saving negative for several years then very positive; investment rises strongly and drives early current account deterioration; GDP rises little in first five years then shoots up.
  - Immediate shock: current account-GDP ratio turns only slightly negative for a couple of years; saving-GDP ratio rises; output jumps immediately.
- Sectoral share experiments:
  - High resource-sector capital share amplifies investment spike and current account swings.
  - Changing resource-sector labor share from 0.13 to 0.4 (keeping resource share constant) mutes investment response and alters hours dynamics: high labor-share case leads to significant positive hours when production begins.
- Robust theoretical synthesis:
  - Current account response to news is robustly negative between news arrival and production across specifications.
  - Hours response differs from one-sector models: two-sector model often predicts hours fall on news arrival and can remain lower depending on sectoral labor shares.

### Robustness checks and sensitivity analyses
- Alternative NPV constructions:
  - Constant production for 20 years and common discount factor of 10 percent produce virtually identical impulse responses.
- Sample composition:
  - Excluding Middle East and North Africa countries does not change main results.
  - Using only countries with at least one discovery yields robust results.
  - Removing top ten largest exporters leaves main results virtually unchanged.
- Predictability tests:
  - Linear probability and panel logit models: no evidence that lagged current account and investment predict discoveries (Table 4 p-values: Linear row 0.61, 0.45, 0.69, 0.57; Logit row 0.75, 0.53, 0.72, 0.62 across specifications).
  - Removing discoveries that follow others or restricting to discoveries with no prior discovery in past three years yields similar impulse responses.
  - Controlling for current and lagged exploration expenditures does not change main results.
- Econometric specifications:
  - Results robust to higher order lags for dependent variable and different lag orders for independent variables.
  - Robust to removal of country-specific quadratic trends for level variables.
- Government behavior and exchange rates:
  - Private investment-GDP rises while public investment-GDP falls following discoveries.
  - Private consumption increases; public consumption jumps when production begins.
  - Real exchange rate: empirical point estimates show appreciation in first five years but are imprecise and not statistically different from zero.

### Appendix and data approximation details
- Production profile approximation (Appendix B.I):
  - Production zero for first 5 years after discovery; then plateau q_p until depletion constraint binds; thereafter exponential decline at constant depletion rate d_m.
  - q_p specified via empirical power law relation between field size and plateau production; d_m set to capture heterogeneous decline rates across field sizes.
  - Parameter estimates for plateau (pooled "All giants"): α = 0.57, β = 0.65, R2 = 0.65.
  - Parameter estimates for maximum depletion rate (pooled "All giants"): γ = 0.64, δ = -0.31, R2 = 0.31.
- Discounting and risk adjustments:
  - Adjusted discount rate = risk free rate set to 5 percent + country specific risk premium.
  - Sovereign spread prediction uses EMBI Global (41 emerging market economies, 1997-2007) and International Country Risk Guide (2012).
  - Estimated ln(EMBI_it) regression includes ln(political risk_it) with coefficient 13.27 (t-statistic 1.76) and reported R-squared 0.41 in the source presentation.
- NPV valuation choices:
  - Gross revenues valued at current international prices at discovery (price assumed best forecast due to random walk behavior).
  - Ultimately recoverable size based on estimates at time of discovery (not subsequent revisions).

### Conclusions and implications
- Empirical conclusions:
  - News about giant oil discoveries produces a statistically and economically significant anticipation effect: immediately after news, current account and saving decrease while investment increases; after production begins, current account and saving increase along with GDP.
  - Employment falls upon arrival of the news and remains depressed for several years.
- Theoretical contribution:
  - Two-sector open economy model shows sector-specific news can produce initial negative effects on hours and large investment-driven current account swings; high capital share in the resource sector amplifies these dynamics.
- Relevance to literature:
  - Results provide direct measures of news shocks and anticipation effects, complementing the literature on news-driven business cycles and current account adjustments.

*Source: Excerpted content from IMF Working Paper _wp15209 (PDF content unit).*

### References .............................................................................................................

### _wp15209 - References

### Introduction and motivation
- Objective: Provide empirical evidence of the effect of news shocks on the current account, saving, investment, GDP, consumption, and employment using plausibly exogenous variation in the timing of worldwide giant oil discoveries as an observable measure of news shocks about higher future output.
- Key empirical sample: 1970–2012 for up to 130 countries.
- Definition: A giant oil discovery is a discovery of an oil and/or gas field that contains at least a total of 500 million barrels of ultimately recoverable oil equivalent.
- Typical discovery-to-production lag: 4 to 6 years.
- Median value of constructed NPV as a percentage of country’s GDP: about 9 percent.
- Historical illustrative case: Norway — borrowing to build North Sea production facilities after discoveries in the late 1960s and early 1970s; saving rate declined; current account deficit approached minus 15 percent of GDP at its trough in 1977; current account improved after massive oil exports began.

### Theoretical framework and model extensions
- Model: Extension of the Jaimovich and Rebelo (2008) small open economy model to include two sectors, where one sector is a resource sector with oil discoveries.
- Purpose: Develop macroeconomic predictions for news about oil discoveries and determine differences from standard aggregate TFP news shocks.
- NPV construction: Sum of gross oil revenue derived from an approximated oil production profile discounted by country-specific discounting factors, and valued at the oil price prevailing at the time of the discovery.

### Empirical strategy
- Identification approach: Exploit natural lags between giant oil discoveries and subsequent production to directly measure news shocks about future output; quasi-natural experiment that does not rely on VAR or parametric DSGE identification assumptions.
- Estimation method: Dynamic panel distributed lag model with year- and country-fixed effects to control for global common shocks and time-invariant cross-country differences.
- Within-country variation: Exploited to mitigate endogeneity concerns from omitted variable bias.

### Main empirical findings
- Short-run (immediately following discoveries):
  - Current account decreases significantly as investment rises and the saving rate declines.
  - Employment rates decline after the news arrives and remain below normal for over 10 years.
- Medium- to longer-run:
  - Five years after the discovery, the average effect on the current account turns positive and significant as output and saving rise and investment declines.
  - Peak effect is reached about eight years following the discovery, after which the effect gradually declines.
- Sectoral composition:
  - Private investment-GDP ratio rises robustly.
  - Public investment-GDP ratio falls.
  - Total consumption response driven by both private and public consumption increases, though estimates are imprecise.

### Robustness checks and alternative specifications
- Alternative NPV measures: Used NPV with constant production profile and NPV with common discount factors across countries; impulse responses virtually the same.
- Sample composition: Results not driven by particular groups of countries; removing Middle East and North Africa countries, major oil exporters, or countries without any discoveries does not alter dynamic effect patterns.
- Predictability concerns: Removing discoveries that immediately follow a discovery, selectively using discoveries when no discoveries occurred in the past three years, and separately controlling for current and lagged exploration expenditures; results virtually unchanged.
- Model specifications: Robust to including higher order lags for the dependent variable and for giant oil discoveries.

### Implications for literature and theory
- News-driven business cycles:
  - Provides direct measures of news about future output and tests model predictions.
  - Findings are not consistent with news as a driver of business cycle fluctuations—contrasting with studies that attribute large fractions of business cycle variation to news shocks (e.g., Beaudry and Portier (2006)).
- Current account theory:
  - Provides direct evidence supporting the classic intertemporal trade model of the current account: anticipation effects on the current account operate through both saving and investment channels following discovery announcements.
  - Contributes to literature on determinants of the current account and adjustment to shocks.

### Methodological contribution
- Identification: Exploiting the exogenous timing and production lag of giant oil discoveries offers a novel, direct approach to identifying news shocks and associated anticipation effects, avoiding many untested identification assumptions required by VAR and DSGE approaches.

*Source: Excerpt from the provided PDF content unit.*

### Section VII discusses robustness checks. Section VIII concludes.

### II. THEORETICAL EFFECTS OF AN OIL DISCOVERY

### A. News Shocks in an Endowment Economy
- Setup and assumptions:
  - Small open economy with infinitely lived households, one tradeable good, exogenous endowment in each period.
  - Households discount future utility by factor ߚ and borrow/lend at exogenous world interest rate r.
  - Representative household maximization: max Σ_{t=0}^∞ ߚ^t U(c_t) subject to budget and No-Ponzi scheme.
  - Utility assumed linear-quadratic (Hall (1978)); discount factor specification and Euler condition imply permanent income result: c_t = E_t(c_{t+1}).
  - Output y_t defined relative to long-run steady state ȳ: y_t ≡ y_t / ȳ and follows AR(1) with coefficient ߩ and shock η_t.
  - Shock decomposition (following Schmitt-Grohe and Uribe (2012)):
    - η_t = η_t^u + η_t^n, where η_t^u is an unanticipated contemporaneous shock and η_{t+1}^n is a one-period-ahead news shock learned in period t.
    - For k = 0,1 both η_{t+k}^· have mean zero and are uncorrelated across time and anticipation horizons.

- Consumption response:
  - Optimal change in consumption: Δc_t = [ρ/(1-ρ^2)] η_t^u + [ρ/(1-ρ^2)] η_t^{n} ??? (text presents: ܥ߂_௧ = ݎ ߩݎ1൅ ߟቆ...)  
  - Key intuition: both contemporaneous shock and news shock in t change expectations and induce updating of consumption path. However, a news shock learned in t-1 and materializing in t does not produce additional information updating at t (consumers adjusted consumption when news arrived), so the realized news shock η_{t}^{n} drops out of the consumption innovation at t.

- Current account / saving response:
  - Current account s_t ≡ S_t = S_{t-1} and in this endowment economy equals saving (no investment).
  - Change in current account given by (text expression):
    - ܣܥ߂_௧ = 1/(ߩݎ1൅) [ ቂെ(ߩ1െ)ଶ ݕ_{௧ିଵ} + (ߩ1െ) ߟ_{௧}଴ েݎ... + (remaining terms) ]
  - Qualitative results:
    - Contemporaneous positive temporary shock (η_t^u positive) → increase in current account (increase in saving).
    - Anticipated positive news shock (η_t^{n} learned earlier) → decrease in current account (anticipation effect reduces saving initially).
    - A prior-period news shock realized in the current period (η_{t-1}^{n} materializes now) → increase in current account.
    - Overall dynamics: upon receiving news, current account should decline first (anticipation effect) and then increase as the output materializes.

- Limitations of the simple endowment model:
  - Does not consider endogenous investment and labor responses.
  - Omits oil-sector specific features.
  - Motivates extension to a two-sector production model with oil discovery news shocks.

### B. A Two-Sector Small Open Economy Model with a News Shock
- Purpose:
  - Extend Jaimovich and Rebelo (2008) one-sector model by adding a resource (oil) sector to capture features of news about oil discoveries.

- Preferences and household problem:
  - Lifetime utility:
    - U = Σ_{t=0}^∞ ߚ^t [ c_t^{1-σ}/(1-σ) - χ (h_t^{1+ψ}/(1+ψ)) ] * X_t^{1-ν} ??? (text gives functional form)
    - The specific form used (Jaimovich and Rebelo (2008, 2009)) nests KPR (σ = 1) and GHH (σ = 0) preferences.
  - Parameters: 0 < ߚ < 1, ߠ ≥ 1, ߰ ≥ 1, ߪ ≥ 0.
  - X_t introduces non-time-separability between consumption and hours; agents internalize X_t dynamics.

- Production structure:
  - Two sectors: non-oil (sector 1) and oil (sector 2).
  - Sector 1 (non-oil) production:
    - y_{1,t} = A_{1,t} h_{1,t}^{α_h} k_{1,t-1}^{α_k}  (Cobb-Douglas with constant returns)
    - Text notation: ܻ_{ଵ,௧} = ܣ_{ଵ,௧} ܰ_{ଵ,௧}^{ईభ} ܭ_{ଵ,௧ିଵ}^{ଵିఈభ}
  - Sector 2 (oil) production:
    - y_{2,t} = A_{2,t} h_{2,t}^{α_h2} k_{2,t-1}^{α_k2} R_{t-1}^{α_R}  (Cobb-Douglas including resource factor R)
    - Text notation: ܻ_{ଶ,௧} = ܣ_{ଶ,௧} ܰ_{ଶ,௧}^{ईమ} ܭ_{ଶ,௧ିଵ}^{ఈౖ} ܴ_{௧ିଵ}^{ଵిఈమిఈౖ}
  - Parameter restrictions: 0 < α_· < 1 and α_h + α_k + α_R ≤ 1 as given in the text.

- Capital accumulation and adjustment costs:
  - Sectoral capital accumulation for sector ζ (ζ = 1,2):
    - k_{ζ,t} = i_{ζ,t} + (1 - ϕ_ζ/2 ((i_{ζ,t}/i_{ζ,t-1} - 1)^2) ) k_{ζ,t-1}
    - Text expression (equation (5)): ܭ_{௛,௧ ܫ} = ௛,௧ ൥ 1െ ߶_௛ 2 ቆ ܫ_௛,௧ ܫ_௛,௧ିଵ 1ቇ ... (functional form with ϕ_ζ)
  - ϕ_ζ > 0; functional form implies no adjustment costs in steady state.
  - Sector-specific labor adjustment costs are introduced (subtracted from flow budget constraint) to impede intratemporal reallocation of labor and capital.

- Market structure and trade:
  - All goods assumed tradeable; households consume only good 1 but can exchange oil for good 1 on international markets.
  - Flow budget constraint (equation (6)):
    - S_t = (1 + r_t) S_{t-1} + [p_t y_{2,t} + y_{1,t}] - c_t - Σ_{ζ=1,2} [Ψ_ζ/2 (h_{ζ,t} - h_{ζ,t-1})^2] - Σ_ζ i_{ζ,t}
    - Text notation preserves exact terms: ܤ_௧ = (ݎ1൅௧)ܤ_{௧ିଵ} ܻା൅(ܵ,௧ ݌൅ ܻ_{ଶ,௧}) ܥ ... (labor adj. cost terms)
  - Relative price of oil ݌_t determined by world market; oil traded internationally.

- Interest rate and stationarity of foreign assets:
  - To induce stationarity of foreign bond holdings, adopt external debt-elastic interest rate (Schmitt-Grohe and Uribe (2003)):
    - r_t = r̄ + ξ [ (S̄ - S_{t-1}) / S̄ ]  ??? (text expression)
    - Text exact equation (7): ݎ_௧ ݎ = ௪ ൅߯ ሾ ݌ݔ݁ ሺ ܤ ത ܤେ ௧ିଵ ሻ ে1 ିሿ
  - Where r̄ is the world interest rate and ߯ ≥ 0 is the interest rate debt elasticity.

- Modeling news about oil discoveries:
  - The model incorporates news shocks about future oil discoveries / output in the resource sector, allowing analysis of anticipation effects on consumption, investment, labor allocation, and the current account.
  - Sectoral specificity of capital and adjustment costs implies slower reallocation across sectors following news, affecting dynamic responses.

- Key analytical insights from the two-sector setup (as contrasted with the endowment model):
  - News about future oil increases expected future income from the resource sector; anticipation effects can induce:
    - Upfront increases in consumption financed by foreign borrowing (current account deterioration) upon news arrival.
    - Reallocation of labor and investment responses contingent on sectoral adjustment costs and preference specification (KPR vs GHH features).
  - Realized oil output (when discovery materializes) reverses some of the anticipation dynamics: saving increases and current account improves as resource income is realized.
  - The presence of sector-specific adjustment costs, labor adjustment costs, and non-time-separable preferences alters the timing and magnitude of consumption, investment, labor, and current account responses compared with the simple endowment model.

*IMF Working Paper _wp15209 — Section II excerpt (theoretical framework on news shocks and two-sector model).*

### part is the risk premium which is decreasing in the country’s aggregate net foreign assets. We

### _wp15209 - part is the risk premium which is decreasing in the country’s aggregate net foreign assets. We

### Definitions and model primitives
- Trade balance (TB) is defined as in equation (8):  
  TB_t = (p_1,t q_1,t + p_2,t q_2,t) − p_1,t c_1,t − p_2,t c_2,t + Σ_k=1^N A_k,t Ψ_k^2 (A_k,t A_k,t−1)^−1  (expression preserved as in source).
- Current account (CA) is defined as in equation (9):  
  CA_t = B_t − B_{t−1}, and thus CA_t = r_t B_t − B_{t−1} (as presented: ܣܥ௧ܤൌ௧ܤെ௧்).
- Saving is defined as in equation (10):  
  S_t = C_t + G_{1,t} + G_{2,t} (preserved notation: ܵܣ௧ܣܥൌ௧ܫ൅ଵ௧ܫ൅ଶ௧).
- Aggregate variables are defined in equation (11):  
  Output: Y_t = y_{1,t} + y_{2,t}; Capital: K_t = k_{1,t} + k_{2,t}; Labor: L_t = l_{1,t} + l_{2,t}; Resources: R_t = r_{1,t} + r_{2,t} (preserving original notation ܻ௧ܻൌଵ,௧݌൅௧ܻଶ,௧, etc.).

### Resource news specification and time-to-build feature
- Resource factor AR(2) specification (equation (12)):  
  R_t = (ρ1 + ρ2) ρ R_{t−1} + ρ1 R_{t−1} + ρ2 R_{t−2} + ε_t (preserving original AR(2) structure and parameter constraints: ߩ1 ߩ2 ൏ 1 and ൒0݆).
- Interpretation and modelling choices:
  - R_t interpreted as a “resource factor” rather than literal reserves; lag on ε_{t−j} captures that resources are not immediately available when news arrives.
  - Sector-specific adjustment costs on investment replicate time-to-build effects.
  - Resource depletion is modelled exogenously via ρ_s rather than optimal extraction dynamics.

### Calibration and baseline parameter choices
- Calibration matches annual data used in empirical analysis; baseline summarized in Table 1 (parameters similar to Jaimovich and Rebelo (2008), converted to annual basis).
- Preference and interest parameters:
  - γ set so preferences are very close to GHH preferences.
  - χ set so elasticity of interest rates is very low.
  - World interest rate set to 10 percent to match the average of the risk adjusted interest rate in the data.
- Resource-sector shares (following Gross and Hansen (2013)):
  - Labor share = 13 percent
  - Capital share = 49 percent
  - Resource share = 38 percent
  - These shares noted as broadly consistent with U.S. data.

### Simulation experiments and key theoretical results
- News shock timing:
  - Typical lag between discovery and initial oil production is five years; authors explore news shock ε_{t+5} in equation (12).
  - Compare two cases: discovery with five-year delay (news) versus discovery with immediate availability (contemporaneous shock ε_t).
- Main dynamic patterns (from Figure II and related discussion):
  - Current account:
    - News about future resources: current account turns sharply negative for five years, then becomes positive.
    - Immediate availability: current account-GDP ratio turns only slightly negative for a couple of years; saving-GDP ratio always positive.
  - Saving-GDP ratio:
    - News shock: negative for several years, then very positive.
    - Immediate shock: saving-GDP ratio rises because households smooth consumption for what is not permanent.
  - Investment-GDP ratio:
    - Immediate shock: rising investment contributes to slight current account decline.
    - News shock: rising investment is a key channel (plus declining savings) causing countercyclical current account.
  - GDP:
    - News shock: GDP rises very little during first five years, then shoots up rapidly before gradually declining.
    - Contemporaneous shock: output jumps immediately and stays high for several years before falling.
  - Consumption:
    - Jumps immediately in both experiments; jump larger for contemporaneous shock (greater present discounted value of resources).
  - Hours (labor input):
    - Contemporaneous shock: rise in hours during first five years, then oscillations around zero.
    - News shock: hours fall in both short-run and medium-run (response of hours small—only one-tenth of aggregate GDP response, per note).
    - Authors emphasize this contrasts with Jaimovich and Rebelo (2008) where news shocks raised hours.
- Experiments isolating sectoral features:
  - Experiment 1: Compare oil news shock to aggregate TFP news shock (both with five-year lead). Results:
    - GDP paths similar (peaks normalized).
    - Current account-GDP, saving-GDP, and investment-GDP show qualitatively similar responses but swings much larger for oil shock due to required capital reallocation and higher capital share in oil sector.
    - For TFP shock: consumption and hours responses are hump-shaped and positive.
  - Experiment 2: Vary Sector 2 (oil) factor shares:
    - Baseline Sector 2: labor share = 0.13, capital share = 0.49, resource share = 0.38.
    - Alternative Sector 2: labor share = 0.4, capital share = 0.22, resource share = 0.38 (ratio equal to Sector 1).
    - Findings:
      - GDP, current account-GDP, saving-GDP similar across parameterizations.
      - Investment-GDP response muted when Sector 2 capital share is lower — much of investment spike in baseline due to high capital share.
      - Hours responses: similar between arrival of news and start of production, but diverge after production starts:
        - Baseline (low labor share): hours continue to decline after production begins (reallocation reduces labor demand).
        - High labor-share case: hours become significantly positive in the fifth year when production begins.
- Synthesis of theoretical findings:
  - Current account response to news is robust: becomes significantly negative for the five years between news arrival and increase in resources across sector-specific and aggregate TFP shocks.
  - Behavior of variables like labor input and consumption depends on shock details:
    - Positive response of hours to news is a feature of one-sector models but not the two-sector model.
    - In two-sector models, hours decline when news arrives but before materialization regardless of oil sector labor share; after production starts, hours fall more if oil sector is low labor share, and rise if oil sector is high labor share.

### Rationale for using giant oil discoveries as empirical news shocks
- Three unique features that make giant oil discoveries suitable proxies for macro-relevant news about future output:
  1. They signal significant increases in future production possibilities (macro-relevance).
  2. There is a significant delay between discovery and start of production (typical lag discussed as five years):
     - UK anecdotal average delay 4.5 years (1954-2011).
     - Experts: giant oil discovery takes between 4 and 6 years to go from drilling to production.
     - Authors’ alternative dataset calculation: average delay 5.4 years.
     - Heterogeneity: average delay 6.4 years for offshore discoveries and 4.6 years for onshore discoveries (from alternative dataset).
  3. Timing is arguably exogenous and unexpected after controlling for country and year fixed effects and previous discoveries:
     - Exploration uncertainty driven by technological innovation and geological uncertainty.
     - Dataset by Mike Horn provides independently verified and documented discovery dates.
- Data coverage and scope:
  - Sample period: 1970 to 2012.
  - Up to 130 countries.
  - Horn (2011) dataset includes country, year, oil/gas indicator, and estimated total ultimately recoverable amount in oil equivalent.
  - Summary statistics:
    - Total discovery events in period 1970-2012: 371 world discoveries.
    - 64 countries had at least one giant oil discovery in the sample period.
    - Regional counts: Middle East and North Africa = 97; Asia = 74; Western Hemisphere = 74; Commonwealth of Independent States and Mongolia = 51.
    - The 1970s are the peak period for giant oil discoveries; number of discoveries has been growing since the 1980s.
    - Distribution of discovery sizes (log million barrels ultimately recoverable) shows significant heterogeneity.
  - Note on data sources: alternative dataset (Global Energy Systems, Uppsala University) includes 358 discoveries covering 47 countries; number shrinks to 157 for period 1970 onwards.

### Empirical strategy and estimation details
- Empirical model: dynamic panel with distributed lags of giant oil discoveries (equation (13)):
  - y_{i,t} = S(L) y_{i,t−1} + S(L) Φ_i + Σ_{j=0}^{J} β_j NPV_{i,t−j} + X_{i,t−1}' Γ + μ_{i,t}
    - y_{i,t}: dependent macro variables — log real GDP (local currency), current account-GDP ratio, saving-GDP ratio, investment-GDP ratio, log real consumption (local currency), employment-population ratio.
    - α_i: country fixed effects (capture time-invariant characteristics).
    - δ_t: year effects controlling for common shocks (global cycles, international crude oil and gas prices).
    - X_{i,t−1}: other control variables used in robustness (e.g., exploration expenditures).
    - NPV_{i,t}: net present value of giant oil discoveries measure (described below).
  - Lag operators S(·) allow autoregressive and distributed lag structure; benchmark uses autoregressive order j = 1 and lag length J = 10.
  - Regressions using log levels and employment rate include country-specific quadratic trends.
- Identification advantages of panel and lag structure:
  - Country fixed effects exploit within-country variation in discovery timing and control for unobserved time-invariant confounders.
  - Distributed lags control for serial correlation in discoveries and allow impulse response interpretation.
  - Large panel cross-section and long time span (at least thirty years) mitigate dynamic panel biases (Nickell bias of order 1/T argued to be negligible in this sample).
  - Authors also tried excluding country fixed effects to verify robustness; main results qualitatively and quantitatively similar, but benchmark retains fixed effects because jointly significant.
- Construction of the NPV discovery measure (percent of GDP):
  - NPV_{i,t} = Σ_{τ} [Revenue_{i,τ} * discount factor] * 100 / GDP_i,t (preserved as formula structure in source).
  - Gross revenue derived from an approximated oil production profile q_{i,τ} from the fifth year following discovery to exhaustion year T, valued at the oil price prevailing at time of discovery.
  - Production profile: piece-wise with reserve-specific plateau production followed by exponential decline (per Höök et al. 2014 and Robelius 2007).
  - Motivation: agents respond to net present value of revealed output shock; timing variation alone would ignore heterogeneity in discovery sizes.

*Source: _wp15209 - part is the risk premium which is decreasing in the country’s aggregate net foreign assets. We (PDF chapter/section).*

### Appendix B.I. describes in detail the approximation

### _wp15209 - Appendix B.I. describes in detail the approximation

### Approximation method and valuation
- Gross revenues are valued at current international prices because "oil price series typically follow a random walk process so that current price is the best price forecast."
- The ultimately recoverable size for each discovery is based on the estimation of the value at the time of the discovery, rather than potentially revised estimates in subsequent years.
- The authors do not use the Hubbert curve for single-field production profiles; instead they rely on a reserve specific piecewise process consisting in a plateau production and then an exponential decline.
- Appendix B.I. relies on estimates using an alternative oil field database to approximate recoverable sizes.

### Risk-adjusted discounting and sovereign spread prediction
- Adjusted discount rate = risk free rate set to 5 percent + country specific risk premium.
- The risk free rate is assumed to be the rate prevailing in the United States.
- Sovereign bond spread data source: Emerging Markets Bond Index Global (EMBI Global) available for 41 emerging market economies for the period 1997-2007.
- Political risk rating source: International Country Risk Guide (2012) covering 133 countries.
- Econometric specification to relate sovereign spreads to political risk ratings (notation preserved):
  - ln(EMBIit) = α0 + α1 ln(IRit) + βi + γt + εit (country fixed effects βi, year effects γt).
- Estimated prediction used to obtain country-specific predicted spreads; adjusted equation forced to predict a spread of zero for countries with political risk equal to or lower than the United States.
- Estimated coefficient summary (as presented):
  - Ln(EMBI_it) regression items presented with coefficients: 13.27 (t-statistic 1.76) ln(political risk_it), (3.71), (-2.03) — t-statistics in parenthesis indicate political risk is a significant determinant; R-squared is 0.41.

### Distribution of NPV of giant oil discoveries
- Histogram findings:
  - Median NPV is 9 percent of GDP (2.2 in logarithms).
  - The largest NPV is estimated to be 63 times of the country’s GDP.
- Results are robust to alternative NPV measures (common discount rates and uniform production profile).

### Data coverage and sample period
- Macroeconomic variables sourced from IMF (2013), the World Bank (2013), and the International Labor Organization.
- Benchmark NPV measure starts in 1970; with 10 lags of discovery sizes, baseline regression covers macro variables from 1980-2012.
- Supplementary tables referenced: data appendix Table B.II (data definitions), Table B.III (summary statistics).

### Benchmark empirical results (dynamic panel autoregressive distributed lag model)
- Estimation details:
  - Panel autoregressive distributed lag model with country and year fixed effects.
  - Confidence bands: 90% and 68% (darker grey) using Driscoll-Kraay (1998) standard errors and the delta method.
- Current account, saving, investment (Figure VII):
  - Immediate aftermath of announcement: negative effect on current account-GDP ratio.
  - Five years after discovery: average effect of giant oil discoveries turns positive.
  - Peak effect reached eight years after discovery, then effect declines.
  - Saving-GDP ratio: negative for about five years following announcement, then robustly positive.
  - Investment-GDP ratio: starts to rise one year after discovery, peaks around 5 years once oil production starts, then returns to normal quickly.
- GDP, consumption, employment (Figure VIII):
  - GDP: slightly negative initially, robust positive effect about 5 years after discovery, peak about 7 to 8 years after discovery, then slowly returns to normal.
  - Log consumption: little response at first; starts to rise three years after discovery; estimates are very imprecise (consumption variable includes private and public consumption).
  - Employment rate: begins to fall immediately after discovery and continues to fall even after start of production; remains depressed for several years; estimates small in magnitude but precisely estimated.
- Hypothesis testing approach:
  - Test integrals of responses over horizons:
    - Example: current account-GDP ratio tested for negative response during first five years (horizons 0 to 4) and positive response for horizons 5 to 11.
  - Findings from Table 3:
    - Current account-GDP ratio: significantly negative between discovery and production (anticipation effect), significantly positive after production starts.
    - Saving-GDP ratio: similar pattern to current account.
    - Investment-GDP ratio: significantly positive (p-value 0.06) for the first five years, not significant for following years.
    - GDP response: not different from zero during first few years, significantly positive after production starts.
    - Consumption integral response: p-value 0.13 (not statistically significant at conventional levels).
    - Employment response: statistically significantly negative.
- Comparison with theoretical model (Figure IX):
  - Empirical and theoretical responses broadly consistent in shape, magnitude, and timing after normalizing the theoretical model so GDP peak matches empirical peak.
  - Notable differences:
    - Empirical positive current account and savings responses are larger than theoretical ones after production starts.
    - Saving response peaks later empirically than theoretically.
    - Consumption response larger empirically but imprecise.
    - Theoretical negative employment response somewhat more pronounced than empirical.
  - Magnitude discrepancy:
    - Empirical shock corresponds to estimated potential stream of oil revenues equal to one percent of initial GDP (NPV).
    - If the theoretical model used a one percent of GDP NPV shock, model GDP peak would be 0.19 instead of 0.03.
    - Possible explanations: agents perceived a lower NPV at announcement, or discounting/risk premium used in NPV calculation does not capture all extraction uncertainty and social tensions.

### Extensions: government behavior and real exchange rates
- Government vs private responses (Figure X):
  - Private investment-GDP ratio increases; public investment-GDP ratio decreases.
  - Private consumption (log levels) increases somewhat; public consumption does not respond initially but jumps once oil production begins.
  - Government spending rises but less than the increase in GDP.
  - Conclusion: government behavior does not appear to significantly change outcomes relative to the simple model; government is nonetheless an important part of total consumption response.
- Real exchange rate:
  - Standard expectation: announcement of giant oil discovery -> immediate appreciation of real exchange rate via relative price of non-tradeables.
  - Empirical finding: real exchange rate appreciated during first five years following discoveries using CPI-based or GDP deflator-based measures, but point estimates very imprecise; no response significantly different from zero.
  - Possible reasons for imprecise results: annual frequency of discovery data, fast-moving nature of real exchange rates, timing issues highlighted by Ramey (2011).

### Robustness checks
- Alternative NPV construction:
  - Constant production rate for 20 years after production starts tested.
  - Common discount factor of 10 percent tested.
  - Impulse responses virtually the same; results not sensitive to NPV construction.
- Sample composition:
  - Removing Middle East and North African countries (region with concentration of discoveries) does not change main results.
  - Using only countries that experienced at least one discovery yields robust results.
- Predictability of discoveries:
  - Linear probability and logit models used to test whether lagged values of current account, saving, and investment predict discoveries (three lags).
  - Table 4: no evidence that lagged current account and investment have predictive power for discoveries.
  - Removing discoveries that follow others (e.g., in the year immediately following a prior discovery) leaves impulse responses virtually unchanged (employment response becomes less informative).
  - Retaining as news shocks only discoveries without prior discoveries in past three years yields qualitatively similar impulse responses.
  - Controlling for current and lagged exploration expenditures from Global Energy Systems does not change main results.
- Borrowing constraints and African countries:
  - Anticipation effect on current account present for African countries only and similar to benchmark results.
- Substitution effect concern:
  - Tested robustness by removing the top ten largest oil or gas exporters; main results virtually unchanged.
- Alternative econometric specifications:
  - Robust to higher order lags for the dependent variable and different lag orders for independent variables (e.g., dependent variable lag order = 2, independent variable order = 8).
  - Also robust to removing controls for country-specific quadratic trends for output, employment rate, and other level variables.

### Conclusion and implications
- Main empirical conclusions:
  - News about giant oil discoveries causes a statistically and economically significant anticipation effect: immediately after news, current account and saving decrease while investment increases.
  - After the start of oil production, current account and saving increase along with GDP.
  - Employment falls upon arrival of the news and continues to decline for several years.
- Theoretical insight:
  - Two-sector open economy model shows sector-specific news can cause initial negative effects on hours; this is not driven solely by low labor share in oil sector but is a general feature of sector-specific TFP news.
- Relation to literature:
  - Findings align with Ramey (2011), Barro and Redlick (2011), and Mertens and Ravn (2012) on news effects and oscillations of macro aggregates.
  - Results support the role of news about future output as an important source of fluctuations in the current account, consistent with Engel and Rogers (2006) and Corsetti and Konstantinou (2012).

*Source: _wp15209 - Appendix B.I. describes in detail the approximation*

### References

### _wp15209 - References

### Literature themes cited
- News shocks and business cycles: multiple studies on news-driven fluctuations and information effects, including Alexopoulos (2011); Barsky and Sims (2011); Barsky and Sims (2012); Barsky, Basu, and Lee (2014); Beaudry and Portier (2004, 2006, 2007, 2014); Blanchard, L’Huillier, and Lorenzoni (2013); Cochrane (1994); Jaimovich and Rebelo (2008, 2009); Krusell and McKay (2010); Schmitt-Grohé and Uribe (2012); Kurmann and Otrok (2013); Kueng (2014); Mertens and Ravn (2012); Ramey (2011).
- Resource discoveries, oil shocks, and macroeconomic adjustment: Eastwood and Venables (1982); Ghosh and Ostry (1995); Glick and Rogoff (1995); Gross and Hansen (2013); Hamilton (2009); Hamilton and Atkinson (2013); Horn (2011); Lei and Michaels (2011); Lucca (2007); Pieschacón (2012); Pigou (1927); Pindyck (1978); Robelius (2007); Samuel Wills (2013); Tsui (2011).
- Open economy dynamics and current account behavior: Bergin and Sheffrin (2000); Chinn and Prasad (2003); Chinn and Wei (2013); Correia, Neves, and Rebelo (1995); Engel and Rogers (2006); Obstfeld and Rogoff (1995); Schmitt-Grohé and Uribe (2003); Uribe and Schmitt-Grohé (2014).
- Empirical methods and data sources: Driscoll and Kraay (1998); International Country Risk Guide (2012); International Monetary Fund (2013a, 2013b); World Bank Group (2013); Norwegian Petroleum Directorate (data for Figures).

### Baseline calibrated parameters (Table 1)
- Discount factor (ߚ): 0.9
- Governs disutility of labor, set so steady-state labor is 20 (߰): 0.0197
- Exponent on labor in the utility function, governing intertemporal substitution (ߠ): 1.2
- Governs intertemporal substitution of the consumption-hours bundle (ߪ): 1
- Governs the wealth effect; GHH preferences = 0 (ߛ): 0.01
- Investment adjustment cost parameter (߶௛): 0.1 in both sectors
- Labor adjustment cost parameter (Ψh): 0.5 in both sectors
- Capital depreciation (δh): 0.1 in both sectors
- Labor share in non-resource sector (α1): 0.64
- Labor share in resource sector (α2): 0.13
- Capital share in resource sector (αk): 0.49
- Elasticity of interest rate with respect to net foreign assets (߯): 0.00038
- Parameter in interest rate function; set so that the steady-state tb/y = 0.02 (ܤ): -13
- Parameters governing exogenous depletion rate of reserves (ߩଵ, ߩ,ଶ): 1.4, -0.5
- Relative price of oil, set so that steady-state oil share of GDP is 7% (݌): 1.1
- TFP in Sector i, i = 1,2 (Ai): 1

### Spatial and temporal distribution of giant oil discoveries (Table 2, 1970-2012)
- Africa: 1970s 5, 1980s 6, 1990s 9, 2000s 9, 2010s 9, Total 38
- Asia: 1970s 17, 1980s 14, 1990s 20, 2000s 23, 2010s 0, Total 74
- Commonwealth of Independent States and Mongolia: 1970s 22, 1980s 12, 1990s 4, 2000s 10, 2010s 3, Total 51
- Europe (include Central and Eastern Europe): 1970s 17, 1980s 5, 1990s 7, 2000s 3, 2010s 5, Total 45
- Middle East and North Africa: 1970s 36, 1980s 15, 1990s 23, 2000s 18, 2010s 5, Total 97
- Western Hemisphere: 1970s 20, 1980s 15, 1990s 16, 2000s 21, 2010s 2, Total 74
- World total by decade: 1970s 117, 1980s 67, 1990s 79, 2000s 84, 2010s 24, World total 371

Note: A discovery event is a dummy variable equal to 1 if during a given year at least one discovery of either a giant oil or gas field was made in any given country. Data are from Mike Horn; country grouping from the International Monetary Fund.

### Hypothesis tests on responses to an oil news shock (Table 3)
- Current Account/GDP
  - Theoretical prediction: Negative response at horizons 0-4; Positive response starting at horizon 5.
  - Test 1 (horizons 0-4): H0: ∑ bh = 0 vs H1: ∑ bh < 0 — p-value: 0.00
  - Test 2 (horizons 5-11): H0: ∑ bh = 0 vs H1: ∑ bh > 0 — p-value: 0.00
- Saving/GDP
  - Theoretical prediction: Negative response at horizons 0-4; Positive response starting at horizon 5.
  - Test 1 (horizons 0-4): p-value: 0.05
  - Test 2 (horizons 5-11): p-value: 0.05
- Investment/GDP
  - Theoretical prediction: Positive response at horizons 0-4; Negative or zero response starting at horizon 5.
  - Test 1 (horizons 0-4): p-value: 0.06
  - Test 2 (horizons 5-11): p-value: 0.58
- GDP
  - Theoretical prediction: Positive or zero response at horizons 0-4; Positive response starting at horizon 5.
  - Test 1 (horizons 0-4): p-value: 0.83
  - Test 2 (horizons 5-11): p-value: 0.00
- Consumption
  - Theoretical prediction: Positive response at horizons 0-11.
  - Test (horizons 0-11): p-value: 0.13
- Employment-Population ratio
  - Theoretical prediction: Negative response at horizons 0-11.
  - Test (horizons 0-11): p-value: 0.03

Notes: bh denotes the estimated impulse response at horizon h. P-values were obtained from Delta method. Hypotheses constructed based on the theory in Section II.

### Tests for predictability of giant oil discoveries (Table 4)
- Reported p-values for non-predictability tests (panel linear probability model with fixed effects and panel logit model):
  - Linear (p-values across specifications): 0.61 0.45 0.69 0.57
  - Logit (p-values across specifications): 0.75 0.53 0.72 0.62
- Model specifications:
  - Previous discoveries in 10 years: No, No, Yes, Yes
  - Country and year fixed effects: Yes, Yes, Yes, Yes
- Note: Linear row contains p-value of F test that three lags of current account and investment have no predictive power for the oil discovery event. Logit row reports p-value for likelihood ratio test.

### Empirical impulse-response findings (figures; selected notes and exact experiment sizes)
- Typical oilfield investment and production patterns (Figure 1): Investment data nominal divided by GDP deflator; production measured in barrels per day; data from The Norwegian Petroleum Directorate.
- Comparison of oil news versus unanticipated oil shock (Figure 2): Baseline model, 5 year lag of news; vertical axis shows percentage changes; shocks constructed so that the peak of GDP is the same across experiments.
- Effect of oil news vs. aggregate TFP news in one-sector model (Figure 3): Baseline model with near GHH preferences, 5 year lag of news; vertical axis shows percentage changes; shocks normalized so peak of GDP is the same.
- News in 2-sector models: Impact of low labor share (Figure 4): Baseline where sector 2 has low labor share versus high labor share case; 5 year lag of news; shocks normalized to same GDP peak.
- Size distribution of giant oil discoveries (Figure 5): Presents logarithm of million barrels of ultimately recoverable oil equivalent for giant discoveries in sample (1970-2012).
- Distribution of Net Present Value of giant oil discoveries (Figure 6): Presents NPV of discoveries in percentage of GDP (log).
- Impact of giant oil discoveries on current account, saving, and investment (Figure 7): Impulse response of an oil discovery with NPV equal to 1% of GDP; line with circles indicates point estimates; grey areas are 90 percent and 68 confidence intervals; vertical axis shows percentage changes.
- Impact of giant oil discoveries on GDP, consumption, and employment (Figure 8): Impulse response for discovery with NPV equal to 1% of GDP; line with circles = point estimates; grey areas are 90 percent and 68 confidence intervals; vertical axis shows percentage changes.
- Comparison of estimates to theoretical predictions (Figure 9): Green dashed lines = theoretical impulse response; solid blue lines = point estimates; grey areas are 90 percent and 68 confidence intervals; responses normalized so peak of GDP is the same.
- Private versus public responses to discovery (Figure 10): Impulse response of discovery on logarithm of real private and government consumption (local currency units) with control for country-specific quadratic trend; includes response of government spending as percentage of GDP; line with circles = point estimates; grey areas are 90 percent and 68 confidence intervals.

*Source: _wp15209 - References (PDF chapter/section).*

### Appendix A. Theoretical Appendix

### _wp15209 - Appendix A. Theoretical Appendix

### First-order conditions (two-sector model)
- Lagrange multipliers:
  - λ: multiplier on the net foreign asset accumulation equation (6).
  - μ: multiplier on the utility function state variable definition in equation (2).
  - η’s: multipliers on the capital accumulation equations in equation (5).
- The appendix provides the economy’s first-order conditions expressed in the model’s notation (equations (A.1)–(A.8) as given in the source).
  - Equations listed include (A.1), (A.2), (A.3), (A.4), (A.5), (A.6), (A.7), (A.8) specifying optimality conditions for choices and multipliers across sectors (full algebraic forms provided in the source PDF).

### Theoretical relationships highlighted
- Conditions equate marginal valuations and shadow prices across consumption, capital, and foreign asset accumulation, using the model’s state variables and multipliers (λ, μ, η).
- Specific functional and parameter structures are written in the original notation (no simplification), forming the basis for the quantitative exercises in the paper.

---

### Appendix B. Data Appendix

### Approximation of production profile of giant oil discovery
- Assumptions and construction:
  - Typical delay between discovery announcement and production start: 4-6 years; the net present value computation is based on a 5-year delay.
  - Production profile uses:
    - A plateau production level q_p that is field-size dependent.
    - A maximum depletion rate of remaining reserves d_m.
  - Production equals the plateau level as long as the depletion rate d(t) is lower than the maximum rate; when exceeded, depletion rate is held constant and production declines exponentially.
- Definitions:
  - Depletion rate of remaining reserves at time t: d(t) = q(t) / RRR(t), where q(t) is annual production rate and RRR(t) is remaining reserves.
  - Remaining reserves: RRR(t) = URR − Q(t), where URR is ultimately recoverable reserves and Q(t) is cumulative production.
- Field production profile (piecewise):
  - For t ≤ 5: production = 0 (no production during initial delay).
  - For t > 5 and while d(t) ≤ d_m: production = plateau q_p.
  - When depletion rate exceeds d_m: depletion held constant at d_m and production declines exponentially.
- Empirical basis:
  - Approximations of q_p and d_m derived from an oil field database by the Global Energy Systems research group at Uppsala University.
  - q_p specified via an empirical power law relationship between field size and plateau production.
  - d_m specified to capture higher depletion and decline rates of smaller fields compared to larger fields.
- Duration of plateau production, N, given by the solution to:
  - d(N) = q_p / (URR − q_p N).

### Parameter estimates for approximation functions (Table B.I)
- Plateau production rate estimates (Fields: α β R2)
  - All giants: 0.57 0.65 0.65
  - OPEC giants: 0.36 0.70 0.75
  - Non-OPEC giants: 0.67 0.64 0.58
- Maximum depletion rate estimates (Fields: γ δ R2)
  - All giants: 0.64 -0.31 0.31
  - OPEC giants: 0.40 -0.26 0.36
  - Non-OPEC giants: 0.66 -0.30 0.26
- Implementation choice:
  - A single set of parameter estimates corresponding to the pooled sample of all giant oil fields is used for computation of production profiles.

### Illustrative example
- Figure B.I illustrates production and depletion rate for a field with 500 million barrels of URR:
  - After the 5 years following discovery announcement, production rises, reaches a production plateau that lasts 5 years, then exponentially declines at a constant depletion rate.
  - Note: plateau duration varies with field size.

### Data description (variables and sources)
- Key macroeconomic variables, definitions, and sources (selected):
  - Real GDP: Logarithm of GDP in constant prices, local current unit. Source: IMF (2013).
  - Current account: % of GDP. Source: IMF (2013).
  - Investment: % of GDP as gross fixed capital formation. Source: World Bank (2013).
  - Saving/GDP: Constructed as the sum of current account and investment; robustness note that estimated dynamic effect of giant oil discoveries on saving is virtually unchanged if using World Bank (2013) saving data.
  - Real consumption: Logarithm of final consumption expenditures in constant local current unit. Source: IMF (2013).
  - Employment: Employment rate (employment to population ratio, percentage), age 15+, available from 1991; source: “emploare" from International Labor Organization website (www.ilo.org/kilm).
  - Real private investment: Logarithm of private gross fixed capital formation (constant price, local currency). Source: IMF (2013).
  - Real public investment: Logarithm of public gross fixed capital formation (constant price, local currency). Source: IMF (2013).
  - Real private consumption: Logarithm of private consumption expenditures (constant price, local currency). Source: IMF (2013).
  - Real public consumption: Logarithm of public consumption expenditures (constant price, local currency). Source: IMF (2013).

### Summary statistics of macro variables (1980-2012) (Table B.III, selected rows)
- Ln(GDP) — Years: 1980-2012; Maximum number of countries: 184; Obs: 5537; Min: -356.15; Median: 579.65; Max: 1477.83
- CA/GDP — Years: 1980-2012; Maximum number of countries: 180; Obs: 5408; Min: -242.19; Median: -3.2; Max: 106.84
- Saving/GDP — Years: 1980-2011; Maximum number of countries: 171; Obs: 4711; Min: -202.89; Median: 18.19; Max: 107.18
- Investment/GDP — Years: 1980-2011; Maximum number of countries: 178; Obs: 4956; Min: -2.42; Median: 21.13; Max: 113.58
- Ln(final consumption) — Years: 1980-2012; Maximum number of countries: 163; Obs: 4600; Min: -82.29; Median: 593.84; Max: 1435.58
- Employment rate — Years: 1991-2012; Maximum number of countries: 165; Obs: 3630; Min: 28.9; Median: 57.4; Max: 88.1
- Public investment/GDP — Years: 1980-2012; Maximum number of countries: 106; Obs: 1992; Min: -13.67; Median: 15.43; Max: 95.87
- Private investment/GDP — Years: 1980-2012; Maximum number of countries: 106; Obs: 1992; Min: -4.5; Median: 4.31; Max: 47.08
- Ln(final private consumption) — Years: 1980-2012; Maximum number of countries: 161; Obs: 4544; Min: -113.86; Median: 569.92; Max: 1426.54
- Ln(final public consumption) — Years: 1980-2012; Maximum number of countries: 162; Obs: 4556; Min: -268.61; Median: 425.7; Max: 1223.26
- Government spending/GDP — Years: 1980-2012; Maximum number of countries: 105; Obs: 1959; Min: 1.82; Median: 21.53; Max: 87.63
- Ln(real exchange rate based on CPI) — Years: 1980-2010; Maximum number of countries: 180; Obs: 4776; Min: 285.07; Median: 463.42; Max: 1528.4
- Ln(real exchange rate based on GDP deflator) — Years: 1980-2011; Maximum number of countries: 180; Obs: 5107; Min: 225.58; Median: 460.69; Max: 2608.46

*Source: _wp15209 - Appendix A. Theoretical Appendix (Appendix A and Appendix B content provided in the source PDF).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15209.pdf_