## wpiea2025015-print-pdf

## Source details

**Canonical URL:** [wpiea2025015-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025015-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025015-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025015-print-pdf.pdf.json)

---

### Background and motivation
- Commodity price shocks are significant sources of macroeconomic fluctuations, particularly for developing countries.
- Terms of trade shocks induced by commodity price fluctuations are used because:
  - They are a salient source of variation for less diversified developing economies.
  - They are plausibly exogenous (most countries are price takers), aiding causal identification.
- Leading exporters of major commodities are excluded from the sample to mitigate endogeneity related to commodity price making.
- Research emphasis: origins of macroeconomic institutions in response to external economic pressures rather than on the effectiveness of already existing institutions.

### Research question and scope
- Investigates whether and how commodity terms of trade shocks drive the adoption of:
  - Fiscal rules (budget balance, debt, expenditure, and revenue rules).
  - Financial liberalization / capital account openness.
- Focus on the adoption/choice of macroeconomic institutions in response to commodity-driven terms of trade shocks.

### Data and measurement
- Fiscal rules:
  - Source: dataset developed by Davoodi et al. (2022).
  - Measure: dummy variable = 1 if a country implemented at least one fiscal rule (budget balance, debt, expenditure, or revenue rule), 0 otherwise.
  - Coverage: 106 economies from 1985 to 2021.
- Capital account openness:
  - Measure: Chinn-Ito capital openness index (KAOPEN), ranges from 0 (closed) to 1 (open).
  - Coverage: 182 economies from 1970 to 2021.
- Commodity terms of trade shocks:
  - Source: Gruss and Kebhaj (2019).
  - Construction (Equations 1 and 2):
    - ∆Log(Index)_{i,t} = sum_{j=1}^{J} ∆P_{j,t} Ω_{i,j,t}
    - Ω_{i,j,t} = (1/3) sum_{s=1}^{3} x_{i,j,τ−s} − m_{i,j,τ−s} / GDP_{i,τ−s}
    - P_{j,t} denotes natural logarithm of the actual price of commodity j in year t.
    - x_{i,j,τ}, m_{i,j,τ}, and GDP_{i,τ} represent export value of commodity j, import value of commodity j, and nominal GDP of country i in year τ, respectively, measured in US dollars.
  - Interpretation: A 1 percent change in this index represents a proportional change in aggregate disposable income relative to GDP.
  - Additional indexes: commodity export price shocks (weighted by x_{i,j,τ−s} / GDP_{i,τ−s}) and commodity import price shocks (weighted by m_{i,j,τ−s} / GDP_{i,τ−s}).
  - Coverage: 182 economies from 1962 to 2022.
- Commodity classification used in price shock construction:
  - Energy: coal, crude oil, natural gas.
  - Metal commodities: aluminum, copper, gold, iron ore, lead, nickel, tin, uranium, zinc.
  - Food and beverages: bananas, barley, beef, cocoa, coffee, corn, fish, fish meal, groundnuts, lamb, olive oil, oranges, palm oil, poultry, rapeseed oil, rice, shrimp, soybean meal, soybean oil, soybeans, sugar, sunflower seed oil, swine meat, tea, wheat.
  - Agricultural raw materials: cotton, hard logs, hard sawn wood, hides, natural rubber, soft logs, soft sawn wood, wool.
- Control variables:
  - GDP growth (% annual) from WDI.
  - Electoral democracy index [0,1] from V-Dem.
  - Control of Corruption z-score (approximately -2.5 to 2.5) from WDI.

### Baseline empirical strategy
- Model (Equation (3)):
  - Macroe Inst_{i,t} = α_i + λ_t + γ(Price Shock)_{i,t−1} + Z′β + u_{i,t}
  - Macroe Inst_{i,t} denotes either the existence of a fiscal rule (binary) or the level of financial openness (KAOPEN).
  - Price shocks are country-specific variations in commodity terms of trade.
  - Z′ includes controls such as GDP growth, democracy, and control of corruption.
  - α_i and λ_t are country and year fixed effects.

### Main empirical findings
- Baseline regression (level-level; Commodity net export price shocks_{t−1}):
  - Fiscal rules (Column 1): 0.004*** (0.001)
  - Financial openness (Column 2): -0.003*** (0.000)
  - Observations: 4,648 (Fiscal rules), 7,117 (Financial openness)
  - R-squared: 0.560 (Fiscal rules), 0.695 (Financial openness)
  - Number of countries: 106 (Fiscal rules), 167 (Financial openness)
  - Number of years: 51 (both)
- Interpretation: higher commodity price shocks causally lead to the adoption of fiscal rules and reduced financial openness, consistent with “macroeconomic prudence.”
- Quantitative magnitudes:
  - A one standard deviation in commodity price shocks is associated with a 0.12 standard deviation increase in the adoption of fiscal rules.
  - A one standard deviation in commodity price shocks is associated with a 0.06 standard deviation decrease in financial openness.
- Asymmetry and heterogeneity:
  - Effects are asymmetric across export and import price shocks.
  - Income-group heterogeneity (interaction with income dummies):
    - Net export price shocks_{t−1}: 0.005*** (Fiscal rules) (0.001) ; -0.005*** (Financial openness) (0.001)
    - × middle income (p25–50): -0.004* (Fiscal rules) (0.002) ; 0.005*** (Financial openness) (0.002)
    - × low income (p25): -0.007** (Fiscal rules) (0.003) ; 0.005*** (Financial openness) (0.001)
    - Net effects by income group: fiscal rules adoption and financial openness reductions are driven by high-income countries (default), with diminished or reversed net effects for middle- and low-income groups.
  - Intensity thresholds (country-specific percentiles):
    - 90th percentile threshold_{t−1}: 0.049*** (Fiscal rules) (0.014) ; -0.025*** (Financial openness) (0.010)
    - 80th percentile threshold_{t−1}: 0.035*** (Fiscal rules) (0.011) ; -0.024*** (Financial openness) (0.008)
    - 70th percentile threshold_{t−1}: 0.025*** (Fiscal rules) (0.009) ; -0.019*** (Financial openness) (0.006)
    - 60th percentile threshold_{t−1}: 0.017** (Fiscal rules) (0.008) ; -0.015*** (Financial openness) (0.005)
    - Interpretation: stronger effects for higher-intensity shocks.

### Robustness checks and alternative specifications
- Alternative estimators:
  - Probit (fiscal rules): 0.011** (0.005)
  - Logit (fiscal rules): 0.021** (0.008)
  - Tobit (financial openness): -0.001* (0.000)
  - Poisson (financial openness): -0.006*** (0.002)
  - Results robust to estimator choice.
- Alternative dependent variables:
  - Exchange rate arrangements (6-category and 3-category) — positive terms of trade shocks linked to more rigid exchange rate regimes (negative coefficients). Examples:
    - Commodity export price shock (t−1) Column (1) OLS (6 categories): -0.007*** (0.002)
    - Ln(Net export price shock)_{t−1} Column (1) OLS: -0.383*** (0.128)
  - Loan-to-value (LTV) macroprudential limits (monthly and quarterly):
    - Monthly Table 7 (LTV average): Commodity export price shock_t: 0.033** (0.014); Commodity import price shock_t: -0.304*** (0.024); Net export price shock_t: 0.090*** (0.012)
    - Quarterly Annex VIII: Commodity export price shock (t) Column (1): 0.034 (0.024); Commodity import price shock (t) Column (2): -0.311*** (0.042); Net export price shock (t): 0.092*** (0.022)
    - Observations monthly: 22,000; Number of countries: 60; Number of months: 384; R-squared: 0.659–0.661 in monthly regressions.
    - Observations quarterly: 7,334; Number of countries: 60; Number of quarters: 128; R-squared: 0.661–0.663 in quarterly regressions.
- Additional controls and panels:
  - Results robust to inclusion of GDP growth, electoral democracy, control of corruption, regional analyses, asymmetry checks between export and import sides, and exclusion of top five leading commodity exporters.
  - Annex-level examples preserve coefficients such as:
    - Commodity net export price shocks_{t−1} on Fiscal Rules: 0.004*** (0.001)
    - Commodity net export price shocks_{t−1} on Financial Openness: -0.003*** (0.000)
    - Commodity export price shocks_{t−1} on Fiscal Rules: 0.005*** (0.001)
    - Commodity import price shocks_{t−1} on Fiscal Rules: -0.005** (0.002)
    - Commodity import price shocks_{t−1} on Financial Openness: 0.003** (0.001)

### Regional and specification heterogeneity (selected figures)
- Regional level-level results for Net export price shocks_{t−1} on Fiscal Rules:
  - AFRICA: 0.002** (0.001)
  - AMERICAS: 0.018*** (0.002)
  - EUROPE: 0.023*** (0.007)
- Regional level-level results for Net export price shocks_{t−1} on Financial Openness:
  - AFRICA: -0.001* (0.001)
  - AMERICAS: -0.003*** (0.001)
  - ASIA: -0.003*** (0.001)

### Policy implications and recommendations
- Adopt robust fiscal rules and strengthen financial regulations to mitigate volatility induced by commodity price shocks.
- Ensure consistent enforcement of fiscal rules to maintain stable and predictable fiscal policies during commodity price fluctuations.
- Consider limiting financial openness during positive commodity price shocks to prevent economic overheating and the accumulation of excessive financial risks.
- Enhance financial regulatory frameworks to manage capital inflows and outflows to ensure financial stability and reduce vulnerability to external shocks.
- Suggested further research: explore differentiated effects of commodity price shocks across types of commodities (point source vs. diffuse resources).

### Key descriptive statistics (Annex I examples)
- Fiscal rules (0 or 1 dummy): Observations 4,996; Mean 0.181; Standard deviation 0.385; Min 0.000; Max 1.000
- Financial Openness [0,1] (Chinn and Ito (2008)): Observations 7,462; Mean 0.459; Standard deviation 0.362; Min 0.000; Max 1.000
- Commodity net export price shock (IMF calculation): Observations 9,855; Mean 99.801; Standard deviation 13.256; Min 35.610; Max 177.005
- Commodity export price shock: Observations 9,855; Mean 94.895; Standard deviation 11.229; Min 35.552; Max 132.391
- Commodity import price shock: Observations 9,855; Mean 95.453; Standard deviation 5.655; Min 47.053; Max 109.748
- GDP growth % (WDI): Observations 7,697; Mean 3.458; Standard deviation 5.918; Min -64.047; Max 88.958
- Electoral democracy index [0,1] (V-Dem): Observations 9,596; Mean 0.433; Standard deviation 0.287; Min 0.007; Max 0.924
- Control of Corruption z-score (WDI): Observations 4,084; Mean -0.042; Standard deviation 0.996; Min -1.782; Max 2.459

_Source: wpiea2025015-print-pdf - IMF Working Paper "Shocks and Shields: Macroeconomic Institutions During Commodity Price Swings"._

### 1. Introduction ........................................................................................................

### 1. Introduction

### Background and motivation
- Commodity price shocks are significant sources of macroeconomic fluctuations, particularly for developing countries.
- Existing literature focuses on macroeconomic consequences of commodity export windfalls (e.g., resource curse, Dutch disease, excessive indebtedness, conflicts, erosion of democracy) but has paid limited attention to how commodity export windfalls influence macroeconomic institutions.
- Terms of trade shocks induced by commodity price fluctuations are used because:
  - They are a salient source of variation for less diversified developing economies.
  - They are plausibly exogenous (most countries are price takers), aiding causal identification.
- Leading exporters of major commodities are excluded from the sample to mitigate endogeneity related to commodity price making.

### Research question and scope
- The paper investigates whether and how commodity terms of trade shocks drive the adoption of:
  - Fiscal rules (budget balance, debt, expenditure, and revenue rules).
  - Financial liberalization / capital account openness.
- Emphasis on origins of macroeconomic institutions in response to external economic pressures rather than on the effectiveness of already existing institutions.

### Key findings (from the Introduction)
- Countries facing commodity (net) export price shocks tend to:
  - Adopt fiscal rules.
  - Close their economies financially (i.e., reduce financial openness).
- Effects are asymmetric across import and export price shocks.
- The impact varies with:
  - Intensity of the shocks.
  - Income levels of countries, with higher income showing the most significant results.
- Findings are robust to:
  - Different estimators.
  - Additional control and dependent variables.
- Interpretation: macroeconomic institutions respond to terms of trade shocks driven by commodity prices, reflecting “macroeconomic prudence.”

### Relation to literature
- Connects to political-economy literature on the timing and drivers of reforms (e.g., Alesina and Drazen, 1991; Ranciere and Tornell, 2015; Alesina et al., 2023).
- Complements literature on the effectiveness of fiscal rules and financial liberalization (e.g., Frankel et al., 2013; Pieschacón, 2012; Bekaert et al., 2005).
- Uses commodity export shocks as an identification strategy to link terms of trade shocks and macroeconomic institutions.

### Data and measures (summary drawn from Section 2)
- Fiscal rules:
  - Source: dataset developed by Davoodi et al. (2022).
  - Measure: dummy variable = 1 if a country implemented at least one fiscal rule (budget balance, debt, expenditure, or revenue rule), 0 otherwise.
  - Coverage: 106 economies from 1985 to 2021.
- Capital account openness (financial openness):
  - Measure: Chinn-Ito capital openness index (KAOPEN), ranges from 0 (closed) to 1 (open).
  - Coverage: 182 economies from 1970 to 2021.
- Commodity terms of trade shocks:
  - Source: Gruss and Kebhaj (2019).
  - Construction (Equations 1 and 2 in the source):
    - ∆Log(Index)_{i,t} = sum_{j=1}^{J} ∆P_{j,t} Ω_{i,j,t}
    - Ω_{i,j,t} = (1/3) sum_{s=1}^{3} x_{i,j,τ−s} − m_{i,j,τ−s} / GDP_{i,τ−s}
    - P_{j,t} denotes natural logarithm of the actual price of commodity j in year t.
    - x_{i,j,τ}, m_{i,j,τ}, and GDP_{i,τ} represent export value of commodity j, import value of commodity j, and nominal GDP of country i in year τ, respectively, measured in US dollars.
  - Interpretation: A 1 percent change in this index represents a proportional change in aggregate disposable income relative to GDP.
  - Additional indexes:
    - Commodity export price shocks weighted by x_{i,j,τ−s} / GDP_{i,τ−s}.
    - Commodity import price shocks weighted by m_{i,j,τ−s} / GDP_{i,τ−s}.
  - Coverage: 182 economies from 1962 to 2022.
- Commodity classification (as used in the price shock construction):
  - Energy: coal, crude oil, natural gas.
  - Metal commodities: aluminum, copper, gold, iron ore, lead, nickel, tin, uranium, zinc.
  - Food and beverages: bananas, barley, beef, cocoa, coffee, corn, fish, fish meal, groundnuts, lamb, olive oil, oranges, palm oil, poultry, rapeseed oil, rice, shrimp, soybean meal, soybean oil, soybeans, sugar, sunflower seed oil, swine meat, tea, wheat.
  - Agricultural raw materials: cotton, hard logs, hard sawn wood, hides, natural rubber, soft logs, soft sawn wood, wool.

### Control variables included
- GDP growth:
  - Source: World Development Indicators (WDI) of the World Bank.
  - Measure: annual percentage growth rate of GDP at market prices calculated using constant local currency.
- Electoral democracy:
  - Source: The V-Dem Dataset.
  - Rationale: democratic countries have different trade and financial policy stances; democracy helps control for coordination issues around macroeconomic institutions.
- Control of Corruption:
  - Source: WDI (World Governance Institute / Kaufmann et al. 2010).
  - Measure: country score on aggregate indicator in units of a standard normal distribution, ranging from approximately -2.5 to 2.5.
- Annex I contains descriptive statistics for variables used in empirical analysis.

*Source: wpiea2025015-print-pdf - 1. Introduction (IMF Working Paper "Shocks and Shields: Macroeconomic Institutions During Commodity Price Swings")*

### 3. Main Results

### 3. Main Results

### Empirical strategy and baseline model
- Estimated model (Equation (3)):  
  Macroe Inst_{i,t} = α_i + λ_t + γ(Price Shock)_{i,t−1} + Z′β + u_{i,t}  
  - Macroe Inst_{i,t} denotes either the existence of a fiscal rule (binary) or the level of financial openness for country i at year t.  
  - Price shocks are country-specific variations in terms of trade commodity price shocks.  
  - Z′ includes controls such as GDP growth, democracy, and control of corruption.  
  - α_i are country fixed effects; λ_t are year fixed effects; u_{i,t} is the error term.  
- Baseline shock measure: net export commodity price index (accounts for changes in both export and import sides).  
- Also examined separately: effects of commodity terms of trade changes on exports and imports (asymmetry explored in Annex II).

### Baseline regression results (Table 1)
- Dependent variables: FISCAL RULES (binary) and FINANCIAL OPENNESS (Chinn-Ito index KAOPEN, continuous [0, 1]).  
- Main coefficient on Commodity net export price shocks_{t−1}:  
  - Fiscal rules (Column 1): 0.004*** (standard error (0.001))  
  - Financial openness (Column 2): -0.003*** (standard error (0.000))  
- Sample and fit statistics:  
  - Observations: 4,648 (Fiscal rules), 7,117 (Financial openness)  
  - R-squared: 0.560 (Fiscal rules), 0.695 (Financial openness)  
  - Number of countries: 106 (Fiscal rules), 167 (Financial openness)  
  - Number of years: 51 (both)  
- Interpretation: higher commodity price shocks causally lead to the adoption of fiscal rules and reduced financial openness — consistent with “macroeconomic prudence” where governments commit fiscal institutions and limit financial openness in response to positive shocks.  
- Quantitative magnitudes:  
  - A one standard deviation in commodity price shocks is associated with a 0.12 standard deviation increase in the adoption of fiscal rules.  
  - A one standard deviation in commodity price shocks is associated with a 0.06 standard deviation decrease in financial openness.  
- Economic significance: statistically significant but relatively small economically on average; potential heterogeneity across income levels and shock intensities noted.

### Income-group heterogeneity (Table 2)
- Augmented regressions include interactions between net export commodity price shocks_{t−1} and income-group dummies (middle-income and low-income).  
- Default income group: higher income (coefficients on shocks correspond to high-income countries).  
- Coefficients (Columns 1 and 2):  
  - Net export price shocks_{t−1}: 0.005*** (Fiscal rules) (0.001) ; -0.005*** (Financial openness) (0.001)  
  - × middle income (p25–50): -0.004* (Fiscal rules) (0.002) ; 0.005*** (Financial openness) (0.002)  
  - × low income (p25): -0.007** (Fiscal rules) (0.003) ; 0.005*** (Financial openness) (0.001)  
- Sample and fit: Observations 4,648 (Fiscal rules), 7,117 (Financial openness); R-squared 0.561 and 0.698; Number of countries 106 and 167; Number of years 51 and 51.  
- Net effects by income group:  
  - Fiscal rules: positive for high-income (default), positive for middle-income (default + interaction), negative for low-income (default + interaction).  
  - Financial openness: default negative for high-income; net effect for middle and low-income is zero (default plus interaction).  
- Interpretation: “macroeconomic prudence” (commitment to fiscal institutions and limiting financial openness) is driven by high-income countries.

### Intensity heterogeneity (Table 3)
- Approach: threshold dummy variables equal to 1 if shock exceeds pre-set country-specific percentiles (90th, 80th, 70th, 60th), 0 otherwise.  
- Selected coefficient magnitudes (Columns across fiscal rules and financial openness):  
  - 90th percentile threshold_{t−1}: 0.049*** (Fiscal rules) (0.014) ; -0.025*** (Financial openness) (0.010)  
  - 80th percentile threshold_{t−1}: 0.035*** (Fiscal rules) (0.011) ; -0.024*** (Financial openness) (0.008)  
  - 70th percentile threshold_{t−1}: 0.025*** (Fiscal rules) (0.009) ; -0.019*** (Financial openness) (0.006)  
  - 60th percentile threshold_{t−1}: 0.017** (Fiscal rules) (0.008) ; -0.015*** (Financial openness) (0.005)  
- Sample and fit (examples): Observations 4,768 (fiscal rule columns), 7,233 (financial openness columns); R-squared around 0.550–0.688; Number of countries 106 (fiscal rules) and 167 (financial openness); Number of years 51.  
- Interpretation: main findings are driven by higher-intensity shocks — coefficients larger in absolute value for higher-intensity thresholds, indicating greater government prudence in response to more intense shocks.

### Extensions and robustness checks (Section 4; Tables 4–7)
- Controls, alternative estimators, and alternative dependent variables considered; main findings robust across these checks.

- Table 4 (additional controls; Panel A level-level and Panel B level-log specifications):  
  - Net export price shocks_{t−1} coefficients (examples):  
    - Level-level (Panel A, Columns 1–3): 0.004*** (Fiscal rules) (0.001) ; Level-level (Panel A, Columns 4–6): -0.001** to -0.002*** to -0.001* (Financial openness) (standard errors vary).  
    - Level-log (Panel B, Columns 1–3): Ln(Net export price shocks)_{t−1}: 0.356*** (0.070), 0.359*** (0.070), 0.328*** (0.094) for Fiscal rules; -0.107*** (0.040), -0.168*** (0.042), -0.081** (0.040) for Financial openness.  
  - Additional control coefficients (examples):  
    - GDP growth: -0.002***, -0.002**, -0.002 (Fiscal rules columns) and 0.002***, 0.001***, -0.000 (Financial openness columns) with standard errors (0.001) or (0.000)/(0.001).  
    - Electoral democracy index: 0.138*** (0.039) and -0.166*** (0.063) in reported columns; -0.141*** (0.024) and -0.080** (0.033) in other columns.  
    - Control of Corruption: 0.101*** (0.025) and 0.044*** (0.011) in reported columns.  
  - Samples and fit vary by column: Observations range (e.g., 4,306; 4,062; 2,266 for Fiscal rules panels) and R-squared range (e.g., 0.572; 0.561; 0.752; 0.721; 0.726; 0.893).  
  - Note: GDP growth (% annual) and control of corruption (z-score) sourced from World Development Indicators; Electoral democracy [0,1] from V-Dem.

- Table 5 (different estimators): discrete choice for fiscal rules (Probit, Logit), Tobit and Poisson for financial openness.  
  - Commodity net export price shocks_{t−1} coefficients:  
    - Probit: 0.011** (0.005)  
    - Logit: 0.021** (0.008)  
    - Tobit: -0.001* (0.000)  
    - Poisson: -0.006*** (0.002)  
  - Observations: 4,648 (Probit), 2,315 (Logit), 7,117 (Tobit), 7,117 (Poisson).  
  - Number of countries: 106 (Probit), 52 (Logit), 167 (Tobit/Poisson).  
  - Number of years: 51 (all).  
  - Interpretation: main results robust to estimator choice.

- Alternative dependent variables (Tables 6 and 7): exchange rate regime choice and loan-to-value (LTV) macroprudential limits.  
  - Table 6 (exchange rate arrangements; OLS, Tobit, Logit; 6-category and 3-category measures): net export price shocks_{t−1} coefficients example:  
    - (6 categories) OLS Column (1): -0.004*** (0.002) ; Tobit Column (2): -0.019*** (0.004) ; Logit Column (5): -0.024*** (0.004)  
    - (3 categories) OLS Column (12): -0.002** (0.001) ; Tobit Column (13): -0.011*** (0.003) ; Logit Column (16): -0.016*** (0.005)  
  - Observations: 8,738 (all reported columns); Number of countries: 170; Number of years: 57.  
  - Interpretation: positive terms of trade shocks are linked to a choice of a more rigid exchange rate regime (negative coefficient).

  - Table 7 (monthly LTV limits): Column 1 shows a positive and statistically significant association between commodity price shock and LTV limits; Columns 2 and 3 confirm asymmetry for export and import sides (detailed coefficients presented in the source).

- Robustness summary: results remain robust to inclusion of additional control variables (Annex VI), to asymmetry checks between export and import sides (Annex II, Annex III, Annex VII), to regional group analysis (Annex IV), and to extensive threshold specifications (Annex V). Quarterly and monthly frequency checks for LTV limits are reported in Annexes VIII and III respectively.

*Source: wpiea2025015-print-pdf - 3. Main Results*

### Annex VIII confirms that our main results are robust to the use of higher frequency data.

### wpiea2025015-print-pdf - Annex VIII confirms that our main results are robust to the use of higher frequency data.

### Robustness to higher-frequency (monthly) data
- Annex VIII confirms that the paper's main results are robust to the use of higher frequency data (monthly).
- Table 7 (monthly frequency) uses the average of regulatory loan-to-value (LTV) limits as the dependent variable and reports:
  - Commodity export price shock_t coefficient: 0.033** (standard error 0.014)
  - Commodity import price shock_t coefficient: -0.304*** (standard error 0.024)
  - Net export price shock_t coefficient: 0.090*** (standard error 0.012)
  - Countries FEs: Yes; Time FEs: Yes
  - Observations: 22,000; Number of countries: 60; Number of months: 384
  - R-squared: 0.659 in Column (1) and 0.661 in Column (2)
- Notes on LTV variable:
  - Dependent variable represents the simple average of regulatory LTV limits for real estate mortgage loans (residential and commercial).
  - If a country does not have any LTV limits, the value is set at 100.
  - Sample excludes the top five world leading exporters of hydrocarbons, minerals, and agriculture based on 2022 data (e.g., the United States, Russia, Australia, Saudi Arabia, and Canada) to account for potential endogeneity.
- Estimation details elsewhere in the document:
  - Exchange rate arrangements variable coded into groups and estimated using Tobit (lower limit 1, upper limit 6) and logistic regression; column [2] reports 4,038 left-censored, 4,341 uncensored, and 359 right-censored observations; columns [5] reports 6,088 left-censored, 2,650 uncensored, and 0 right-censored observations.
  - Ordered logistic LR test p-value reported as 0.0000 (ordered logistic model fits significantly better than the simpler model).

### Key empirical findings across panels and specifications
- Main conclusion (summary):
  - Countries facing commodity (net) export price shocks tend to adopt fiscal rules and to financially close their economies; effects are asymmetric across import and export price shocks.
  - The impact varies with shock intensity and income levels, with higher-income countries driving main results.
  - Results are robust to different estimators and additional control variables.
- Selected coefficient estimates (preserving reported precision and significance):
  - Annex II (level-level panel; lagged shocks):
    - Commodity net export price shocks_{t−1} on Fiscal Rules: 0.004*** (0.001)
    - Commodity net export price shocks_{t−1} on Financial Openness: -0.003*** (0.000)
    - Commodity export price shocks_{t−1} on Fiscal Rules: 0.005*** (0.001)
    - Commodity import price shocks_{t−1} on Fiscal Rules: -0.005** (0.002); on Financial Openness: 0.003** (0.001)
    - Observations in panels: 4,648 and 7,117; Number of countries: 106 and 167; Number of years: 51
    - R-squared values: 0.560, 0.695, 0.696 as reported
  - Annex II (level-log panel):
    - Ln(Commodity import price shocks)_{t−1} on Fiscal Rules: 0.365*** (0.064)
    - Ln(Commodity net export price shocks)_{t−1} on Fiscal Rules: 0.364*** (0.072)
    - Ln(Commodity export price shocks)_{t−1} on Fiscal Rules: -0.374** (0.188)
    - Ln specifications R-squared: 0.560 and 0.695
  - Annex III (income interaction results):
    - Export price shocks_{t−1} on Fiscal Rules: 0.005*** (0.001); Export price shocks_{t−1} × middle income(p25−50): -0.006*** (0.002)
    - Import price shocks_{t−1} on Fiscal Rules: -0.009*** (0.002); Import price shocks_{t−1} × low income(p25): 0.011*** (0.003)
    - Import price shocks_{t−1} on Financial Openness: 0.006*** (0.001); Import price shocks_{t−1} × middle income(p25−50) on Financial Openness: -0.013*** (0.002)
    - Observations: 4,648 and 7,117; Number of countries: 106 and 167; R-squared: 0.561 and 0.696 (or 0.560 and 0.697 depending on panel)
  - Annex IV (regional analysis, level-level):
    - Net export price shocks_{t−1} on Fiscal Rules, selected regions:
      - AFRICA: 0.002** (0.001)
      - AMERICAS: 0.018*** (0.002)
      - EUROPE: 0.023*** (0.007)
    - Net export price shocks_{t−1} on Financial Openness, selected regions:
      - AFRICA: -0.001* (0.001)
      - AMERICAS: -0.003*** (0.001)
      - ASIA: -0.003*** (0.001)
    - Observations by region example: AFRICA fiscal rules observations 2,138; AMERICAS 1,002; Number of countries examples: AFRICA 46; AMERICAS 21
  - Annex V (threshold analysis):
    - Fiscal Rules: coefficients for exceeding percentile thresholds (lagged) are positive and significant at higher percentiles, e.g.:
      - 90th percentile threshold_{t−1}: 0.049*** (0.014)
      - 80th percentile threshold_{t−1}: 0.035*** (0.011)
      - 60th percentile threshold_{t−1}: 0.017** (0.008)
    - Financial Openness: coefficients for exceeding percentile thresholds (lagged) are negative and significant, e.g.:
      - 90th percentile threshold_{t−1}: -0.025*** (0.010)
      - 50th percentile threshold_{t−1}: -0.017*** (0.005)
      - 20th percentile threshold_{t−1}: -0.042*** (0.007)
    - Observations: Fiscal Rules panels have 4,768 observations; Financial Openness panels have 7,233 observations; Number of countries: 106 for Fiscal Rules, 167 for Financial Openness; Number of years: 51
  - Annex VI (with controls):
    - Commodity export price shocks_{t−1} on Fiscal Rules with controls: 0.004*** (0.001)
    - Commodity import price shocks_{t−1} on Fiscal Rules with controls: -0.011*** (0.003) and in some specifications -0.010*** (0.003)
    - GDP growth reported coefficient on Fiscal Rules: -0.002*** or -0.002** (standard error 0.001) across specifications
    - Electoral democracy index effects on Fiscal Rules: 0.139*** (0.039) in some columns and -0.163** (0.063) in interactions
    - Control of Corruption on Fiscal Rules: 0.102*** (0.025)
    - Commodity export shocks (ln)_{t−1} on Fiscal Rules (level-log with controls): 0.333*** (0.079), 0.341*** (0.078), 0.361*** (0.106)
    - Commodity export shocks (ln)_{t−1} on Financial Openness (level-log with controls): -0.119*** (0.045), -0.133*** (0.045)
    - Observations and sample sizes in Annex VI:
      - Fiscal Rules with controls Observations: 4,306; 4,062; 2,266 in different columns
      - Financial Openness with controls Observations: 6,331; 5,948; 3,295 in different columns
      - Number of countries and years vary by specification (examples: 106, 100 countries; 49, 48, 23 years)

### Policy implications emphasized in the paper
- Recommended policy responses based on findings:
  - Adopt robust fiscal rules and strengthen financial regulations to mitigate volatility induced by commodity price shocks.
  - Ensure consistent enforcement of fiscal rules to maintain stable and predictable fiscal policies during commodity price fluctuations.
  - Consider limiting financial openness during positive commodity price shocks to prevent economic overheating and the accumulation of excessive financial risks.
  - Enhance financial regulatory frameworks to manage capital inflows and outflows to ensure financial stability and reduce vulnerability to external shocks.
- Suggested directions for further research:
  - Explore differentiated effects of commodity price shocks across types of commodities (point source vs. diffuse resources), noting Isham et al. (2005) finding that point source natural resources such as minerals and hydrocarbons are associated with lesser growth performance than diffuse resources such as agriculture products.

### Data, variables, and measurement notes
- Key variables and descriptive statistics (Annex I):
  - Fiscal rules (0 or 1 dummy, Davoodi et al. (2022)): Observations 4,996; Mean 0.181; Standard deviation 0.385; Min 0.000; Max 1.000
  - Financial Openness [0,1] (Chinn and Ito (2008)): Observations 7,462; Mean 0.459; Standard deviation 0.362; Min 0.000; Max 1.000
  - Commodity net export price shock (IMF calculation): Observations 9,855; Mean 99.801; Standard deviation 13.256; Min 35.610; Max 177.005
  - Commodity export price shock: Observations 9,855; Mean 94.895; Standard deviation 11.229; Min 35.552; Max 132.391
  - Commodity import price shock: Observations 9,855; Mean 95.453; Standard deviation 5.655; Min 47.053; Max 109.748
  - GDP growth % (WDI): Observations 7,697; Mean 3.458; Standard deviation 5.918; Min -64.047; Max 88.958
  - Electoral democracy index [0,1] (V-Dem): Observations 9,596; Mean 0.433; Standard deviation 0.287; Min 0.007; Max 0.924
  - Control of Corruption z-score (WDI): Observations 4,084; Mean -0.042; Standard deviation 0.996; Min -1.782; Max 2.459
- Data sources and notes:
  - Commodity net export price shocks represent windfall gains and losses in income associated with fluctuations in global prices, considering both export and import prices.
  - Fiscal rule dummy equals 1 if at least one of budget balance rules (BBR), debt rules (DR), expenditure rules (ER), or revenue rules (RR) is enacted for central/general/public sector; equals 0 if none implemented or data missing.
  - Financial openness measured primarily by the Chinn-Ito index (KAOPEN).
  - Several samples exclude the top five world leading exporters of hydrocarbons, minerals, and agriculture based on 2022 data (the United States, Russia, Australia, Saudi Arabia, and Canada) to mitigate potential endogeneity from commodity price making.

*Source: IMF Working Paper "Shocks and Shields: Macroeconomic Institutions During Commodity Price Swings" (Annexes I–VI and Table 7 as provided).*

### Annex VII. Commodity Export and Import Price

### Annex VII. Commodity Export and Import Price Shock and Exchange Rate Arrangements

### PANEL A: Level-level regression model (Dependent variables: Exchange Rate Arrangements — 6 categories; Exchange Rate Arrangements — 3 categories)
- Estimators reported: OLS, Tobit, Logit (columns (1), (2), (5) for 6 categories; columns (12), (13), (16) for 3 categories).
- Commodity export price shock (t−1) coefficients:
  - Column (1) OLS: -0.007*** (0.002)
  - Column (2) Tobit: -0.024*** (0.004)
  - Column (5) Logit: -0.029*** (0.005)
  - Column (12) OLS: -0.003*** (0.001)
  - Column (13) Tobit: -0.018*** (0.003)
  - Column (16) Logit: -0.026*** (0.005)
- Commodity import price shock (t−1) coefficients:
  - Column (1) OLS: -0.002 (0.004)
  - Column (2) Tobit: -0.006 (0.007)
  - Column (5) Logit: 0.004 (0.007)
  - Column (12) OLS: -0.002 (0.002)
  - Column (13) Tobit: -0.024*** (0.006)
  - Column (16) Logit: -0.036*** (0.009)
- Fixed effects and sample statistics:
  - Countries FEs: Yes
  - Year FEs: Yes
  - Observations: 8,738 (for all columns)
  - R-squared: 0.507 (reported for one specification), 0.450 (reported for another)
  - Number of countries: 170 (for all columns)
  - Number of years: 57 (for all columns)

### PANEL B: Level-log regression model (Dependent variables: Exchange Rate Arrangements — 6 categories; Exchange Rate Arrangements — 3 categories)
- Estimators reported: OLS, Tobit, Ologit (columns (1), (2), (5) for 6 categories; columns (12), (13), (16) for 3 categories).
- Ln(Net export price shock) (t−1) coefficients:
  - Column (1) OLS: -0.383*** (0.128)
  - Column (2) Tobit: -1.291*** (0.310)
  - Column (5) Ologit: -1.640*** (0.338)
  - Column (12) OLS: -0.138** (0.065)
  - Column (13) Tobit: -0.847*** (0.245)
  - Column (16) Ologit: -1.230*** (0.388)
- Fixed effects and sample statistics:
  - Countries FEs: Yes
  - Year FEs: Yes
  - Observations: 8,738 (for all columns)
  - R-squared: 0.506 (reported for one specification), 0.450 (reported for another)
  - Number of countries: 170 (for all columns)
  - Number of years: 57 (for all columns)

### Annex VIII. Commodity Export and Import Price Shock and Loan to Value limits (Quarterly data)

### Regression results (Dependent variable: Average of the regulatory loan-to-value (LTV) limits)
- Columns reported: (1), (2)
- Commodity export price shock (t) coefficients:
  - Column (1): 0.034 (0.024)
- Commodity import price shock (t) coefficients:
  - Column (2): -0.311*** (0.042)
- Net export price shock (t) coefficients:
  - Column (1) or as reported: 0.092*** (0.022)
- Fixed effects and sample statistics:
  - Countries FEs: Yes (both columns)
  - Time FEs: Yes (both columns)
  - Observations: 7,334 (both columns)
  - R-squared: 0.661 (Column (1)), 0.663 (Column (2))
  - Number of countries: 60
  - Number of quarters: 128
- Notes on dependent variable:
  - The dependent variable represents the simple average of the regulatory loan-to-value (LTV) limits (LTV_average).
  - Column [1] specifically focuses on LTV limits for real estate mortgage loans (both residential and commercial), while dummy-type indicators and text information may cover other types of loans (e.g., auto loans).
  - If a country does not have any LTV limits, the value is set at 100.

*IMF WORKING PAPERS — Shocks and Shields: Macroeconomic Institutions During Commodity Price Swings*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025015-print-pdf.pdf_