## Annex 1. Data

## Source details

**Canonical URL:** [Annex 1. Data](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1636.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1636.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1636.pdf.json)

---

### Dataset and scope
- New dataset covering 1970–2014 for non-renewable resource-rich countries (oil/gas and metals).
- Country selection criterion: share of commodity exports in total exports and share of commodity revenues in total government revenues of at least 15 percent on average for a five year period (either 2007-11 or 2009-13, depending on data availability).
- Commodity price series are real prices in USD deflated by the U.S. GDP deflator (Table 1 provides descriptive statistics).
- Commodity price index for each country: P_i = Σ_j P_ij * w_ij where j indexes commodity types (oil, gas, gold, tin, zinc, lead, aluminum, nickel, copper, and silver), P is the real commodity price (deflated by the U.S. CPI in formula exposition), and w is the commodity weight (commodity export share in GDP).

### Size and dynamics of commodity price fluctuations
- Average duration of commodity price upswings (downswings): 2-4 years.
- Standard deviation of durations is large; some expansions or contractions can last up to 10 years (Table 2).
- Average amplitude of booms (percentage change from trough to peak) and busts (percentage change from peak to trough):
  - Booms: large, ranging from 40-50 percent (example: iron ore) and 80 percent (example: natural gas).
  - Busts: range from 35-80 percent.
  - Some booms or busts can exceed 200 percent in amplitude.
- Metals, minerals, and oil sectors tend to have relatively longer boom/bust durations due to long lags between investment and supply response.

### Direct impact on fiscal revenues and exports
- Estimated average direct impact of commodity price swings (Figure 1):
  - Fiscal revenues: 8-13 percent of GDP.
  - Exports: 2-10 percent of GDP.
- Relative impact larger on fiscal revenues in oil exporters, suggesting stronger transmission through the fiscal channel.
- Asymmetry: impact is stronger in downswings compared to upswings.

### Historical macroeconomic consequences
- 1970s-80s example: many countries experienced real revenue increases of close to 10 percent a year during the boom and subsequent large falls in the bust (Figure 2).
- After the 1970s-80s bust many commodity exporters experienced long periods of negative or stagnant growth.
- Recent large revenue windfalls in the 2000s have been followed by large falls in commodity prices, forcing many countries into large fiscal contractions during weak economic activity.

### Fiscal cyclicality — conceptual approaches used
- Two empirical approaches applied:
  1. Measure 1 — responsiveness of public expenditure growth rates to year-to-year changes in commodity prices (uses Δlog(RG_it) = α_i + β Δlog(P_it) + ε_it). Positive β implies procyclicality.
  2. Measure 2 — reaction of cyclically adjusted non-resource balance to the non-resource output gap (relationship between non-resource fiscal balance and non-resource output gap).
- Rationale:
  - Commodity prices are exogenous to domestic cycles and spending policies, alleviating endogeneity issues for Measure 1.
  - Cyclically adjusted non-resource balance isolates fiscal stance excluding commodity component for Measure 2.
- Analysis uses changes (differences) to abstract from long-run level correlations; panel cointegration tests (Westerlund, 2007) found no co-integration between levels of spending and prices.

### Empirical findings on procyclicality (Measure 1 results)
- Commodity prices have a positive impact on government spending (Table 4): fiscal policy has been procyclical.
- Magnitude:
  - A 10 percent increase (fall) in commodity prices leads to a 1.2 percent increase (fall) in real expenditure growth (example coefficient 0.119** [0.052] reported).
  - Example implication: If oil prices fall by 50 percent (as in the second half of 2014), expenditures would need to contract by 6.5 percent on average—this occurs while economic growth is rapidly decelerating.
- Capital spending is even more procyclical compared to total spending (reported coefficient noted as 0.15 in text fragment; examples: Column V 0.147* [0.081], Column VIII 0.169** [0.086]).
- Procyclicality is stronger during commodity price expansions than contractions (interaction Δ log of commodity prices * expansion dummy positive and significant; interaction with contraction dummy generally not significant).
- Procyclicality has not changed significantly over time in the sample overall.
- Some countries exhibited countercyclical behavior temporarily (e.g., in 2009), but these episodes were short-lived relative to overall patterns.

### Measure 2: The cyclically-adjusted non-resource balance
- Specification: CA_BAL_NC (cyclically adjusted non-resource balance), GDP_NC (non-resource GDP), GAP_NC (non-resource GDP gap). Coefficient β captures fiscal cyclicality (negative ⇒ procyclicality). Fixed effects used; robustness check replaces output gap with non-resource GDP growth.
- Main findings:
  - Governments loosen fiscal stance when the domestic non-resource economy strengthens and tighten when it weakens (Table 5).
  - A 1 percentage point improvement in the non-resource output gap leads to a 1 percentage point deterioration of the cyclically adjusted non-resource balance as a share of potential non-commodity GDP (example coefficient -0.945*** [0.225]).
  - Replacing output gap with real GDP growth rates does not alter the negative association (example coefficient for non-commodity GDP growth -0.215*** [0.011]).
  - Commodity exporters tend to be more procyclical than other emerging economies (comparison: IMF (2015b) coefficient ~0.5 for emerging markets vs. ~1 here).

### Can institutions help reduce procyclicality?
- Analytical approach:
  - Focus on rules designed to regulate accumulation/use of resource revenues, including rules for functioning of a fund (saving or stabilization). Estimation includes dummies for fiscal rules and for when a fund is in place.
  - Institutional measures from World Governance Indicators and International Risk Group: bureaucratic quality, corruption, political risk, institutional and legal setting; Polity IV used for democratic institutions.
  - Specification interacts commodity price index with institutional quality and fiscal rules/resource funds. Coefficient γ measures effect on procyclicality (negative γ ⇒ reduction).
- Results on rules and funds:
  - Interaction term negative (consistent with reduction in procyclicality) but generally not statistically significant (example Δ log of commodity prices * Fiscal rule dummy -0.078 [0.088]; * Savings fund dummy -0.022 [0.071]; * Stabilization fund dummy -0.058 [0.064]).
  - Many countries struggle with compliance and efficacy of special fiscal institutions; notable successes limited to Botswana, Chile, and Norway.
- Reasons for limited success:
  - Existence of a rule/fund does not imply de facto compliance; many rules breached, especially in bad times.
  - Causes: lack of political will, poor rule design, absence of monitoring/enforcement.
  - Country examples of weaknesses provided in source text.
- Institutional quality effects:
  - Evidence that higher political/institutional quality helps limit procyclical bias in spending.
  - Procyclicality would be eliminated in countries with bureaucratic quality or institutional/legal setting around two standard deviations above the mean (interaction Δ log commodity prices * Bureaucratic quality -0.087*** [0.029]; * Political risk -0.007*** [0.002]; * Institutional and legal setting -0.003* [0.001]).
  - Average institutional quality tends to be weaker in resource-rich countries than in other countries.
- Successful country examples:
  - Norway, Chile, Botswana: higher institutional quality supported disciplined fiscal policy and stronger long-term growth.
  - Norway specifics: non-oil deficit target anchored on 4 percent of assets in oil fund over the economic cycle; oil/gas revenue saved in oil fund with only returns used for budget; rule insulates budget from yearly oil/gas price movements and supported sustained GDP per capita growth over last 4 decades.

### How do non-resource revenues respond to commodity revenue shocks?
- Context and methods:
  - Broader sample of commodity exporters; commodity and non-commodity revenues scaled by non-commodity GDP to alleviate endogeneity.
  - Long-run and short-run reactions analyzed using Pooled Mean Group (PMG) estimator (Pesaran and others, 1999).
  - Specification: Y = nominal GDP (non-commodity), R = government non-commodity (NC) and commodity (C) revenues; β measures long-run reaction; δ short-term effect; ϕ is speed of adjustment; μ country fixed effect.
- Main quantitative findings:
  - A permanent increase in commodity revenues by 1 percent of non-commodity GDP tends to reduce non-commodity revenues by 0.03-0.04 percent of non-commodity GDP (long-run coefficients across specifications -0.033*** [0.007]; -0.035*** [0.008]; -0.040*** [0.007]).
  - Temporary changes in commodity revenues (up to three years lag) do not have a significant impact on non-commodity revenues.
  - Half of the deviation from the long-run association is corrected in four years (half life examples: 4.1; 4.3; 3.7 years; speed of adjustment -0.156*** [0.042]; -0.150*** [0.044]; -0.169*** [0.045]).

### Conclusions and policy implications
- Core findings:
  - Fiscal policy in resource-rich countries tends to be highly procyclical and more so than for other economies.
  - No evidence that procyclicality has declined over time in this sample.
  - Adoption of fiscal rules or resource funds does not have a significant impact on fiscal cyclicality by itself; general political institutions do help reduce procyclicality.
  - Lack of progress in institutional quality likely partly explains why fiscal procyclicality has not declined in recent years.
- Policy recommendations:
  - Establish comprehensive fiscal policy frameworks in resource-rich countries to cope with heightened uncertainty and volatility.
  - Frameworks should be based on a solid long-term anchor and explicitly incorporate commodity price uncertainty.
  - Build precautionary savings during good times to enable countercyclical responses.
  - Improve institutional frameworks, including enhancing transparency and accountability.
  - Adopt tax policies aimed at diversifying the revenue base to reduce overdependence on commodity revenues and improve ability to run countercyclical policies.
  - Promote economic diversification beyond the commodity sector.

### Data sources, sample, and descriptive statistics
- Primary data sources: IMF’s International Financial Statistics (IFS), Balance of Payments Statistics, Direction of Trade Statistics, World Economic Outlook database, fiscal rules databases; World Bank’s World Development Indicators and World Governance Indicators; Macro Data Guide Political Constraint Index Dataset (POLCON); POLITY IV; International Country Risk Guide.
- Temporal coverage: annual data from 1970 to 2013 (where available); monthly commodity prices over 1957-2015.
- Sample composition: 48 commodity-exporting countries where commodities represent a large share of exports (20 percent or more) or fiscal revenues (15 percent or more) over a large part of the period. Countries listed in source include: Algeria, Angola, Australia, Azerbaijan, Bahrain, Bolivia, Botswana, Brunei Darussalam, Cameroon, Canada, Chad, Chile, Colombia, Democratic Republic of Congo, Republic of Congo, Côte d'Ivoire, Ecuador, Gabon, Ghana, Guinea, Guyana, Indonesia, Iran, Iraq, Kazakhstan, Kuwait, Libya, Mali, Mauritania, Mexico, Mongolia, Nigeria, Norway, Oman, Papua New Guinea, Peru, Qatar, Russia, Saudi Arabia, South Africa, Sudan, Suriname, Syria, Trinidad and Tobago, United Arab Emirates, Venezuela, Yemen, and Zambia.
- Note: The sample varies for each analysis depending on data availability.

### Commodity price growth rates (selected descriptive statistics)
- Aluminium: Sample Feb 1957-Jan 2015; Obs. 696; Mean -0.133; Median -0.188; St. Dev. 4.66; Skewness -0.504; Kurtosis 8.380
- Copper: Sample Feb 1957-Jan 2015; Obs. 696; Mean -0.009; Median 0.084; St. Dev. 6.63; Skewness -0.477; Kurtosis 6.351
- Gold: Sample Feb 1957-Jan 2015; Obs. 6120; Mean 0.254; Median -0.199; St. Dev. 4.69; Skewness 1.091; Kurtosis 11.563
- Iron ore: Sample Feb 1975-Jan 2015; Obs. 4800; Mean 0.145; Median -0.295; St. Dev. 5.56; Skewness 3.812; Kurtosis 36.452
- Gas (EU): Sample Feb 1985-Jan 2015; Obs. 3600; Mean 0.033; Median -0.206; St. Dev. 6.43; Skewness -0.635; Kurtosis 18.460
- Gas (US): Sample Feb 1991-Jan 2015; Obs. 2880; Mean 0.063; Median -0.310; St. Dev. 13.34; Skewness -0.042; Kurtosis 3.950
- Tin: Sample Feb 1957-Jan 2015; Obs. 696; Mean 0.008; Median -0.107; St. Dev. 4.98; Skewness -0.474; Kurtosis 6.719
- Oil (Brent): Sample Feb 1957-Jan 2015; Obs. 696; Mean 0.146; Median -0.243; St. Dev. 8.38; Skewness 4.354; Kurtosis 65.932
- Oil (Texas): Sample Feb 1957-Jan 2015; Obs. 696; Mean 0.099; Median -0.279; St. Dev. 7.11; Skewness 2.088; Kurtosis 34.853

### Duration of commodity price expansions and contractions (selected)
- Oil-Texas
  - Expansions: Mean 29.3; Median 18.0; St. Dev. 24.4; Freq. 11; Min 1; Max 186
  - Contractions: Mean 34.1; Median 34.0; St. Dev. 14.6; Freq. 11; Min 2; Max 65
- Iron ore
  - Expansions: Mean 29.2; Median 13.0; St. Dev. 34.6; Freq. 12; Min 9; Max 99
  - Contractions: Mean 43.7; Median 47.0; St. Dev. 20.9; Freq. 11; Min 7; Max 172
- Gas-US
  - Expansions: Mean 23.8; Median 19.0; St. Dev. 15.9; Freq. 13; Min 5; Max 135
  - Contractions: Mean 20.9; Median 15.0; St. Dev. 14.5; Freq. 6; Min 4; Max 76
- Gold
  - Expansions: Mean 28.3; Median 22.5; St. Dev. 21.3; Freq. 10; Min 12; Max 278
  - Contractions: Mean 37.6; Median 23.0; St. Dev. 35.1; Freq. 11; Min 8; Max 115

### Amplitude of commodity price expansions and contractions (selected)
- Oil-Texas
  - Expansions: Mean 40.95; Median 52.4; St. Dev. 35.01; Freq. 1; Min 1.0; Max 6107.1
  - Contractions: Mean -51.6; Median -49.6; St. Dev. 42.2; Freq. 1; Min -140.0; Max -6.2
- Oil-Brent
  - Expansions: Mean 63.32; Median 28.9; St. Dev. 64.3; Freq. 8; Min 7.3; Max 166.1
  - Contractions: Mean -47.5; Median -48.5; St. Dev. 24.4; Freq. 8; Min -91.1; Max -9.9
- Gas-US
  - Expansions: Mean 80.18; Median 84.3; St. Dev. 25.1; Freq. 6; Min 0.3; Max 110.5
  - Contractions: Mean -78.9; Median -69.5; St. Dev. 49.3; Freq. 7; Min -154.9; Max -14.4
- Copper
  - Expansions: Mean 63.05; Median 57.3; St. Dev. 51.5; Freq. 11; Min 1.0; Max 175.4
  - Contractions: Mean -57.8; Median -51.2; St. Dev. 43.6; Freq. 12; Min -162.7; Max -2.9
- Gold
  - Expansions: Mean 65.56; Median 63.0; St. Dev. 50.6; Freq. 10; Min 0.4; Max 166.7
  - Contractions: Mean -51.8; Median -45.7; St. Dev. 40.1; Freq. 11; Min -138.6; Max -9.3

### Estimation results: government spending and commodity prices (selected coefficients)
- Dependent variables: real total expenditure and capital expenditure growth rates. Fixed effects with AR(1) residuals and time effects.
- Examples of reported coefficients:
  - Δ log of commodity prices (total expenditure): 0.119** [0.052]; 0.121** [0.057]
  - Δ log of commodity prices (capital): 0.147* [0.081]; 0.169** [0.086]
  - Interaction Δ log commodity prices * Expansion dummy: 0.148** [0.059]; 0.157** [0.065]; capital 0.205** [0.095]; 0.223** [0.101]
  - Share of commodity sector value added in GDP: 0.004*** [0.001]
  - Observations (examples): total expenditure regressions 902, 902, 824, 824; capital expenditure regressions 1346, 1346, 1239, 1239
  - Number of countries typically 41 or 38 depending on specification.
  - R^2 examples: 0.07, 0.08, 0.10, 0.11

### Non-commodity output gap and cyclically-adjusted non-commodity balance (selected)
- Non-commodity output gap coefficient: -0.945*** [0.225]
- Non-commodity GDP growth coefficient: -0.215*** [0.011]
- Observations: 770 and 765; Number of countries: 41 and 41; R^2: 0.14 and 0.27

### Impact of fiscal rules and institutions on procyclicality (selected)
- Interactions of Δ log commodity prices with institutional/fund dummies generally negative but mostly not statistically significant:
  - *Savings fund dummy*: -0.022 [0.071]
  - *Stabilization fund dummy*: -0.058 [0.064]
  - *Fiscal rule dummy*: -0.078 [0.088]
  - *Fiscal rule or savings/stabilization fund dummy*: -0.075 [0.063]
- Interactions with institutional quality measures (examples showing statistical significance in some cases):
  - Δ log commodity prices * Bureaucratic quality: -0.087*** [0.029]
  - Δ log commodity prices * Political risk: -0.007*** [0.002]
  - Δ log commodity prices * Institutional and legal setting: -0.003* [0.001]
- Observations and panels vary across specifications (examples: Obs. 902, 464, 805; Countries 41, 22, 41).

### Impact of commodity revenue shocks on non-commodity revenues (PMG results, selected)
- Long-run coefficients (commodity revenue/non-commodity GDP, lagged): -0.033*** [0.007]; -0.035*** [0.008]; -0.040*** [0.007]
- Speed of adjustment (ϕ): -0.156*** [0.042]; -0.150*** [0.044]; -0.169*** [0.045]
- Short-run Δ commodity revenue coefficients generally not significant: examples -0.072 [0.075]; -0.046 [0.102]; -0.085 [0.150]
- Observations: 744, 703, 676
- Half life (years): 4.1; 4.3; 3.7

### Resource funds and fiscal rules (overview of sample coding)
- Sovereign Wealth Funds (SWF) capture saving and stabilization funds; columns indicate presence (1) or absence (0) and dates where applicable.
- Selected entries preserve source values and dates (examples):
  - Algeria: 1 0 1 2000 0
  - Botswana: 1 1 1993 1972 1 1994
  - Chile: 1 1 1985 1985 1 2006
  - Ecuador: 1 1 2005 1999-2007 1 2002
  - Kazakhstan: 1 1 2000 2000 1 2000
  - Norway: 1 1 1985 1985 1 2002
  - Timor Leste: 1 1 2005 2005 1 2005
  - Trinidad and Tobago: 1 1 1999 2005 1 2007
  - Venezuela: 1 0 1 1999 2000 1

### Key methodological notes
- Phases of expansions and contractions identified using Harding and Pagan (2002) algorithm.
- Estimations use fixed effects with AR(1) residuals and time effects for panel regressions; Pooled Mean Group (PMG) estimator used for dynamics of non-commodity revenues.
- Robust standard errors reported; significance levels indicated by *, **, *** for 10, 5, and 1 percent respectively.

*Source: ANNEX 1. DATA, _wp1636 - ANNEX 1. DATA*

### Annex 1. Data ..........................................................................................................

### Annex 1. Data

### Dataset and scope
- New dataset covering 1970–2014 for non-renewable resource-rich countries (oil/gas and metals).
- Country selection criterion: share of commodity exports in total exports and share of commodity revenues in total government revenues of at least 15 percent on average for a five year period (either 2007-11 or 2009-13, depending on data availability).
- Commodity price series are real prices in USD deflated by the U.S. GDP deflator (Table 1 provides descriptive statistics).
- Commodity price index for each country: P_i = Σ_j P_ij * w_ij where j indexes commodity types (oil, gas, gold, tin, zinc, lead, aluminum, nickel, copper, and silver), P is the real commodity price (deflated by the U.S. CPI in formula exposition), and w is the commodity weight (commodity export share in GDP).

### Size and dynamics of commodity price fluctuations
- Average duration of commodity price upswings (downswings): 2-4 years.
- Standard deviation of durations is large; some expansions or contractions can last up to 10 years (Table 2).
- Average amplitude of booms (percentage change from trough to peak) and busts (percentage change from peak to trough):
  - Booms: large, ranging from 40-50 percent (example: iron ore) and 80 percent (example: natural gas).
  - Busts: range from 35-80 percent.
  - Some booms or busts can exceed 200 percent in amplitude.
- Metals, minerals, and oil sectors tend to have relatively longer boom/bust durations due to long lags between investment and supply response.

### Direct impact on fiscal revenues and exports
- Estimated average direct impact of commodity price swings (Figure 1):
  - Fiscal revenues: 8-13 percent of GDP.
  - Exports: 2-10 percent of GDP.
- Relative impact larger on fiscal revenues in oil exporters, suggesting stronger transmission through the fiscal channel.
- Asymmetry: impact is stronger in downswings compared to upswings.

### Historical macroeconomic consequences
- 1970s-80s example: many countries experienced real revenue increases of close to 10 percent a year during the boom and subsequent large falls in the bust (Figure 2).
- After the 1970s-80s bust many commodity exporters experienced long periods of negative or stagnant growth.
- Recent large revenue windfalls in the 2000s have been followed by large falls in commodity prices, forcing many countries into large fiscal contractions during weak economic activity.

### Fiscal cyclicality — conceptual approaches used
- Two empirical approaches applied:
  1. Measure 1 — responsiveness of public expenditure growth rates to year-to-year changes in commodity prices (uses Δlog(RG_it) = α_i + β Δlog(P_it) + ε_it). Positive β implies procyclicality.
  2. Measure 2 — reaction of cyclically adjusted non-resource balance to the non-resource output gap (relationship between non-resource fiscal balance and non-resource output gap).
- Rationale:
  - Commodity prices are exogenous to domestic cycles and spending policies, alleviating endogeneity issues for Measure 1.
  - Cyclically adjusted non-resource balance isolates fiscal stance excluding commodity component for Measure 2.
- Analysis uses changes (differences) to abstract from long-run level correlations; panel cointegration tests (Westerlund, 2007) found no co-integration between levels of spending and prices.

### Empirical findings on procyclicality (Measure 1 results)
- Commodity prices have a positive impact on government spending (Table 4): fiscal policy has been procyclical.
- Magnitude: A 10 percent increase (fall) in commodity prices leads to a 1.2 percent increase (fall) in real expenditure growth.
  - Example implication: If oil prices fall by 50 percent (as in the second half of 2014), expenditures would need to contract by 6.5 percent on average—this occurs while economic growth is rapidly decelerating.
- Capital spending is even more procyclical compared to total spending (reported coefficient noted as 0.15 in text fragment).
- Procyclicality is stronger during commodity price expansions than contractions.
- Procyclicality has not changed significantly over time in the sample overall.
- Some countries exhibited countercyclical behavior temporarily (e.g., in 2009), but these episodes were short-lived relative to overall patterns.

### Institutions, fiscal rules, and revenue mobilization (summary of findings described)
- Regression analysis suggests that adoption of fiscal rules alone does not significantly reduce procyclicality unless supported by strong political institutions.
- Examination of non-commodity revenues indicates:
  - Revenue mobilization efforts decline with rising commodity prices.
  - Non-commodity revenues adjust mainly in response to persistent changes in commodity revenues; the adjustment tends to be sluggish.

*Source: IMF staff estimates and analysis in Annex 1. Data.*

### 0.17 when controlling for dependence on resource revenue).

### _wp1636 - 0.17 when controlling for dependence on resource revenue).

### B. Measure 2: The cyclically-adjusted non-resource balance
- Empirical specification: CA_BAL_NC (cyclically adjusted non-resource balance, assuming elasticities of 1 for revenues and 0 for expenditures), GDP_NC (non-resource GDP), GAP_NC (non-resource GDP gap). Coefficient β captures fiscal cyclicality (negative ⇒ procyclicality). Equation estimated using fixed effects; robustness check uses non-resource GDP growth instead of non-resource output gap.
- Main findings:
  - Governments tend to loosen the fiscal stance when the domestic non-resource economy strengthens, and tighten when it weakens (Table 5).
  - A 1 percentage point improvement in the non-resource output gap leads to a 1 percentage point deterioration of the cyclically adjusted non-resource balance as a share of potential non-commodity GDP.
  - Replacing output gap with real GDP growth rates (non-resource economy) does not alter the negative association.
  - Commodity exporters tend to be more procyclical than other emerging economies.
  - Comparison: IMF (2015b) found emerging markets and developing economies act procyclically in expansions with a coefficient around 0.5, about half the size of the figure found here for commodity exporters.

### C. Can institutions help reduce procyclicality?
- Analytical approach:
  - Focus on rules designed to regulate accumulation/use of resource revenues, including rules established for functioning of a fund (saving or stabilization). Estimation includes a dummy for fiscal rules and a dummy for when a fund is in place (with or without a rule).
  - Institutional measures selected from World Governance Indicators and International Risk Group databases: bureaucratic quality, corruption, political risk, strength of institutional and legal setting. Polity variable from Polity IV dataset captures democratic institutions and rule of law.
  - Empirical specification interacts commodity price index with measures of institutional quality and fiscal rules/resource funds. Coefficient γ measures extent institutions/rules/funds affect procyclicality (negative γ ⇒ reduction in procyclicality).
- Results on rules and funds:
  - Interaction term is negative (consistent with reduction in procyclicality after adoption of fiscal rules/resource funds) but not significant (Table 6).
  - Many countries struggle to improve compliance and efficacy of special fiscal institutions; notable successes limited to Botswana, Chile, and Norway.
- Reasons for limited success:
  - Existence of a rule/fund does not imply de facto compliance; many rules breached, especially in bad times.
  - Causes include lack of political will, poor rule design, absence of monitoring/enforcement.
  - Examples: Nigeria (weak enforcement); Chad, Ecuador, Timor Leste (rules breached due to incompatibility with budget/development priorities); Mongolia, Azerbaijan, Kazakhstan, Libya (extra-budgetary operations undermining rules); Ghana, Trinidad and Tobago (accumulation in funds while governments borrowed expensively).
- Institutional quality effects:
  - Empirical support that political institution quality helps limit procyclical bias in spending (Table 7).
  - Procyclicality would be eliminated in countries with bureaucratic quality or institutional/legal setting around two standard deviations above the mean.
  - Average institutional quality tends to be weaker in resource-rich countries than in other countries (Figure 4).
  - Lack of success of rules/funds may owe more to weak institutional frameworks than to the rules themselves.
- Successful examples:
  - Norway, Chile, Botswana have higher institutional quality which helped support fiscal policy and stronger long-term growth (see Figure 2).
  - Rules/funds can help achieve objectives if supported by strong institutions, political commitment, good design, and close linkage to broader policy objectives.
  - Chile and Norway examples: rules disciplined policies while allowing flexibility; large financial buffers built during booms and strong market credibility enabled countercyclical policy.
  - Norway specifics: non-oil deficit target anchored on 4 percent of assets in oil fund over the economic cycle; oil/gas revenue saved in oil fund with only returns used for budget; rule insulates budget from yearly oil/gas price movements and supported sustained GDP per capita growth over last 4 decades.

### V. HOW DO NON-RESOURCE REVENUES RESPOND TO COMMODITY REVENUE SHOCKS?
- Context:
  - Previous studies focus on expenditures; countries can also change tax effort in response to shocks.
  - Existing studies (Bornhorst and others, 2009; Thomas and Trevino, 2013; Crivelli and Gupta, 2014) find countries tend to offset rising commodity revenue by reducing non-resource tax effort.
- Empirical expansion in this study:
  - Broader set of commodity exporters; scale commodity and non-commodity revenues by non-commodity GDP to alleviate endogeneity from commodity price changes on denominator.
  - Analyze long-run and short-run reactions using Pooled Mean Group (PMG) estimator (Pesaran and others, 1999).
  - Specification details: Y = nominal GDP (non-commodity), R = government non-commodity (NC) and commodity (C) revenues; β measures long-run reaction of non-commodity revenues to permanent change in commodity revenues; δ measures short-term effect; ϕ (denoted ߶) is speed of adjustment to long-run equilibrium; μ is country-specific fixed effect.
- Main quantitative findings:
  - A permanent increase in commodity revenues by 1 percent of non-commodity GDP tends to reduce non-commodity revenues by 0.03-0.04 percent of non-commodity GDP (Table 8).
  - Temporary changes in commodity revenues (up to three years lag) do not have a significant impact on non-commodity revenues.
  - Half of the deviation from the long-run association between commodity and non-commodity revenues is corrected in four years, indicating sluggish adjustment.

### VI. CONCLUSIONS
- Core findings:
  - Fiscal policy in resource-rich countries tends to be highly procyclical and more so than for other economies.
  - No evidence found that procyclicality has declined over time (contrary to other studies).
  - Adoption of fiscal rules or resource funds does not have a significant impact on fiscal cyclicality; however, general political institutions do help reduce procyclicality.
  - Lack of progress in institutional quality likely partly explains why fiscal procyclicality has not declined in recent years.
- Policy implications and recommendations:
  - Establish comprehensive fiscal policy frameworks in resource-rich countries to cope with heightened uncertainty and volatility.
  - Frameworks should be based on a solid long-term anchor and explicitly incorporate commodity price uncertainty.
  - Emphasize building precautionary savings during good times to weather shocks in a countercyclical fashion.
  - Further efforts to improve institutional framework, including enhancing transparency and accountability.
  - Adopt tax policies aimed at diversifying the revenue base to reduce overdependence on commodity revenues and improve ability to run countercyclical policies.
  - Promote economic diversification beyond the commodity sector.

*Source: _wp1636 - 0.17 when controlling for dependence on resource revenue).*/*

### ANNEX 1. DATA

### ANNEX 1. DATA

### Data sources and sample
- Primary data sources: IMF’s International Financial Statistics (IFS), Balance of Payments Statistics, Direction of Trade Statistics, World Economic Outlook database, and fiscal rules databases; World Bank’s World Development Indicators and World Governance Indicators; Macro Data Guide Political Constraint Index Dataset (POLCON); POLITY IV; International Country Risk Guide.
- Temporal coverage: annual data from 1970 to 2013 (where available); monthly commodity prices over 1957-2015.
- Sample composition: 48 commodity-exporting countries where commodities represent a large share of exports (20 percent or more) or fiscal revenues (15 percent or more) over a large part of the period. Countries: Algeria, Angola, Australia, Azerbaijan, Bahrain, Bolivia, Botswana, Brunei Darussalam, Cameroon, Canada, Chad, Chile, Colombia, Democratic Republic of Congo, Republic of Congo, Côte d'Ivoire, Ecuador, Gabon, Ghana, Guinea, Guyana, Indonesia, Iran, Iraq, Kazakhstan, Kuwait, Libya, Mali, Mauritania, Mexico, Mongolia, Nigeria, Norway, Oman, Papua New Guinea, Peru, Qatar, Russia, Saudi Arabia, South Africa, Sudan, Suriname, Syria, Trinidad and Tobago, United Arab Emirates, Venezuela, Yemen, and Zambia.
- Note: The sample varies for each analysis depending on data availability.

### Commodity price growth rates (descriptive statistics)
Source: IMF World Economic Outlook. Reported are descriptive statistics for real m-o-m growth rates. Sample periods indicated per commodity.

- Aluminium: Sample Feb 1957-Jan 2015; Obs. 696; Mean -0.133; Median -0.188; St. Dev. 4.66; Skewness -0.504; Kurtosis 8.380
- Copper: Sample Feb 1957-Jan 2015; Obs. 696; Mean -0.009; Median 0.084; St. Dev. 6.63; Skewness -0.477; Kurtosis 6.351
- Gold: Sample Feb 1957-Jan 2015; Obs. 6120; Mean 0.254; Median -0.199; St. Dev. 4.69; Skewness 1.091; Kurtosis 11.563
- Iron ore: Sample Feb 1975-Jan 2015; Obs. 4800; Mean 0.145; Median -0.295; St. Dev. 5.56; Skewness 3.812; Kurtosis 36.452
- Gas (EU): Sample Feb 1985-Jan 2015; Obs. 3600; Mean 0.033; Median -0.206; St. Dev. 6.43; Skewness -0.635; Kurtosis 18.460
- Gas (US): Sample Feb 1991-Jan 2015; Obs. 2880; Mean 0.063; Median -0.310; St. Dev. 13.34; Skewness -0.042; Kurtosis 3.950
- Tin: Sample Feb 1957-Jan 2015; Obs. 696; Mean 0.008; Median -0.107; St. Dev. 4.98; Skewness -0.474; Kurtosis 6.719
- Oil (Brent): Sample Feb 1957-Jan 2015; Obs. 696; Mean 0.146; Median -0.243; St. Dev. 8.38; Skewness 4.354; Kurtosis 65.932
- Oil (Texas): Sample Feb 1957-Jan 2015; Obs. 696; Mean 0.099; Median -0.279; St. Dev. 7.11; Skewness 2.088; Kurtosis 34.853

### Duration of commodity price expansions and contractions
Source: IMF World Economic Outlook and authors’ calculations. Phases defined using Harding and Pagan (2002) algorithm. Sample period: January 1957-March 2015.

- Oil-Texas
  - Expansions: Mean 29.3; Median 18.0; St. Dev. 24.4; Freq. 11; Min 1; Max 186
  - Contractions: Mean 34.1; Median 34.0; St. Dev. 14.6; Freq. 11; Min 2; Max 65
- Tin
  - Expansions: Mean 38.5; Median 24.5; St. Dev. 33.0; Freq. 8; Min 1; Max 112
  - Contractions: Mean 43.2; Median 44.0; St. Dev. 26.9; Freq. 12; Min 8; Max 1
- Oil-Brent
  - Expansions: Mean 36.4; Median 26.0; St. Dev. 33.0; Freq. 8; Min 1; Max 125
  - Contractions: Mean 40.3; Median 35.0; St. Dev. 22.1; Freq. 8; Min 17; Max 79
- Iron ore
  - Expansions: Mean 29.2; Median 13.0; St. Dev. 34.6; Freq. 12; Min 9; Max 99
  - Contractions: Mean 43.7; Median 47.0; St. Dev. 20.9; Freq. 11; Min 7; Max 172
- Gas-EU
  - Expansions: Mean 30.8; Median 24.5; St. Dev. 23.9; Freq. 6; Min 7; Max 76
  - Contractions: Mean 29.3; Median 27.0; St. Dev. 13.6; Freq. 15; Min 5; Max 53
- Gas-US
  - Expansions: Mean 23.8; Median 19.0; St. Dev. 15.9; Freq. 13; Min 5; Max 135
  - Contractions: Mean 20.9; Median 15.0; St. Dev. 14.5; Freq. 6; Min 4; Max 76
- Aluminium
  - Expansions: Mean 25.1; Median 23.0; St. Dev. 8.8; Freq. 11; Min 2; Max 238
  - Contractions: Mean 35.1; Median 19.0; St. Dev. 46.5; Freq. 12; Min 14; Max 179
- Copper
  - Expansions: Mean 28.3; Median 21.0; St. Dev. 20.5; Freq. 11; Min 1; Max 279
  - Contractions: Mean 32.2; Median 23.5; St. Dev. 30.3; Freq. 12; Min 8; Max 107
- Gold
  - Expansions: Mean 28.3; Median 22.5; St. Dev. 21.3; Freq. 10; Min 12; Max 278
  - Contractions: Mean 37.6; Median 23.0; St. Dev. 35.1; Freq. 11; Min 8; Max 115

(Note: table layout in source groups rows differently; numbers preserved exactly as presented.)

### Amplitude of commodity price expansions and contractions
Source: IMF World Economic Outlook and authors’ calculations. Phases defined using Harding and Pagan (2002) algorithm. Sample period: January 1957-March 2015.

- Oil-Texas
  - Expansions: Mean 40.95; Median 52.4; St. Dev. 35.01; Freq. 1; Min 1.0; Max 6107.1
  - Contractions: Mean -51.6; Median -49.6; St. Dev. 42.2; Freq. 1; Min -140.0; Max -6.2
- Tin
  - Expansions: Mean 78.67; Median 70.2; St. Dev. 45.6; Freq. 8; Min 25.2; Max 172.0
  - Contractions: Mean -72.4; Median -85.4; St. Dev. 34.5; Freq. 9; Min -113.1; Max -17.1
- Oil-Brent
  - Expansions: Mean 63.32; Median 28.9; St. Dev. 64.3; Freq. 8; Min 7.3; Max 166.1
  - Contractions: Mean -47.5; Median -48.5; St. Dev. 24.4; Freq. 8; Min -91.1; Max -9.9
- Iron ore
  - Expansions: Mean 50.4; Median 14.4; St. Dev. 96.0; Freq. 6; Min 0.9; Max 245.6
  - Contractions: Mean -35.9; Median -27.3; St. Dev. 35.5; Freq. 7; Min -107.6; Max -5.9
- Gas-EU
  - Expansions: Mean 63.65; Median 54.2; St. Dev. 58.7; Freq. 6; Min 3.0; Max 168.8
  - Contractions: Mean -63.4; Median -58.7; St. Dev. 30.5; Freq. 6; Min -113.0; Max -21.5
- Gas-US
  - Expansions: Mean 80.18; Median 84.3; St. Dev. 25.1; Freq. 6; Min 0.3; Max 110.5
  - Contractions: Mean -78.9; Median -69.5; St. Dev. 49.3; Freq. 7; Min -154.9; Max -14.4
- Aluminium
  - Expansions: Mean 50.64; Median 49.3; St. Dev. 34.2; Freq. 11; Min 0.9; Max 117.1
  - Contractions: Mean -47.8; Median -44.6; St. Dev. 48.2; Freq. 12; Min -192.2; Max -12.3
- Copper
  - Expansions: Mean 63.05; Median 57.3; St. Dev. 51.5; Freq. 11; Min 1.0; Max 175.4
  - Contractions: Mean -57.8; Median -51.2; St. Dev. 43.6; Freq. 12; Min -162.7; Max -2.9
- Gold
  - Expansions: Mean 65.56; Median 63.0; St. Dev. 50.6; Freq. 10; Min 0.4; Max 166.7
  - Contractions: Mean -51.8; Median -45.7; St. Dev. 40.1; Freq. 11; Min -138.6; Max -9.3

### Estimation results: Government spending and commodity prices (summary)
Note: Dependent variables are real total expenditure (Columns I-IV), and capital expenditure (Columns V-VIII) growth rates. Estimations use fixed effects with AR(1) residuals and time effects. Robust standard errors in brackets. *, **, *** denote significance at 10, 5, and 1 percent.

- Across specifications (Columns I-IV for total expenditure; V-VIII for capital expenditure):
  - Δ log of commodity prices coefficients (examples shown):
    - Column I (total expenditure): 0.119** [0.052]
    - Column II: 0.121** [0.057]
    - Column V (capital): 0.147* [0.081]
    - Column VIII: 0.169** [0.086]
  - Interaction Δ log of commodity prices * Dummy (=1 for commodity price expansions):
    - Column I: 0.148** [0.059]
    - Column II: 0.157** [0.065]
    - Column V: 0.205** [0.095]
    - Column VIII: 0.223** [0.101]
  - Interaction Δ log of commodity prices * Dummy (=1 for commodity price contractions):
    - Coefficients reported include 0.073, 0.070, 0.074, 0.102 with associated standard errors [0.068], [0.072], [0.103], [0.108] (not statistically significant at conventional levels).
  - Share of commodity sector value added in GDP: 0.004*** (multiple columns) [0.001]
  - Constants reported vary by column; example constants: 0.124** [0.058], 0.038 [0.045], -1.193 [1.281], etc.
  - Observations and panels:
    - Observations for total expenditure regressions: 902, 902, 824, 824
    - Observations for capital expenditure regressions: 1346, 1346, 1239, 1239
    - Number of countries typically 41 or 38 depending on specification.
  - R^2 examples: 0.07, 0.08, 0.10, 0.11 (by column)

Interpretation (from reported coefficients):
- Positive and statistically significant coefficients on Δ log of commodity prices and its interaction with expansion dummy indicate that government spending (both total and capital) tends to increase when commodity prices rise, especially during expansion phases.
- The share of the commodity sector in GDP is positively associated with higher expenditure growth.

### Non-commodity output gap and cyclically-adjusted non-commodity balance
Note: Dependent variable is cyclically-adjusted non-commodity balance. Fixed effects with time effects. Robust standard errors in brackets. *, **, *** denote significance at 10, 5, and 1 percent.

- Selected coefficients (examples):
  - Non-commodity output gap: -0.945*** [0.225]
  - Non-commodity GDP growth: -0.215*** [0.011]
  - Constants in two reported columns: -0.266*** [0.039], -0.254*** [0.039]
- Observations: 770 and 765
- Number of countries: 41 and 41
- R^2: 0.14 and 0.27

Interpretation:
- Negative and significant coefficients indicate that worse non-commodity output gaps and lower non-commodity GDP growth are associated with deteriorations in the cyclically-adjusted non-commodity balance.

### Impact of fiscal rules on fiscal procyclicality
Note: Dependent variable is real expenditure growth rate. Fixed effects with AR(1) residuals and time effects. Robust standard errors in brackets. *, **, *** denote significance at 10, 5, and 1 percent.

- Δ log of commodity prices:
  - Column I: 0.119** [0.052]
  - Column II: 0.143** [0.066]
  - Column III: 0.164** [0.069]
  - Column IV: 0.151** [0.063]
  - Column V: 0.176** [0.070]
- Interactions with institutional/fund dummies:
  - Δ log of commodity prices * Savings fund dummy: -0.022 [0.071]
  - Δ log of commodity prices * Stabilization fund dummy: -0.058 [0.064]
  - Δ log of commodity prices * Fiscal rule dummy: -0.078 [0.088]
  - Δ log of commodity prices * Fiscal rule or savings/stabilization fund dummy: -0.075 [0.063]
- Observations vary by specification: 902, 718, 718, 718, 718
- Number of countries: 41, 34, 34, 34, 34
- R^2 examples: 0.079, 0.083, 0.084, 0.084, 0.085

Interpretation:
- Baseline positive association between commodity price changes and expenditure growth persists.
- Interactions with savings/stabilization funds and fiscal rule dummies are negative but generally not statistically significant at conventional levels in the reported specifications.

### Impact of institutions on fiscal procyclicality
Note: Dependent variable is real expenditure growth rate. Fixed effects with AR(1) residuals and time effects. Robust standard errors in brackets. *, **, *** denote significance at 10, 5, and 1 percent.

- Δ log of commodity prices (selected columns): 0.119** [0.052]; 0.142* [0.076]; 0.341*** [0.090]; 0.214** [0.095]; 0.609*** [0.178]; 0.266** [0.115]
- Interactions with institutional measures:
  - Δ log of commodity prices * Polity: -0.008 [0.006]
  - Δ log of commodity prices * Bureaucratic quality: -0.087*** [0.029]
  - Δ log of commodity prices * Corruption: -0.027 [0.024]
  - Δ log of commodity prices * Political risk: -0.007*** [0.002]
  - Δ log of commodity prices * Institutional and legal setting: -0.003* [0.001]
- Observations and panel counts vary across columns (examples):
  - Observations: 902, 464, 805, 805, 804, 716
  - Number of countries: 41, 22, 41, 41, 33, 32
- R^2 examples: 0.079, 0.101, 0.089, 0.079, 0.088, 0.092

Interpretation:
- Higher bureaucratic quality, lower political risk, and stronger institutional/legal settings are associated with a reduction in the procyclical response of expenditure to commodity price changes (coefficients negative and, in cases, statistically significant).

### Impact of commodity revenue shocks on non-commodity revenues
Note: Dependent variable is the change in non-commodity revenue ratio. Estimations performed using the Pooled Mean Group (PMG) estimator. *, **, *** denote significance at 10, 5, and 1 percent.

- Long-run coefficients:
  - Commodity revenue/non-commodity GDP (lagged): -0.033*** [0.007]; -0.035*** [0.008]; -0.040*** [0.007] (across three specifications)
  - Constant: 19.576*** [0.412]; 19.829*** [0.463]; 20.006*** [0.380]
- Short-run coefficients:
  - Speed of adjustment: -0.156*** [0.042]; -0.150*** [0.044]; -0.169*** [0.045]
  - Δ Commodity revenue/non-commodity GDP: -0.072 [0.075]; -0.046 [0.102]; -0.085 [0.150]
  - Δ Commodity revenue/non-commodity GDP (1 lag): -0.057 [0.132]; -0.071 [0.207]
  - Δ Commodity revenue/non-commodity GDP (2 lags): -0.244 [0.175]
- Observations: 744, 703, 676
- Log likelihoods: -1545.3; -1432.0; -1321.0
- Half life (years): 4.1; 4.3; 3.7

Interpretation:
- Long-run negative and significant relationship: higher lagged commodity revenue relative to non-commodity GDP is associated with lower non-commodity revenue ratios in the long run.
- Speed of adjustment is negative and significant, implying convergence dynamics with half-lives around 3.7–4.3 years.

### Resource funds and fiscal rules (overview)
Note: Sovereign Wealth Funds (SWF) capture only saving and stabilization funds. Columns report Yes=1; No=0 and Dates when applicable.

- Selected country entries (Saving Fund; Stabilization Fund; Fiscal Rule):
  - Algeria: 1 0 1 2000 0
  - Angola: 1 1 2012 2012 0
  - Azerbaijan: 1 1 1999 1999 1 1999
  - Bahrain: 1 0 1 2000 0
  - Bolivia: 0 0 0 0 0
  - Botswana: 1 1 1993 1972 1 1994
  - Brunei Darussalam: 0 0 0 0 0
  - Cameroon: 0 0 0 0 0
  - Chad: 1 1 2008 2008 0
  - Chile: 1 1 1985 1985 1 2006
  - Colombia: 1 0 1995 0
  - Congo: 0 0 0 0 0
  - Congo DRC: 0 0 0 0 0
  - Cote D'Ivoire: 0 0 0 0 0
  - Equatorial Guinea: 1 1 2002 2001 2002
  - Ecuador: 1 1 2005 1999-2007 1 2002
  - Gabon: 1 1 1998 1998 0
  - Ghana: 1 1 2011 2012 2011
  - Guinea: 0 0 0 0 0
  - Guyana: 0 0 0 0 0
  - Indonesia: 0 0 0 0 0
  - Iran: 1 0 1 2000 0
  - Iraq: 0 0 0 0 0
  - Kazakhstan: 1 1 2000 2000 1 2000
  - Kuwait: 1 1 1960 1960 0
  - Libya: 1 1 1995 0
  - Mali: 0 0 0 0 0
  - Mauritania: 1 0 1 2000 0
  - Mexico: 1 0 1 2000 0
  - Mongolia: 1 0 1 2011 0
  - Mozambique: 0 0 0 0 0
  - Niger: 0 0 0 0 0
  - Nigeria: 1 1 2011 2004 0
  - Norway: 1 1 1985 1985 1 2002
  - Oman: 1 1 1980 0
  - Papua New Guinea: 1 0 1 1974-2001 0
  - Peru: 1 0 1 1999 0
  - Qatar: 1 0 1 2000 0
  - Russia: 1 0 1 2004 2008 1
  - Saudi Arabia: 0 0 0 0 0
  - Sudan: 1 0 1 2002 0
  - Suriname: 0 0 0 0 0
  - Syria: 0 0 0 0 0
  - Timor Leste: 1 1 2005 2005 1 2005
  - Trinidad and Tobago: 1 1 1999 2005 1 2007
  - UAE: 0 0 0 0 0
  - Venezuela: 1 0 1 1999 2000 1
  - Yemen: 0 0 0 0 0
  - Zambia: 0 0 0 0 0

(Columns in source indicate Saving Fund, Stabilization Fund, Fiscal Rule and dates; entries above preserve values and dates as presented.)

### Key methodological notes
- Phases of expansions and contractions identified using Harding and Pagan (2002) algorithm.
- Estimations use fixed effects with AR(1) residuals and time effects for panel regressions; Pooled Mean Group (PMG) estimator used for dynamics of non-commodity revenues.
- Robust standard errors reported; significance levels indicated by *, **, *** for 10, 5, and 1 percent respectively.

*Source: ANNEX 1. DATA, _wp1636 - ANNEX 1. DATA*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1636.pdf_
