## ppea2024054 — EXECUTIVE SUMMARY (EBA-Lite 3.0)

## Source details

**Canonical URL:** [ppea2024054 — EXECUTIVE SUMMARY (EBA-Lite 3.0)](https://www.imf.org/-/media/files/publications/pp/2024/english/ppea2024054.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/pp/2024/english/ppea2024054.pdf.md)
- [Structured JSON version](/-/media/files/publications/pp/2024/english/ppea2024054.pdf.json)

---

### Role, context, and operational use
- EBA-Lite methodology is a key input in external sector assessments for over three-quarters of IMF member countries, representing a broad set of emerging and low-income economies.
- EBA introduced in 2012 for systemic economies; EBA-Lite introduced in 2015 to extend assessments to the full membership in line with the 2014 Triennial Surveillance Review.
- Reviews conducted in 2018 and 2022; this paper describes the current model referred to as EBA-Lite 3.0.
- Operational role in Article IV surveillance:
  - Assess the position and trajectory of foreign assets and liabilities, current accounts, real exchange rates, capital flows and policy measures, and foreign exchange intervention and reserve levels.
  - Bottom-line assessment must be consistent with numerical inputs from EBA and EBA-Lite models, analytically grounded judgment, and country-specific insights.
  - The CA model is typically preferred; staff judgment required to select model and adjust results where country-specific factors are not captured.

### Over-arching objectives and main refinements in EBA-Lite 3.0
- Objectives:
  - Further enhance robustness of econometric model estimates.
  - Maintain conceptual consistency with EBA models.
  - Ensure appropriate adaptations for diverse characteristics of emerging and developing economies.
- Main refinements and updates:
  - Data updates and refinements to construction of some explanatory variables.
  - Improvements in model robustness, including use of Bayesian Model Averaging (BMA) to exclude non-robust variables.
  - Operational guidance updates for External Sustainability (ES) model and commodity modules.
  - Flexibility for staff to apply most relevant approach depending on country circumstances.

### Key characteristics of EBA-Lite economies shaping model design
- Ninety percent of EBA-Lite economies are either low income or emerging market economies.
- Nearly 30 percent experienced annual damages from natural disasters that, on average, accounted for more than 0.5 percent of GDP since 1995.
- Higher incidence of armed conflicts relative to EBA economies.
- Remittances are an important external financing source for many EBA-Lite economies; remittances are included in EBA-Lite models (not in EBA).
- Larger share are less financially developed; EBA-Lite considers both credit cycle and level of financial development.
- More than 20 percent of sample economies are fuel and non-fuel commodity exporters, motivating commodity modules.
- Many have high levels of external indebtedness, including foreign-currency denominated debt; ES approach accounts for this.
- Greater heterogeneity: wider range and larger variance of current account balances and larger country-specific CA balance heterogeneity over time compared with EBA economies.

### CA model specification, estimation, and data
- Model form (panel regression): CAit = α + β'Fit + λ'Pit + γ'Cit + eit
  - CAit is current account balance in percent of GDP in country i in year t.
  - Cit: cyclical factors and shocks.
  - Pit: policy variables.
  - Fit: macroeconomic and structural fundamentals.
  - eit: zero mean, normally distributed regression residual, assumed to follow an AR(1) process.
- Most variables measured relative to the world average (GDP-weighted average of all countries in the EBA-Lite estimation sample).
- Country fixed effects are not included.
- Estimation method: pooled Generalized Least Squares (GLS) with a panel-wide AR(1) correction.
- Estimation sample includes all countries where data are available, including those in the EBA estimation sample.
- Sample period extended from 1995–2016 to 1995–2019.
- Model estimated with data available as of September 2021, except NFA from October 2021 EWN revision.

### Explanatory variable groups (examples)
- Cyclical factors and shocks:
  - Output gap; commodity terms of trade; dummies for natural disasters and conflicts.
- Policy variables:
  - Fiscal balance; change in reserves/GDP (interacted with capital controls); public health spending; Credit/GDP; Credit/GDP growth.
- Macroeconomic and structural fundamentals:
  - Net foreign assets (NFA); output per worker; share in FX world reserves; migrant share; oil and gas trade balance; demographics; institutional quality.

### Key variable refinements and BMA model selection
- Commodity terms-of-trade gap:
  - New approach identifies “supercycles” in each commodity price series, builds gaps per commodity, aggregates using country-specific trade weights; provides transparent decomposition of contributions.
  - Although estimated coefficient on terms-of-trade is smaller in EBA-Lite 3.0, a one standard deviation change in the terms-of-trade series yields at least as large a cyclical adjustment in the model.
- Oil and gas trade balance proxied by a 3-year moving average of oil and gas net export (relative to GDP); net exports assessed as if prices at long-term trend level.
- Capital controls: EBA-Lite 3.0 uses the Financial Account Restrictiveness Index (FARI) replacing the Chinn-Ito index; correlation coefficient between Chinn-Ito and FARI was 0.83 in 2019.
- BMA approach:
  - Analyzes all possible variable combinations to calculate posterior inclusion probability (PIP).
  - Over 33 million rounds of estimations performed given EBA-Lite 2.0 variables.
  - Variable considered robust if PIP > 50 percent.
  - Variables removed by BMA: life expectancy at age 45 (relative to world average); lagged NFA/GDP interacted with dummy for NFA/GDP >-60 percent; lagged demeaned VIX*capital openness; lagged demeaned VIX*capital openness*share in world reserves.
  - Removing these had little impact on overall fit and other coefficient estimates.

### Estimation results and selected coefficients
- Model stability and fit:
  - Most coefficient estimates stable relative to prior specifications.
  - Model fit marginally improved compared to EBA-Lite 2.0.
  - R-squared marginally higher than EBA model but RMSE higher, reflecting higher prevalence of outliers.
- Cyclical factors and shocks:
  - Output gap: a 1 percent increase in the relative output gap → reduction in CA by about 0.13 percentage points of GDP.
  - Commodity terms-of-trade (interacted with trade openness): for openness degree of 50 percent, a 1 percent improvement → increase in CA by about 0.1 percentage points of GDP.
  - Natural disaster dummy: impact depends on capital account openness; example FARI indices: Solomon Islands FARI = 0.78 in 2020; Papua New Guinea FARI = 0.14 in 2020.
  - Conflicts dummy: estimated marginal impact of an armed conflict is to raise the current account by 0.7 percent of GDP.
- Macroeconomic fundamentals:
  - Lagged NFA coefficient = 0.026 (positive), similar to EBA-Lite 2.0 but lower than 2022 EBA model.
  - Domestic currency share in world FX reserves: for every 10 percent of global reserves held in domestic currency, current account deficit is lower by about 0.7 percent.
  - Output per worker (lagged): a 1 percent higher relative output per worker → increase of about 0.087 in CA balance; interacted with capital account openness (impact increases for more open economies).
  - Expected real GDP growth (5 years ahead): a 1 percent relatively higher expected real GDP growth → lower CA by about 0.64 percentage points of GDP.
- Structural fundamentals:
  - Migrant share: a 9 percentage point (one standard deviation) increase in outward migration lowers CA by about 1 percent of GDP.
  - Institutional quality (ICRG): higher institutional quality tends to lower CA.
  - Demographics: included variables have signs aligning with priors (negative on population growth and old-age dependency; positive for prime-age savers).
- Policy variables and instruments:
  - Fiscal balance (instrumented cyclically adjusted overall general government fiscal balance): estimated coefficient = 0.43 — a 1 percent increase (relative to rest of world) increases CA; EBA coefficient = 0.4.
  - Health spending (lagged): a 1 percent of GDP higher public health expenditure → lower CA by 0.57 percentage points of GDP.
  - FXI (change in reserves) interacted with capital controls (instrumented): FX purchases tend to improve the CA.
  - Credit variables: greater Credit/GDP and rapid Credit/GDP growth tend to lower CA.

### CA norms, gaps, decomposition, and multilateral consistency
- Cyclically adjusted current account: ˆAdjitCA = CAit − C'itβ̂
- Current account norm: ˆˆNormit = F'itα̂ + P'*itλ̂γ̂
- Policy gap = (P*it − Pit); contribution to CA gap = (ˆ*(it)it PP γ − ) [as in source].
- Current account gap: gapitCA = AdjitCA − Normit = (− (P*it − Pit)γ̂) + eit; equivalently gapitCA equals sum of model-identified policy gaps and regression residual.
- Residual interpreted as structural factors not captured by the model; persistent large residuals may justify a norm adjustor.
- Multilateral consistency: estimated CA gaps should sum to zero; in practice gaps do not sum to zero due to statistical discrepancies and some variables not expressed as deviations from world average; gaps receive a small adjustment attributed to CA norm to ensure multilateral consistency.

### Setting desirable policy levels (P*) — operational guidance (summary)
- Fiscal policy P*: cyclically adjusted general government fiscal balance recommended by country teams over the medium term (five years); may be based on debt-stabilizing primary balance, fiscal rules, long-term adjustment needs, or higher spending needs.
- Public health expenditure P*: benchmark regressions guide P*; regressors include GDP per capita (PPP), old-age dependency ratio, income inequality, and fiscal revenue; P* is a medium-term objective not expected to change year-to-year.
- Private credit P*: desirable Credit/GDP consistent with financial development role; benchmark regression links credit-to-GDP to financial development, GDP per capita, public debt, inflation, etc.; P* for change in private credit set to achieve desirable credit-to-GDP level.
- FXI P*: proxied by change in reserves/GDP; desirable medium-term level is zero change in reserves/GDP; in exceptional cases P* could be annualized change needed to close gap to adequate reserves.
- Capital controls P*: desirable level is either country’s current level or contemporaneous cross-country average of controls index, whichever is smaller (least restrictive).
- Real short-term interest rates (REER equation only): desirable short-term interest rate equals monetary policy stance appropriate for domestic output and inflation objectives; if current policy appropriate, P* = actual.

### Adjustors — temporary and structural
- Adjustors applied transparently and conservatively; internal review ensures theory-grounded, appropriately computed, evenhanded, and multilaterally consistent.
- Types:
  - Cyclical adjustors for temporary shocks not fully captured by model (e.g., one-off import-intensive projects, prolonged yet transitory natural disasters, one-off remittance shocks); typically calibrated to temporary portion and used for one or two years.
  - Mandatory adjustors for broad-based developments not covered by model (example: Covid-19 adjustors for transitory net impacts on tourism, commodity volumes, medical goods trade, and shift in consumer preferences).
  - Norm adjustors for country-specific structural factors not included in model (e.g., large migrant share with very low remittances); once introduced, should be maintained for subsequent assessments.

### REER gap and elasticities (conversion)
- CA gap converted to REER gap using semi-elasticity η_TB:
  - η_TB = ∆(TB/GDP) / ∆REER/REER
  - REER_gap = gapitCA / η_TB
  - η_TB = η_X s_X − η_M s_M, where η_X and η_M are elasticities of exports and imports wrt REER; s_X and s_M are nominal shares of exports and imports to GDP.
- Country-specific semi-elasticities use panel-estimated η_X and η_M and country-specific export/import shares; eleven-year moving average smooths shares.
- Note: 2022 EBA refinements introduced income balance semi-elasticities that were generally close to zero and were not added to EBA-Lite framework; staff may use alternative well-justified estimates if appropriate.

### REER model (EBA-Lite 3.0) — estimation, results, and caveats
- Export/import elasticities (unbalanced panel 1980–2019, nominal USD):
  - Import elasticity (long-run): 0.25
  - Export elasticity (long-run): -0.46
  - EBA-Lite 2.0: import = 0.29; export = -0.44
- REER model features:
  - Focuses on country-specific determinants of REER movements with country fixed effects (residuals average to zero for each country).
  - Sample extended 1995–2019; uses refined commodity terms-of-trade and FARI.
  - Insignificant variables dropped (also from CA model): global risk aversion variables; output per worker interacted with capital openness; oil and gas reserves; conflict dummy; aid.
  - Adds relative output gap among cyclical factors; adds cyclically adjusted fiscal balance as policy variable.
- Key REER coefficients and interpretations:
  - Output gap: a 1 percent increase in relative output gap → REER appreciation of about 0.4 percent.
  - Commodity terms-of-trade (interacted with trade openness): more favorable terms → appreciated REER.
  - Natural disasters (interacted with capital openness): tend to appreciate REER but impact mitigated for more open capital accounts.
  - Lagged NFA: ambiguous; interaction with high-debt dummy (NFA below -60 percent) positive and significant.
  - Own currency share in world FX reserves: positive → more appreciated REER.
  - Remittances: increase tends to depreciate REER (contrast with CA model treatment).
  - Fiscal policy: a 1 percentage point increase in cyclically adjusted fiscal balance relative to GDP → REER appreciation of 0.2 percent.
  - FXI interacted with capital controls: FX purchases tend to depreciate REER; effectiveness depends on capital account openness.
  - Monetary policy interacted with capital controls: a 1 percentage point increase in relative real short-term interest rate → 1 percent appreciation in countries with open capital accounts.
  - Credit variables: greater financial depth and strong credit growth tend to appreciate REER.
- Caveats and guidance:
  - REER model assesses REER relative to own history (fixed effects), less robust to structural breaks, generally poorer fit than CA models.
  - Preference usually given to CA models in external sector assessments; REER model is complementary and secondary.

### External Sustainability (ES) approach and commodity modules
- External vulnerabilities:
  - Nearly a quarter of EBA-Lite users have NFAs (or NIIP) below -60 percent of GDP.
  - High net external borrowing increases exposure to sudden global financial condition changes, roll-over risks, and borrowing cost increases.
- ES approach overview:
  - Compares medium-term expected CA/GDP to CA/GDP that would stabilize NFA/GDP at a specified benchmark.
  - More forward-looking; CA gaps not attributed to specific policies.
  - EBA-Lite 2.0 expanded Blanchard and Das (2017) approach to include trade and financial factors with deterministic and probabilistic modules.
  - Accounts for valuation effects linked to foreign-currency denomination of assets and liabilities; generally does not consider return differentials except for a few countries.
  - Uses medium-term CA projection and stabilizes debt at its level today; recommended REER adjustments may be necessary but sometimes insufficient for very large negative NFA.
- Commodity exporter modules:
  - Consumption allocation model: derives CA norms by distributing resource wealth across periods to smooth inter-temporal consumption; annuity = present value of below-the-ground wealth (exports of exhaustible commodities) + above-ground external wealth (NIIP); often yields higher CA norms than CA regression; most useful in middle- and high-income countries.
  - Investment module: incorporates investment needs and efficiency; better suited for countries with large investment needs, lack of market access, and reasonably high investment efficiency; calibration sensitive to investment efficiency parameters.
  - In EBA-Lite 3.0 most parameters fixed at empirically supported values; only private and public investment efficiency parameters free and guided by PIMA or empirical studies.
  - Evidence (Araujo and others (2016)): introducing investment frictions can lead to CA norms over 5 percentage points higher than without frictions.
  - Distinction between investment inefficiencies (one dollar invested yields < one dollar of productive capital) and absorptive capacity constraints (costly adjustment of capital stocks).
- Use and interpretation:
  - Commodity-tailored models complement standard approaches; regression-based CA models fit relatively poorly for commodity exporters and residuals increase with commodity exports.
  - Staff should select preferred approach based on country characteristics and keep model choice stable year-to-year; numerical results from different models should not be averaged.

### Consumption allocation and investment allocation — practical guidance
- Consumption allocation model:
  - Permanent income hypothesis applied to derive constant real per capita annuity as normative consumption; yields saving norm applied to CA.
  - Tractable and often useful in middle- and high-income countries; less appropriate in some LICs where immediate investment needs dominate.
- Investment module:
  - Can produce lower CA norms than consumption allocation when domestic returns to investment are high.
  - Calibration sensitive; parameters for investment efficiency should be transparently justified; avoid frequent small changes.
  - Implementation challenges include sensitivity to parameterization and dependence on institutional capacity (PIMA assessments).

### Methodological guidance and ongoing work
- Flexibility: EBA-Lite toolkit allows staff to apply the most relevant approach for country circumstances; model results can be adjusted transparently to reflect country-specific factors.
- Preference for CA model as main workhorse for most EBA-Lite economies due to better empirical fit and interpretability.
- Where external sustainability is overriding concern, use ES approach or apply ES adjustor to CA models.
- Commodity modules provide additional perspectives for commodity exporters, noting sensitivity of investment module to parameter choices.
- Ongoing work includes improving empirical robustness, exploring new variables such as climate change impacts on CA balances, and other adaptations to improve country-specific tailoring while maintaining multilateral consistency.

*Prepared by the External Policy Division in the Strategy Policy and Review Department (SPR). Source: ppea2024054.*

### EXECUTIVE SUMMARY

### ppea2024054 - EXECUTIVE SUMMARY

### Role and context
- The EBA-Lite methodology is a key input in external sector assessments for over three-quarters of IMF member countries, representing a broad set of emerging and low-income economies.
- The External Balance Assessment (EBA) Methodology was introduced in 2012 for systemic economies; EBA-Lite was introduced in 2015 to extend assessments to the full membership beyond more systemic economies, in line with the recommendations of the 2014 Triennial Surveillance Review.
- Staff conducted reviews of the EBA-Lite methodology in 2018 and 2022; this paper provides a detailed description of the current model (referred to as EBA-Lite 3.0).

### Over-arching objectives of EBA-Lite 3.0
- Further enhance the robustness of econometric model estimates.
- Maintain high-level conceptual consistency with the EBA models.
- Ensure appropriate adaptations vis-à-vis the EBA models to help explain implications for current account balances given diverse characteristics of emerging and developing economies.

### Main refinements and updates in EBA-Lite 3.0
- Data updates, refinements to the construction of some explanatory variables, and improvements in model robustness.
- Use of Bayesian Model Averaging (BMA) to exclude non-robust variables from the current account model.
- Operational guidance for other EBA-Lite tools (External Sustainability model and commodity modules) updated based on surveillance experience.

### Operational role in IMF surveillance
- EBA-Lite models are a key input in Article IV staff reports. External sector assessments should assess:
  - (i) the position and trajectory of foreign assets and liabilities,
  - (ii) current accounts,
  - (iii) real exchange rates,
  - (iv) capital flows and policy measures, and
  - (v) foreign exchange intervention and reserve levels.
- The bottom-line assessment must be consistent with numerical inputs from EBA and EBA-Lite models, analytically grounded judgment, and country-specific insights.
- The CA model is typically the preferred approach; staff judgment is required to select the appropriate model and, where necessary, adjust model results to capture country-specific factors not accounted for by the models.

### Key characteristics of EBA-Lite economies that shape model design
- Ninety percent of EBA-Lite economies that use the framework are either low income or emerging market economies (contrast with EBA economies which are predominantly advanced economies or emerging markets).
- Nearly 30 percent of EBA-Lite economies experienced annual damages from natural disasters that, on average, accounted for more than 0.5 percent of GDP since 1995.
- Higher incidence of armed conflicts in EBA-Lite economies relative to EBA economies.
- Remittances: important source of external financing for many low-income countries and emerging markets; accounted for in EBA-Lite models (not in EBA).
- Financial development: larger share of EBA-Lite economies are less financially developed; EBA-Lite considers both credit cycle and level of financial development.
- Exhaustible commodity exporters: more than 20 percent of the EBA-Lite sample economies are fuel and non-fuel commodity exporters; motivates commodity modules.
- External indebtedness: many EBA-Lite economies have high levels of external indebtedness, including foreign-currency denominated debt; EBA-Lite External Sustainability (ES) approach accounts for this.
- Greater heterogeneity: EBA-Lite economies exhibit a significantly wider range and larger variance of current account balances compared with EBA economies; country-specific CA balance heterogeneity over time is larger.

### EBA-Lite current account (CA) model — specification and estimation
- Model form (panel regression): CAit = α + β'Fit + λ'Pit + γ'Cit + eit, where:
  - CAit is the current account balance in percent of GDP in country i in year t.
  - Cit: cyclical factors and shocks.
  - Pit: policy variables.
  - Fit: macroeconomic and structural fundamentals.
  - eit is the zero mean, normally distributed regression residual, assumed to follow an AR(1) process.
- Most variables measured relative to the world average (GDP-weighted average of all countries in the EBA-Lite estimation sample).
- Country fixed effects are not included.
- Estimation method: pooled Generalized Least Squares (GLS) with a panel-wide AR(1) correction.
- Estimation sample includes all countries where data are available, including those in the EBA estimation sample.

### Explanatory variable groups and examples
- Cyclical factors and shocks:
  - Output gap (as in EBA models).
  - Commodity terms of trade (revised construction; see below).
  - Two additional variables capturing shocks related to natural disasters and conflicts (included in EBA-Lite but not in EBA).
- Policy variables (directly affected by policy actions):
  - Fiscal balance.
  - Change in reserves/GDP (interacted with capital controls).
  - Public health spending.
  - Credit/GDP.
  - Credit/GDP growth.
- Macroeconomic and structural fundamentals (slow-moving):
  - Net foreign assets (NFA).
  - Output per worker.
  - Share in FX world reserves.
  - Migrant share.
  - Oil and gas trade balance.
  - Demographics.
  - Institutional quality.

### Operational guidance updates
- External Sustainability (ES) approach and commodity modules operational guidance updated based on experience and insights from the use of these models in Fund surveillance.
- EBA-Lite toolkit retains flexibility: staff can apply the most relevant approach depending on country circumstances; model results can be adjusted where appropriate to reflect country-specific factors not captured in the models.

*Prepared by the External Policy Division in the Strategy Policy and Review Department (SPR).*

### 3.0 model as a one standard deviation change in the EBA-Lite 2.0 model.

### ppea2024054 - 3.0 model as a one standard deviation change in the EBA-Lite 2.0 model.

### Key variable changes and definitions
- Oil and gas trade balance
  - Proxied by a 3-year moving average of the oil and gas net export (relative to GDP) to smooth out price fluctuations.
  - Net exports of oil and gas are assessed as if their prices were at their long-term trend level.
- Capital controls
  - EBA-Lite 3.0 uses the Financial Account Restrictiveness Index (FARI) constructed by the Fund, replacing the Chinn-Ito index.
  - The Chinn-Ito and FARI indexes correlation coefficient was 0.83 in 2019.
  - FARI avoids sample-dependent principal component variation inherent in Chinn-Ito and facilitates stability of the CA norm.

### Model selection and Bayesian Model Averaging (BMA)
- BMA approach
  - Analyzes all possible variable combinations to calculate posterior inclusion probability (PIP).
  - Given the number of explanatory variables in the EBA-Lite 2.0 model, over 33 million rounds of estimations were performed.
  - A potential explanatory variable is considered robust if PIP > 50 percent.
- Variables removed by BMA (from EBA-Lite 2.0)
  - life expectancy at age 45 (relative to world average)
  - lagged NFA/GDP interacted with dummy for NFA/GDP >-60 percent
  - lagged demeaned VIX*capital openness
  - lagged demeaned VIX*capital openness*share in world reserves
- Impact of removals
  - Removing these variables had little impact on overall fit and on other coefficient estimates.

### Estimation results and model fit
- Final refined CA model incorporates:
  - new commodity terms-of-trade, new oil and gas trade balance, capital controls measured with FARI, and model selection per BMA.
- Stability and fit
  - Most coefficient estimates are remarkably stable relative to prior specifications.
  - Model fit is marginally improved compared to the EBA-Lite 2.0 model.
  - In terms of model fit, the R-squared is marginally higher than that for the EBA model but the RMSE is higher, reflecting the higher prevalence of outliers in the sample.
- Coefficient changes
  - Two demographic variables (dependence ratio and population growth) have smaller coefficients than in the previous model.
  - The estimated impact of natural disasters (interacted with capital openness) is larger in the revised model.

### Cyclical factors and shocks
- Output gap
  - An increase in the relative output gap of 1 percent is associated with a reduction in the current account balance by about 0.13 percentage points of GDP.
- Commodity terms-of-trade (interacted with trade openness)
  - For an economy with an openness degree of 50 percent, a 1 percent improvement in the term-of-trade is associated with an increase in the current account balance by about 0.1 percentage points of GDP.
- Natural disasters
  - The model includes a dummy for natural disaster occurrence; impact depends on capital account openness.
  - Example FARI indices: Solomon Islands FARI = 0.78 in 2020; Papua New Guinea FARI = 0.14 in 2020.
  - Ceteris paribus, a natural disaster would reduce the current account of the latter (more open) but could increase the current account of the former (more closed).
- Conflicts
  - A dummy for armed conflict is included.
  - The estimated marginal impact of an armed conflict is to raise the current account by 0.7 percent of GDP.

### Macroeconomic fundamentals
- Net foreign asset (NFA) position (lagged)
  - Estimated coefficient = 0.026, positive and similar to EBA-Lite 2.0 but lower than in the 2022 EBA model.
- Domestic currency share in world FX reserves
  - Estimated coefficient indicates that for every 10 percent of global reserves held in domestic currency, a country’s current account deficit is lower by about 0.7 percent.
- Output per worker (lagged)
  - Higher relative output per worker by one percent is associated with an increase of about 0.087 in the current account balance.
  - This variable is interacted with capital account openness; impact increases for more open economies.
- Expected real GDP growth (5 years ahead)
  - A one percent relatively higher expected real GDP growth is associated with a lower current account balance by about 0.64 percentage points of GDP.

### Structural fundamentals
- Demographics
  - Included variables: population growth, old age dependency ratio, share of prime-age savers as proportion of working age population, life expectancy of prime-age savers interacted with expected future old age dependency ratio.
  - Estimated signs align with priors: negative coefficients on population growth and old age dependency; positive association of prime-age savers with higher CA balances.
- Institutional quality
  - Proxied by the International Country Risk Guide (ICRG).
  - Estimated coefficient has expected sign: higher institutional quality tends to lower the current account balance.
- Migrant share (outward migration as a share of population)
  - Used as a proxy related to remittances.
  - Estimated impact: a 9 percentage point (or one standard deviation) increase in outward migration lowers the current account balance by about 1 percent of GDP.

*Source: ppea2024054 - 3.0 model as a one standard deviation change in the EBA-Lite 2.0 model.*

### 26.      Oil and gas reserves. Exporters of exhaustible natural resources tend to save a higher share

### 26.      Oil and gas reserves. Exporters of exhaustible natural resources tend to save a higher share

### Oil and gas reserves and current account
- Model includes a variable combining the size of the oil and natural gas balance (in percent of GDP) with a measure of temporariness based on the ratio of current extraction to proven reserves.
- Exports of oil and gas are assessed as if prices are at the long-term trend level to insulate this variable from short-term price fluctuations.
- Estimated coefficient has the expected sign and indicates that an increase in the oil and gas reserves increases the current account.

### Policy variables — role and interpretation
- Distinction between macroeconomic/structural variables and policy variables enables:
  - Computation of a current account norm based on desirable policies (P*).
  - Computation of the contribution of policy gaps (actual minus desirable policies) to deviations of the current account from its norm.
- Policy gaps enter the decomposition of the current account gap; the contribution of policy gaps is defined as (''ˆ*(it)it PP γ − ).

### Fiscal policy (cyclically adjusted fiscal balance, instrumented)
- Fiscal policy is instrumented because fiscal decisions are endogenous.
- Instrumentation: cyclically adjusted overall general government fiscal balance is instrumented with lagged world real GDP growth, lagged world output gap, lagged world cyclically adjusted general government fiscal balance, lagged global risk aversion (proxied by lagged U.S. corporate credit spreads), GDP per capita, and a democracy ranking; first stage also controls for independent current account regressors.
- Estimated coefficient: 0.43 — an increase in the fiscal balance of 1 percent (relative to the rest of the world) increases the current account balance.
- Comparison: EBA coefficient is 0.4.

### Health spending (lagged)
- Public health expenditure used as proxy for social safety net; correlated with World Bank Aspire measures.
- Expected effect: more generous social safety net reduces precautionary savings.
- Estimated effect: countries where health expenditure is 1 percent of GDP higher tend to have lower current account balances by 0.57 percentage points of GDP.

### Foreign exchange intervention (FXI) interacted with capital controls (instrumented)
- FXI proxied by change in reserves; interaction with FARI index (capital controls).
- Instrumentation: FXI instrumented with global reserves accumulation, US corporate spreads, M2 relative to GDP (all interacted with capital controls index); first stage also controls for independent CA regressors.
- Estimated sign: expected — an increase in FX purchases tends to improve the current account.

### Credit variables
- EBA-Lite includes two credit variables: credit-to-GDP ratio and change in credit-to-GDP ratio.
- Greater financial depth relative to the rest of the world associated with a lower current account balance.
- Rapid increase in credit-to-GDP signals credit boom associated with wider current account deficits.
- Estimated coefficients have the expected sign: both greater financial depth and rapid credit growth tend to lower the current account.

### Estimating CA norms and gaps — definitions and decomposition
- Cyclically adjusted current account (AdjitCA) adjusts CA for short-term cyclical factors and shocks (C'it), with coefficient estimates denoted with a hat:
  - 'ˆAdjitCA = CAit − C'itβ̂
- Current account norm (Normit) is consistent with medium-term fundamentals (F'it) and medium-term desirable policies (P'*):
  - ''ˆˆNormit = F'itα̂ + P'*itλ̂γ̂
  - Policy gap = (P*it − Pit), contribution to CA gap = (''ˆ*(it)it PP γ − )
- Current account gap (gapitCA):
  - gapitCA = AdjitCA − Normit = (− (P*it − Pit)γ̂) + eit
  - Equivalently, gapitCA equals the sum of model-identified policy gaps and the regression residual.
- Residual interpretation: may reflect structural factors not captured by the model; persistent large residuals may justify consideration of a norm adjustor.

### Multilateral consistency
- Excess deficits should sum to excess surpluses; estimated CA gaps should sum to zero.
- In practice, CA gaps do not sum to zero due to global statistical discrepancies and some variables not expressed as deviations from world average.
- Gaps receive a small adjustment attributed to the CA norm to ensure multilateral consistency.

### Setting desirable policy levels (P*) — operational guidance (summary)
- Fiscal policy P*: cyclically adjusted general government fiscal balance recommended by country teams over the medium term (five years); may be based on debt-stabilizing primary balance, fiscal rules, long-term adjustment needs, or higher spending needs.
- Public health expenditure P*: benchmark regressions provide guidance; regressors include GDP per capita (PPP), old-age dependency ratio, income inequality, and fiscal revenue; P* not expected to change year-to-year as medium-term objective.
- Private credit P*: desirable private credit-to-GDP consistent with role as indicator of financial depth; benchmark regression links credit-to-GDP to financial development, GDP per capita, public debt, inflation, etc.; P* for change in private credit set to achieve desirable credit-to-GDP level.
- FXI P*: proxied by change in reserves relative to GDP; desirable medium-term level is zero change in reserves relative to GDP (no further accumulation); in exceptional cases P* could be set to annualized change needed to close gap to adequate reserves.
- Capital controls P*: benchmark desirable level is either the country’s current level or the contemporaneous cross-country average level of the controls index, whichever is smaller (least restrictive).
- Real short-term interest rates (REER equation only): desirable short-term interest rate equals the monetary policy stance appropriate for domestic output and inflation objectives; if current policy appropriate, P* = actual.

### Adjustors — temporary and structural
- Adjustors applied transparently and conservatively; internal review to ensure adjustors are justified, theory-grounded, appropriately computed, evenhanded, and multilaterally consistent.
- Adjustments can be to cyclically adjusted current account or to the norm:
  - Cyclical adjustors for temporary shocks not fully captured by model (e.g., one-off import-intensive projects, prolonged yet transitory natural disasters, one-off shocks to remittances); calibrate to temporary portion; typically one or two years.
  - Mandatory adjustors for broad-based developments not covered by model (example: Covid-19 adjustors to account for transitory net impacts on tourism, commodity volumes, medical goods trade, and shift in consumer preferences).
  - Norm adjustors for country-specific structural factors not included in model (e.g., large migrant share with very low remittances). Norm adjustors, once introduced, should be maintained for subsequent assessments.

### REER gap and elasticities
- CA gap converted to REER gap using semi-elasticity of trade balance-to-GDP ratio with respect to REER (η_TB):
  - η_TB = ∆(TB/GDP) / ∆REER/REER
- REER gap formula:
  - REER_gap = gapitCA / η_TB
- Semi-elasticity η_TB is estimated as weighted average of export and import elasticities:
  - η_TB = η_X s_X − η_M s_M
  - where η_X and η_M are elasticities of exports and imports wrt REER; s_X and s_M are nominal shares of exports and imports to GDP.
- Country-specific semi-elasticities: use panel-estimated η_X and η_M and country-specific export/import shares; an eleven-year moving average smooths cyclical fluctuations in shares.
- Note: 2022 EBA refinements introduced estimates of income balance semi-elasticities; they were generally close to zero and were not added to the EBA-Lite framework, though staff can use alternative well-justified estimates if appropriate.

*Source: ppea2024054 - 26.      Oil and gas reserves. Exporters of exhaustible natural resources tend to save a higher share*

### 37.      Specifically, dynamic export (X) and import (M) equations are estimated using an

### ppea2024054 - 37

### Export and import equation estimation (EBA-Lite sample and elasticities)
- An unbalanced panel covering EBA-Lite countries with quarterly data between 1980 and 2019 is used to estimate dynamic export (X) and import (M) equations.
- Exports and imports are expressed in nominal USD.
- Reduced-form specifications relate:
  - Exports to lagged exports, real effective exchange rates (REER), and world demand proxied by trading partners’ real GDP (with time and country fixed effects).
  - Imports to lagged imports, REER, and domestic demand proxied by domestic real GDP (with time and country fixed effects).
- Long-run elasticities are computed from estimated short-run coefficients and lag coefficients according to the panel regression formulation.
- Estimated long-run elasticities (EBA-Lite estimation sample):
  - Import elasticity: 0.25
  - Export elasticity: -0.46
- Comparison with EBA-Lite 2.0 elasticities:
  - Import elasticity (EBA-Lite 2.0): 0.29
  - Export elasticity (EBA-Lite 2.0): -0.44

### EBA-Lite REER model: purpose, specification, and refinements
- Purpose and feature:
  - Focuses on country-specific determinants of movements in REER indices.
  - REER indices are normalized to 100 in the base year; the model uses country fixed effects, so model residuals for each country average to zero over the sample period.
  - Fixed effects complicate interpretation and reduce robustness to structural breaks; REER model provides a historical (own-history) perspective rather than cross-country comparisons.
- Sample and data updates:
  - EBA-Lite REER model sample extended from 1995–2016 to 1995–2019.
  - Uses updated database aligned with the CA model.
  - Terms-of-trade gap is made consistent with the refined CA model; FARI index is used as measure of capital controls.
- Model selection and refinements (EBA-Lite 3.0):
  - Insignificant variables dropped (also dropped from refined CA model): global risk aversion variables; output per worker interacted with capital openness; oil and gas reserves; conflict dummy variable; and aid.
  - Medium-term growth and population growth variables removed due to unexpected sign and limited impact on fit.
  - Adds output gap (relative to world average) among cyclical factors to align with EBA REER Index model.
  - Adds cyclically adjusted fiscal balance as a policy variable; health spending dropped as insignificant.
  - Aim: improve robustness and closer alignment with EBA REER model.

### Key estimation results and interpretation (EBA-Lite 3.0 REER)
- Cyclical factors and shocks:
  - Output gap (relative to world average): statistically significant; a 1 percent increase in the relative output gap is associated with a real appreciation of about 0.4 percent.
  - Commodity terms-of-trade (interacted with trade openness): positive coefficient; more favorable commodity terms-of-trade associated with more appreciated REER.
  - Natural disasters (with interaction with capital account openness): tend to appreciate the REER, but the impact is mitigated for economies with more open capital accounts.
- Macroeconomic fundamentals:
  - Lagged net foreign asset (NFA) position: relationship with REER is ambiguous; lagged NFA not statistically significant, but interaction with a high-debt dummy (NFA below -60 percent) is positive and statistically significant.
  - Own currency share in world FX reserves: positive coefficient; reserve-currency status associated with more appreciated REER.
  - Output per worker (lagged): rise in relative output per worker associated with higher REER (partial Balassa-Samuelson effect).
  - Remittances (relative to GDP): increase in remittances tends to depreciate the REER (contrasts with CA model treatment).
- Structural fundamentals:
  - Demographics (dependency ratio, life expectancy, interaction): economies with higher savings needs related to demographics tend to have more depreciated REERs.
  - Institutional quality: expected sign (weaker institutions → more depreciated REER), not statistically significant; impact smaller than in EBA-Lite 2.0.
- Policy variables:
  - Fiscal policy (instrumented cyclically adjusted fiscal balance): statistically significant; a 1 percentage point increase in the cyclically adjusted fiscal balance relative to GDP is associated with a REER appreciation of 0.2 percent.
  - Foreign exchange intervention (FXI) interacted with capital controls (instrumented): FX purchases tend to depreciate the REER; effectiveness depends on capital account openness.
  - Monetary policy interacted with capital controls: higher real short-term interest rates tend to appreciate the REER; an increase in the relative real short-term interest rate by 1 percentage point is associated with 1 percent appreciation in countries with open capital accounts.
  - Credit variables (credit-to-GDP ratio and change in credit-to-GDP ratio): greater financial depth and strong credit growth tend to appreciate the REER.

### REER model caveats and preference for CA models
- Caveats:
  - REER model assesses current REER relative to its own history (due to fixed effects), whereas CA model compares current account to desirable policies and fundamentals of peers; models answer different questions and can produce inconsistent results.
  - Inclusion of country fixed effects complicates coefficient interpretation and makes the REER model less robust to structural breaks.
  - Overall fit of REER models tends to be poorer than CA models.
- Guidance:
  - Preference is usually given to CA models in external sector assessments; REER model provides complementary insights but is secondary.

### External sustainability (ES) approach and commodity modules
- External vulnerabilities:
  - Nearly a quarter of economies that use the EBA-Lite framework have NFAs (or NIIP) below -60 percent of GDP, a benchmark frequently used to indicate higher external vulnerabilities.
  - High net external borrowing increases exposure to sudden changes in global financial conditions, roll-over risks, and increases in borrowing costs; concerns about external imbalances have risen recently reflecting post-pandemic increases in debt-to-GDP ratios amid global financial tightening.
- ES approach overview:
  - Models sustainability by comparing medium-term expected CA/GDP to the CA/GDP that would stabilize NFA/GDP at a specified benchmark level.
  - Differs from CA and REER approaches: CA gaps are not attributed to specific policies; ES is more forward-looking.
  - EBA-Lite 2.0 expanded the ES approach (based on Blanchard and Das (2017)) to include trade and financial factors and provides a deterministic approach plus a probabilistic module.
  - Defines net external liabilities as sustainable if ≤ present value of net exports plus rate-of-return differential times gross asset position.
  - When the intertemporal constraint is violated (former larger than latter), real exchange rate depreciation may be needed to ensure external sustainability; framework can estimate needed REER depreciation to maintain NFA at current level in the medium-term.
- ES approach refinements and limitations:
  - Accounts for valuation effects of exchange rate changes linked to foreign currency denomination of assets and liabilities.
  - Generally does not consider return differentials except for a few countries.
  - Requires debt at a future point to be stabilized at its level today (rather than comparing discounted future debt to today).
  - Uses medium-term CA projection rather than entire CA stream.
  - Stabilizing NFA at current level in medium-term does not guarantee optimality; for countries with very large negative NFA, recommended REER adjustments may be necessary but insufficient; alternative benchmarks may be needed but require careful justification.
- Commodity exporter considerations:
  - Commodity exporters face distinct external-sector dynamics; exhaustible resources generate potentially large but temporary income streams, suggesting a role for smoothing domestic absorption and saving export proceeds.
  - Regression-based CA models fit relatively poorly for commodity exporters; residuals generally increase with commodity exports.
  - EBA-Lite introduced two complementary models for commodity exporters:
    - Consumption allocation model (current account norms for smoothing inter-generational consumption of assets).
    - Investment needs model (assesses returns to investment of resource wealth for countries with low capital-to-assess).
  - These commodity-tailored models complement standard EBA-Lite approaches.

*Italic: Source — ppea2024054 (excerpt provided).*

### 66.      Consumption allocation rules that distribute resource wealth across periods can be

### Consumption allocation rules that distribute resource wealth across periods

### Consumption allocation model: concept and application
- Consumption allocation rules derive current account (CA) gaps by distributing resource wealth across periods to smooth inter-temporal consumption.
- Motive to save exhaustible resources arises from inter-temporal smoothing concerns because resource revenue is non-renewable.
- Permanent income hypothesis application: each individual in each generation is allocated the same real resources out of the country’s wealth, implying a constant real per capita annuity as a normative measure for external sector assessments.
- Annuity defined as sum of:
  - below-the-ground wealth (the present value of exports of exhaustible commodities), plus
  - above-ground external wealth (net international investment position).
- The constant real per capita annuity yields a norm for consumption; from this a saving norm is derived and applied to the current account.
- Empirical implications:
  - This approach often suggests current account norms that are higher relative to those from the EBA-Lite CA regression model.
  - The consumption allocation model is tractable and most useful in middle- and high-income countries.
  - In some LICs, immediate investment or development needs may overwhelm longer-term savings considerations, making the consumption allocation model less appropriate.

### Investment module: role, calibration, and suitability
- Models that explicitly account for investment needs can yield lower CA norms in resource-rich developing countries relative to the consumption allocation model.
- Rationale:
  - Consumption allocation rules that hold investment fixed implicitly assume additional savings is invested abroad.
  - In low-income countries where capital is scarce, returns to investing resource wealth domestically may exceed returns to saving externally, so the consumption allocation rule can overstate savings-investment norms.
- Araujo and others (2016) propose a small open economy model that incorporates investment and credit constraints, with separate public and private investment efficiency parameters to illustrate how external misalignments could change if investment efficiency is improved.
- Suitability:
  - The investment-needs model is better suited for countries with large investment needs, lack of market access, but reasonably high investment efficiency.
  - Calibration depends on investment efficiency, which in turn depends on institutional frameworks to implement investments effectively; this can be low in many resource-rich developing countries.
  - Lower investment efficiency => lower levels of optimal investment => higher CA norms.
- Implementation challenges:
  - Results are sensitive to parameterization.
  - In EBA-Lite 3.0 all parameters other than those related to the efficiency of private investment and of public investment have been fixed at empirically supported values to reduce free parameters.
  - Selection of investment efficiency parameters can be guided by a country’s Public Investment Management Assessment (PIMA), or empirical results in Gupta and others (2014) or Pritchett (2020).
  - Assessments should be transparent about underlying assumptions for investment parameters and their justification, and teams should avoid making small changes from year to year.
- Evidence from Araujo and others (2016):
  - Introducing investment frictions can lead to current account norms over 5 percentage points higher than in the absence of such frictions.
  - Distinction in frictions:
    - Investment inefficiencies: one dollar invested translates into less than one dollar of productive capital (lowers optimal investment across all periods).
    - Absorptive capacity constraints: adjusting capital stocks is costly (typically lead to gradual investment strategies).

### Final remarks and methodological guidance
- EBA-Lite framework provides flexibility to tailor external sector assessments to country-specific characteristics; staff should identify a preferred approach based on country characteristics and the model choice is expected not to change from year to year.
- Numerical results from different models should not be averaged.
- Where data availability is very limited and EBA-Lite tools are difficult to apply, assessments may rely on descriptive and qualitative discussions.
- For most EBA-Lite economies, the current account model remains the main workhorse approach due to better empirical fit and interpretability.
- Caveats:
  - There are several caveats with the REER model reinforcing preference for the CA empirical approach.
  - Where external sustainability is an overriding concern, assessments should also be informed by directly using the ES approach (based on the current level of the NFA or an alternative benchmark) or by applying an external sustainability adjustor to the current account models.
  - Commodity modules provide an additional perspective for commodity exporters, although the investment module is sensitive to parameter choices and staff continue to draw lessons from its application.
- Ongoing work to improve the EBA-Lite methodology includes:
  - Improving empirical robustness of the models,
  - Exploring new variables such as the impact of climate change on current account balances,
  - Other adaptations to improve country-specific tailoring while maintaining multilateral consistency.

*EBA-LITE  INTERNATIONAL MONETARY FUND*

---


_Source: https://www.imf.org/-/media/files/publications/pp/2024/english/ppea2024054.pdf_