## wpiea2019065 ## Source details **Canonical URL:** [wpiea2019065](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019065.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2019/wpiea2019065.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019065.pdf.json) --- ### I. Purpose, evolution, and governance of IMF external assessments - Objective: identify when current account imbalances reflect appropriate intertemporal smoothing or raise macrofinancial vulnerabilities. - Historical evolution: - Model-based, multilaterally-consistent assessments began in the early 1990s. - External Balance Assessment (EBA) framework launched in 2012 with new current account and REER models. - Key EBA innovations: expanding policy variables affecting external balances; defining “norms” as levels consistent with fundamentals and desired policies. - Refinements: 2015 (demographic modeling nonlinearities, new REER model); 2018 (strengthened demographics and institutional quality, FX intervention, credit excesses, financial center features). - Governance and process: - Numerical model inputs combined with country-specific judgment. - Year-round dialogue between country teams and External Sector Coordinating Group to ensure multilateral consistency for the largest 30 economies (about 90 percent of global GDP). - Assessments inform bilateral and multilateral policy recommendations. ### II. External Sector Assessment Framework and approach - Objective: distinguish appropriate current account surpluses/deficits from “excess” imbalances relative to fundamentals and desired policies over the medium term. - Method: - Combine multiple numerical inputs from EBA models with analytically-grounded country-specific judgment. - Consider multiple indicators (real unit labor costs, export/import performance). - Use External Sustainability (ES) approach when large net debtor positions raise concerns. - Multilateral consistency: - Most regressors specified as deviations from GDP-weighted global averages to help global aggregation properties. - Consistency procedure applies uniform GDP-term adjustments so policy gaps and residuals sum to zero across the EBA sample. ### III. EBA current account model: sample, estimation, and structure - Sample and data: - Sample of 49 countries; annual data 1986-2016 (originally 1986-2010 with three-year extensions per refinement). - Incorporates BPM6 migration and UN World Population Prospects (2017 Revision). - Estimation and identification: - Pooled Generalized Least Squares (GLS) with panel-wide AR(1) correction; country fixed effects excluded. - Lagged CA excluded despite statistical significance. - Endogeneity treatment: instrumentation for fiscal policy and FX intervention; lags for other fundamentals/policies. - Instrumentation used for identification only; policy gaps computed at actual policy variable values. - Dependent variable and regressor groups: - Dependent variable: CA-to-GDP ratio (CA/GDP). - Determinants grouped: cyclical factors; fundamentals (macro and structural); policy variables. - Many regressors specified as deviations from GDP-weighted world averages (marked with # in tables). ### IV. Cyclical factors and cyclically-adjusted CA - Output gap (relative to world average): - Measured in percent of potential GDP. - Estimated coefficient: an increase in the relative output gap of 1 percent reduces the current account balance by about 0.35 percentage points of GDP. - Table 1 coefficient: Output gap #: -0.356*** (0.000). - Commodity terms-of-trade gap (interacted with trade openness): - Commodity TOT index constructed from 43 export/import categories relative to manufactured goods prices in advanced economies. - Estimated example: a 10 percent temporary terms-of-trade improvement is associated with a 0.8 percentage points of GDP increase in CA for a country with openness 50 percent of GDP. - Table 1 coefficient: Commodity terms of trade gap (interacted): 0.161*** (0.000). ### V. Macroeconomic and structural fundamentals (2018 specification) - Selected estimated coefficients and interpretations (Table 1 and text): - Net foreign assets (lagged): coefficient 0.023 (Table 1: 0.023*** (0.000)); non-linearity: dummy if NFA/GDP < -60 percent interacts negatively. - Output per worker (lagged, relative to frontier): a 10 percent increase in relative output per worker increases CA by about 0.64 percent of GDP in fully open economies (text); Table 1 entry: Output per worker (lagged) 0.023 (0.259). - Expected real GDP growth (5 years ahead): a 1 percentage point increase lowers CA by about 0.3 percent of GDP; Table 1: -0.302*** (0.004). - Reserve currency status: for each 10 percentage points of global reserves held in currency, CA weakened by about 0.3 percent of GDP; Table 1: Reserve currency status: -0.030*** (0.009). - Demographics (static and dynamic): - Static measures: population growth, old-age dependency ratio (OAD), share of prime-aged savers. - Dynamic measures: life expectancy of current prime-aged saver and its interaction with future OAD. - Table 1 coefficients: - Old-age dependency ratio (OAD) #: -0.069 (0.109). - Population growth #: -0.692* (0.061). - Share of prime-aged savers #: 0.138** (0.013). - Life expectancy #: -0.005*** (0.000). - Interaction life expectancy × future OAD: 0.013*** (0.005). ### VI. Policy variables and estimated quantitative effects - Fiscal policy (cyclically-adjusted general government overall balance, instrumented): - Estimated coefficient: 0.33 (Table 1: Fiscal policy #: 0.329*** (0.000)). - Public health spending (lagged): - An increase in public health expenditures of 1 percent of GDP reduces CA by about 0.4 percentage points of GDP. - Table 1: Health spending (lagged) #: -0.399*** (0.003). - Foreign exchange intervention (FXI, instrumented and interacted with capital controls): - FXI proxied by transaction-based change in reserves plus derivatives operations. - Estimated effects (text and Table 1): - Table 1 reports FXI (interacted with capital controls) #: 0.754*** (0.001). - Text illustrative effects: For a country at the 75th percentile of capital controls, a 1 percent of GDP in FX purchases leads to a 0.19 percent of GDP CA improvement (earlier EBA 0.11); at the 90th percentile, effect is 0.38 percent. - Financial excesses (credit gap): - Credit gap measured via one-sided HP filter (penalty parameter 1,600). - Estimated effect: a 10 percent of GDP increase in credit relative to trend associated with a 1 percent of GDP deterioration in CA. - Table 1: Credit gap #: -0.104*** (0.000). - Capital controls (Quinn index, 0 to 1) enter via interactions: - Example interaction: Output per worker (lagged, interacted with capital openness) 0.041* (0.051) in Table 1. ### VII. Model fit, robustness, and sensitivity - Overall fit: - Table 1: Observations: 1,367; Number of countries: 49; R-squared IV: 0.524; R-squared Fit: 0.550; Root MSE: 0.031. - R-squared roughly unchanged relative to original 2012 model; about 10 percent higher than 2015 version; RMSE 6 percent smaller relative to 2015. - Robustness highlights: - Institutional quality: alternative ICRG/WGI combinations yield similar baseline results; ICRG selected for coverage. - Financial excesses: alternative HP lambda values, demeaned credit, and change-in-credit specifications tested; one-sided HP filter (lambda 1,600) retained for superior fit. - Demographics: replacing life expectancy with aging speed weakens significance; interaction life expectancy × future OAD remains significant. - FXI instrumentation: FXI not instrumented yields smaller coefficients (endogeneity downward bias); narrow FXI (spot-only) gives larger coefficients; broader FXI including derivatives necessary given off-balance-sheet operations. ### VIII. REER models: structure, determinants, and fit - Two models: REER-Index (captures country-specific movements; country fixed effects) and REER-Level (captures persistent cross-country REER levels using ICP-based 2011 base). - Sample period: 1990-2016; Index: 40 countries; Level: 39 countries. - Common determinants and illustrative magnitudes (text and Tables 7–8): - Commodity terms-of-trade: 10 percent increase appreciates REER by 1.8 percent (Index) and 0.6 percent (Level). - Output per worker (10 percent increase) appreciates REER by about 2 percent in both models. - Expected real GDP growth (1 percentage point) appreciates REER by about 2 percent in both models. - Trade openness (10 percentage points) depreciates REER by about 2 percent (Index) and 3 percent (Level). - Monetary policy (1 percent real rate differential, fully open capital accounts) appreciates REER by about 0.6 percent. - Model-specific highlights and coefficients (Table 7): - REER-Index notable coefficients: - Commodity TOT (log): 0.184*** (0.000). - NFA (lagged): -0.109*** (0.000). - Output per worker (log, lagged): 0.217*** (0.000). - Expected real GDP growth 5 years ahead: 2.012*** (0.003). - Institutional quality: 0.653*** (0.000). - Health spending (lagged): 2.040** (0.015). - FX intervention (interacted): -2.479* (0.078). - REER-Index fit: Observations 1,004; Number of countries 40; R-squared 0.548; RMSE 0.089. - REER-Level notable coefficients: - NFA (lagged): 0.056*** (0.003). - Output per worker (log, lagged): 0.171*** (0.000). - Expected real GDP growth 5 years ahead: 1.961** (0.048). - Institutional quality: 0.653*** (0.000). - REER-Level fit: Observations 990; Number of countries 39; R-squared 0.895; RMSE 0.146. - REER model diagnostics: - REER-Level R-squared = 0.9; REER-Index R-squared = 0.58. - Unit root and cointegration tests on residuals reject non-stationarity; residuals are persistent but not permanent. ### IX. Constructing CA and REER norms and gaps - Current account regression notation and decomposition: - CACA_GG = α + X_fund' β_fund + X_cyc' β_cyc + P' γ + e (e ~ AR(1)). - Predicted CA: CACA_GG_hat = α_hat + X_fund'_hat β_fund_hat + X'_hat β_hat + P'_hat γ_hat. - Predicted CA decomposed into: - cyclically-adjusted “CA norm” (fundamentals at actual values, policy variables set at desirable medium-term levels P*), - cyclical component (output gap and terms-of-trade), - policy gaps (P − P*), - regression residual. - Definition: - Cyclically-adjusted CA = predicted CA net of cyclical factors: CACA_cyc_adj.P = CACA_GG.P − X_cyc' β_cyc_hat. - Total EBA CA gap = cyclically-adjusted CA − cyclically-adjusted norm = sum of model-identified policy gaps + residual. - Residual captures other policy distortions, unmodeled fundamentals, measurement errors. - CA–REER translation: - CA–REER semi-elasticity η_TT converts CA gaps into REER gaps via η_TT = η_X * s_X − η_M * s_M. - Two approaches for η_TT: 1. Original CGER calibration: η_X = 0.71; η_M = 0.92; semi-elasticities computed with country-specific trade shares (excluding oil). 2. CGER-inspired panel estimation (1980Q1–2017Q4): panel regressions yield η_X = 0.11 and η_M = 0.57 (volume elasticities); trade shares computed on aggregate imports and exports (including oil). - Table 11 provides country-specific semi-elasticities under both approaches; correlation between approaches = 0.9 (examples preserved exactly): - Argentina: -0.13 / -0.11. - Australia: -0.20 / -0.15. - Austria: -0.45 / -0.36. - United States: -0.12 / -0.10. ### X. Benchmarks for policy variables (P*) and uncertainty - Required P* benchmarks: fiscal balance, public health spending, capital controls, FX intervention, monetary policy; guidance for credit gaps. - Fiscal P*: anchored around cyclically-adjusted fiscal balance desirable in medium/long term (debt-stabilizing primary balance or long-term aging costs). - Public health spending P*: regression-based benchmark using PPP GDP per capita, current OAD, and Gini; Table A.2 regression results: - Relative Income: 0.018***. - Old age dependency ratio: 0.096***. - Gini coefficient: 0.0175. - Constant: 0.021*. - Observations: 49; R-squared: 0.835. - Capital controls P*: medium-term benchmark = min(country actual, cross-country average 0.16 in 2017). - FXI P*: normally zero (if reserves adequate per ARA); nonzero desirable levels may be set if reserves significantly below ARA. - Monetary policy P*: desired short-term interest rate = neutral real rate consistent with output and inflation objectives; if inflation near target P* often equals actual. - Credit gap P*: desirable credit gap P* should be zero in most cases; adjustments possible for financial deepening or historical busts. - Standard errors: - 2018 refinements introduce standard errors for estimated CA norms computed as sqrt(Var( X_a' β̂ + P*_a' γ̂ )), reported with EBA CA norms; staff may use larger uncertainty ranges if justified. ### XI. External Sustainability (ES) approach: formulation and usage - Purpose: determine CA-to-GDP ratio that stabilizes NFA-to-GDP at a benchmark over medium term; complementary to EBA models when debtor/creditor positions are large. - Key formulation points: - Law of motion: change in NFA = CA + valuation changes; ES typically assumes valuation changes VC = 0. - Steady-state expressions link required CA to desired lnNFA, medium-term inflation π, potential growth g, and real return r (see equation (11) in source). - Implementation steps in EBA: 1. Choose benchmark NFA-to-GDP (often current level; for large debtors choose stronger target informed by peers). 2. Derive CA that stabilizes NFA at benchmark using π and g. 3. Produce staff medium-term CA under current policies and REER; adjust for cyclical components. 4. CA gap = staff medium-term CA under current policies − CA that stabilizes NFA. 5. Convert CA gap to REER gap using REER–CA semi-elasticities. - Limitations: requires normative benchmark for NFA; does not provide specific policy gaps. ### XII. Complementary tools: measurement-bias adjustor and structural-policy tool - Measurement-bias adjustor: - Motivated by BPM6 transaction recording differences and inflation component in nominal investment income that bias NIIP valuation changes. - Retained earnings treatment and inflation in interest income can create systematic measurement biases. - 2018 approach: remove financial center dummy; introduce outside-of-model adjustors when empirical evidence of retained earnings and inflation measurement biases is sizable and consistent with stock-flow differences. - Magnitude determined case-by-case; data constraints (stock-flow reconciliation, IIP currency exposure) limit precision. - Structural-policy tool: - Two-stage approach: regress EBA CA residuals on structural policy indicators S̃_{a j}. - Indicators (examples): OECD LPS, OECD EPL, WEF SBP, WEF CLER. - Findings (Table 12 samples): - OECD specification: LPS coefficient 0.0049** (Observations 374; R-squared 0.026). - WEF specification: EPL -0.0048**; SBP 0.0242**; CLER 0.0508*** (Observations 533; R-squared 0.053). - Normative exercise: compare S to best-practice frontier S* to estimate the extent reforms can explain residuals. ### XIII. Data, country coverage, and priority research directions - Country coverage (Annex IV): current account model includes 49 countries (list preserved in source); euro area treated by GDP-weighted averages of largest 11 members corrected for intra-European balances; euro area not included directly in EBA models. - Key data sources and definitions (Annexes II–III): WEO, Lane & Milesi-Ferretti NFA, UN WPP 2017, ICRG, Quinn index, BIS, Penn World Table, ICP 2011, etc. (variable definitions and measurement conventions preserved in source). - Priority data needs and research directions: - Better labor and product market policy measures. - External statistics improvements: currency composition of cross-border positions, stock-flow reconciliation. - Research on sectoral net saving and corporate savings, measurement biases from multinational activity, risks from stock imbalances, aging and pension design interactions, and exchange rate adjustment role under dollar dominance and global value chains. ### XIV. Selected quantitative results and robustness (excerpted exact figures) - EBA current account regression main coefficients (Table 1 exact entries): - Output gap #: -0.356*** (0.000). - Commodity TOT gap (interacted): 0.161*** (0.000). - NFA (lagged): 0.023*** (0.000). - Expected real GDP growth 5 years ahead #: -0.302*** (0.004). - Reserve currency status: -0.030*** (0.009). - Population growth #: -0.692* (0.061). - Share of prime-aged savers #: 0.138** (0.013). - Life expectancy #: -0.005*** (0.000). - Life expectancy × future OAD: 0.013*** (0.005). - Institutional quality #: -0.047** (0.015). - Exhaustible resources of oil and natural gas: 0.310*** (0.000). - Fiscal policy #: 0.329*** (0.000). - Health spending (lagged) #: -0.399*** (0.003). - FX intervention (interacted) #: 0.754*** (0.001). - Credit gap #: -0.104*** (0.000). - Observations: 1,367; Number of countries: 49; R-squared IV: 0.524; R-squared Fit: 0.550; Root MSE: 0.031. - REER model fit (Table 7 exact entries): - REER-Index: Observations 1,004; Number of countries 40; R-squared 0.548; RMSE 0.089. - REER-Level: Observations 990; Number of countries 39; R-squared 0.895; RMSE 0.146. - CA–REER semi-elasticities (Table 11 selected exact examples): - Argentina: Original CGER -0.13 / CGER-inspired -0.11. - Ireland: Original CGER -0.92 / CGER-inspired -0.65. - United States: Original CGER -0.12 / CGER-inspired -0.10. *Source — wpiea2019065 (excerpts from IMF EBA methodology paper).* ### References ............................................................................................................. ### wpiea2019065 - References ### I. Introduction: purpose and evolution of IMF external assessments - The IMF assesses external positions and exchange rates to identify when current account imbalances reflect appropriate intertemporal smoothing or raise macrofinancial vulnerabilities. - Historical context and evolution: - Model-based, multilaterally-consistent assessments began in the early 1990s. - The External Balance Assessment (EBA) framework was launched in 2012 with new current account and REER models. - Key innovations of the EBA framework: - Expanding the set of policy variables that affect external balances. - Defining “norms” as levels of the current account or real exchange rate consistent with fundamentals and policies at their “desired” levels. - Subsequent refinements: - 2015: revisions to demographic modeling (nonlinear effects) and introduction of another REER model. - 2018: strengthened modeling of demographics and institutional quality, macroeconomic policies (foreign exchange intervention and credit excesses), and country-specific features (financial centers); REER models largely unchanged but refined for consistency with current account model changes. - Complementarity and judgment: - EBA models provide numerical inputs but do not capture all country characteristics or policy distortions; assessments are complemented by country-specific judgment. - A process exists to achieve multilaterally-consistent assessments for the largest 30 economies (about 90 percent of global GDP). ### II. The External Sector Assessment Framework - Objective: distinguish appropriate current account surpluses/deficits from “excess” imbalances relative to fundamentals and desired policies over the medium term. - Approach: - Multiple numerical inputs from EBA models are combined with analytically-grounded, country-specific judgment. - Consideration of multiple external sector and competitiveness indicators (e.g., real unit labor costs, export and import performance). - External Sustainability approach used where large net debtor positions raise concerns. - Governance: - Year-round dialogue between country teams and an interdepartmental External Sector Coordinating Group to vet assessments and ensure multilateral consistency. - Assessments inform policy recommendations at bilateral and multilateral levels. ### III. The EBA Current Account Model: scope, principles, and innovations - Conceptual foundation: - Builds on literature on macroeconomic determinants of saving and investment. - Most regressors are specified as deviations from the GDP-weighted global average, ensuring multilateral consistency and enabling decomposition into domestic and foreign components. - Determinants selected based on conceptual underpinning and consistency of estimated coefficients with theoretical priors; policy variables generally required to be statistically significant. - Key model principles: - Normative interpretation of model outputs to identify desirable policy-related levels. - Use of complementary tools to assess structural policy roles and measurement biases in current account statistics. ### IV. Sample and Estimation - Sample and data coverage: - Estimated for a sample of 49 countries using annual data for the 1986-2016 period. - Incorporates data revisions including migration to the IMF’s Balance of Payments Manual, 6th edition (BPM6) and demographic estimates/projections from the 2017 Revision of UN World Population Prospects. - Estimation method: - Pooled Generalized Least Squares (GLS) with a panel-wide AR(1) correction to address autocorrelation of current account data. - Country fixed effects are excluded to avoid capturing persistent policy distortions without economic explanation. - The model excludes the lagged current account despite its statistical significance. - Endogeneity treatment: - Some policy variables are instrumented (fiscal policy and foreign exchange intervention). - Other country fundamentals and policies (net foreign asset position, productivity, global financial conditions, health spending) are lagged. ### V. Regression model structure and dependent variable - Dependent variable: - CA-to-GDP ratio. - Determinant grouping: - Cyclical factors. - Fundamentals (macroeconomic and structural). - Policy variables. - Modeling choices and rationale: - Deviations from the GDP-weighted global average imply, for example, that population aging affects a country’s current account only relative to the world average. - The approach allows analysis of how much of a country’s current account position is due to domestic policies versus foreign counterparts. - Policy variable inclusion requires higher evidentiary standards (conceptual and statistical). *Source: IMF — wpiea2019065 (excerpts from the EBA methodology paper).* ### Annex IV provides a list of the EBA countries. The euro area is not included in the EBA models, but the current ### wpiea2019065 - Annex IV provides a list of the EBA countries. The euro area is not included in the EBA models, but the current ### EBA model scope and estimation - Annex IV lists the EBA countries; the euro area is not included in the EBA models. - The current account assessment of the Euro Area is conducted by computing the GDP-weighted averages of the CA norms and gaps of the common currency area’s 11 largest members, corrected for intra-European balances. - The original version of the EBA model was estimated for the period 1986-2010, and with each refinement the sample period was extended an additional three years. - Instrumentation is used for identification purposes only (i.e. properly estimating the coefficient of the policy variables); when calculating policy gaps, the model is evaluated at the actual value of policy variables. - Annex I includes a qualitative description of the refinements introduced since 2013 and highlights situations when the size and statistical significance of the coefficient vary substantially across versions. - Annexes II and III describe the data sources and definitions of the variables included in the regression models. - An online appendix presents actual and fitted values for all EBA models and countries in the sample, including details on the contribution of each regressor. ### Cyclical factors (cyclically-adjusted current account) - Purpose: Strip temporary and cyclical factors from the actual current account balance to derive a cyclically-adjusted measure that can be compared to the medium-term current account norm. - Output gap: - Measured in percent of potential GDP, used to capture business cycle state. - Measured in relative terms with respect to the world average. - Estimated coefficient: an increase in the relative output gap of 1 percent reduces the current account balance by about 0.35 percentage points of GDP. - Commodity terms-of-trade gap (interacted): - Commodity terms-of-trade measured as the ratio of a geometric weighted-average price of 43 commodity export categories to the equivalent geometric weighted-average price of commodity imports, each relative to manufactured goods prices in advanced economies. - Model includes deviations of this index from its trend interacted with trade openness to capture temporary component. - Estimated coefficient: a 10 percent temporary terms-of-trade improvement is associated with a 0.8 percentage points of GDP increase in the current account balance for a country with an openness degree of 50 percent of GDP. ### Macroeconomic fundamentals included in the 2018 EBA specification - Net foreign assets (lagged): - Lagged NFA-to-GDP ratio included to account for effect on net income balance and measurement issues associated with nominal interest income and retained earnings on portfolio equity positions. - Estimated coefficient: 0.023. - Non-linearity: model includes a dummy for countries with an NFA position below -60 percent of GDP (interacted with the NFA-to-GDP ratio); the associated coefficient is negative. - Output per worker (lagged): - Measured as a country’s GDP per working age population (in PPP terms) relative to the average of the top three economies (Germany, Japan and the United States) taken as the frontier. - Interacted with the capital account openness policy variable. - Estimated effect: a 10 percent increase in relative output per worker would increase the current account balance by about 0.64 percent of GDP in fully open economies. - Expected real GDP growth (5 years ahead): - Proxy: Real GDP growth 5 years ahead from the World Economic Outlook (WEO). - Estimated effect: an increase of 1 percentage point in expected real growth lowers the current account by about 0.3 percent of GDP. - Reserve currency status: - Measured by the share of a country’s currency in world reserve holdings. - Estimated effect: for each 10 percentage points of global reserves held in its currency, a country’s current account balance is weakened by about 0.3 percent of GDP. ### Structural fundamentals: demographics - Specification separates static effects (age composition) from dynamic effects (longevity / old age survival risk). - Static/age-composition variables included: - Population growth (partly proxies the share of young). - Old-age dependency ratio (OAD). - Share of prime-aged savers as a proportion of the total working age population. - Estimated coefficient signs align with economic priors: relatively high shares of young or elderly tend to dissave; higher proportion of prime-aged savers tend to save more. - Dynamic effects: - Captured by life expectancy of a current prime-aged saver and its interaction with future (20 years ahead) OAD. - Intuition from a life-cycle model: countries save more when prime-aged savers expect to live longer (or have longer retirement periods), and more so when they cannot rely on future generations for old-age support. - Estimated life expectancy term and its positive interaction with future OAD capture observed non-linearities in the reduced-form relationship between life expectancy and the current account balance. *Source: wpiea2019065 - Annex IV provides a list of the EBA countries. The euro area is not included in the EBA models, but the current* ### Annex V and Dao and Jones (2018) provide further details on the current demographic specification. ### wpiea2019065 - Annex V and Dao and Jones (2018) provide further details on the current demographic specification. ### Institutional quality and natural resources - Institutional quality proxy: International Country Risk Guide (ICRG) indicator expanded to include government stability, law and order, bureaucratic quality, in addition to earlier sub-indicators. - Estimated effect: a country at the 75th percentile of the institutional quality distribution has, all else equal, a 0.5 percentage points of GDP lower current account balance compared to the median country. - Exhaustible oil and natural gas resources: model includes a “temporariness-adjusted” energy balance (size of oil and natural gas balance, in percent of GDP, adjusted by current extraction to proven reserves). - Estimated effect for energy exporters: a 1 percent of GDP increase in the temporariness-adjusted energy balance increases the current account balance by about 0.31 percent of GDP. This coefficient applies to 10 out of the 49 economies in the sample where the net oil and gas balance is positive. ### Policy variables affecting the current account - Fiscal policy (instrumented): - Fiscal measure: cyclically-adjusted general government overall balance, instrumented using lags of global factors and country-specific features (world real GDP growth, world output gap, world cyclically-adjusted fiscal balance, U.S. corporate credit spread as global risk aversion proxy; lagged GDP per capita; lagged output gap; exchange rate regime; democracy index ranking). - Estimated coefficient: 0.33 (interpreted as effect on current account balance). - Health spending (lagged): - Proxy: public health spending relative to GDP; included with a lag to mitigate endogeneity. - Estimated effect: an increase in public health expenditures of 1 percent of GDP reduces the current account by an average of about 0.4 percentage points of GDP. - Foreign exchange intervention (FXI) (interacted and instrumented): - FXI proxied by transaction-based change in reserves (BOP) plus comparable derivatives market operations; instrumented to address endogeneity. - Interaction: FXI-to-GDP ratio interacted with the Quinn index of capital controls. - Estimated effects: - For a country at the 75th percentile of capital controls: a 1 percent of GDP in FX purchases leads to a 0.19 percent of GDP improvement in the current account (compared to 0.11 under the earlier EBA specification). - For a country at the 90th percentile of capital controls: a 1 percent of GDP in FX purchases leads to a 0.38 percent of GDP improvement in the current account (compared to 0.22 under the earlier EBA specification). - Financial excesses (credit gap): - Credit gap measured using a one-sided Hodrick-Prescott (HP) filter on credit-to-GDP with a penalty parameter of 1,600 (annual counterpart to BIS quarterly lambda). - Estimated effect: a 10 percent of GDP increase in credit relative to its trend (credit gap) is associated with a 1 percent of GDP deterioration in the current account. - Coefficient is highly significant; financial cycle has an independent effect beyond the business cycle. - Capital controls (interacted): - Measure: Quinn index (ranging from 0 for full mobility to 1 for no mobility); included via interactions with fundamentals and policy variables rather than as an independent regressor. - Output per worker (lagged) interaction: - Implied effect: a 10 percent increase in output per worker implies about 0.6 percent effect in a country with fully open capital account, and 0.5 percent in a country at the 75th percentile of the capital controls index. - Global risk aversion (lagged, VXO) interaction: - A 10 percent increase in global risk aversion leads to a 0.17 percent of GDP increase in the current account balance of a non-reserve currency country with an average degree of capital account openness. - VXO also interacted with capital openness and share of a country’s currency in world reserves to capture safe-haven offsets; corresponding coefficient is small in quantitative and statistical significance. ### Model fit, robustness, and sensitivity analyses - Overall fit: - R-squared unchanged relative to the original 2012 model (smaller sample), about 10 percent higher relative to the 2015 version; root mean squared error is 6 percent smaller relative to the 2015 version. - Institutional quality robustness: - Alternative combinations of ICRG and Worldwide Governance Indicators (WGI) yield similar baseline model results; ICRG selected for wider coverage and broader indicator set. - Financial excesses robustness: - Alternatives tested: larger HP lambda (25,000), demeaned credit-to-GDP (historical mean), change in credit-to-GDP (current and one-year lag). - Larger HP lambda increases cyclical volatility and lowers estimated parameter; reverting to demeaned credit raises RMSE to 3.3 percent and reduces fit. - Including both demeaned and detrended measures: demeaned credit becomes small and statistically insignificant while detrended coefficient remains. - Change-in-credit specifications yield similar fit and significant coefficients; one-sided HP filter credit gap retained for superior fit and ability to capture sustained financial imbalances. - Demographics robustness: - Replacing life expectancy with aging speed renders demographic coefficients statistically insignificant (aging speed confounds forces and correlates with prime-aged saver share). - Nonlinear effects: squared life expectancy significant alone but insignificant when interaction of life expectancy with future OAD is included; interaction term remains significant, supporting the theoretically-based interaction specification. - Foreign exchange intervention sensitivity: - FXI not instrumented: coefficient still significant but considerably smaller, indicating endogeneity leads to downward bias. - Narrow FXI (spot-only): somewhat larger coefficient, possibly because spot and derivative operations offset each other. - Given increased off-balance-sheet FXI, broader FXI measure (including derivatives) is necessary. ### EBA real exchange rate (REER) models — overview and common determinants - Two REER models: REER-Index and REER-Level (sample period 1990-2016; Index: 40 countries; Level: 39 countries). - REER-Index model captures country-specific REER movements using country fixed effects; does not permit persistent level deviations. - REER-Level model captures cross-country differences in REER levels (persistent deviations) using ICP-based REER levels for base year 2011 extended across 1990-2016 by rescaling REER indices. - Common policy variables: health spending, foreign exchange intervention, financial excesses (fiscal balance excluded from REER models). - Monetary policy included via real short-term interest rate differential (interacted with capital account openness). - Cyclical and macro fundamentals (common effects and magnitudes): - Commodity terms-of-trade: - A 10 percent increase in terms-of-trade appreciates the REER by 1.8 percent in the Index model, and by 0.6 percent in the Level model. - Output per worker (lagged): - A 10 percent increase in output per worker appreciates the REER by about 2 percent in both models (Balassa-Samuelson effect). - Expected real GDP growth (5 years ahead): - A 1 percent increase in expected growth appreciates the REER by about 2 percent in both models. - Trade openness (lagged): - A 10 percentage points increase in openness (exports+imports/GDP) depreciates the REER by about 2 percent in the Index model and 3 percent in the Level model. - Share of administered prices in CPI: - A 1 percentage point increase in the share of administered prices depreciates the REER by about 2 percent in the Index model and 3 percent in the Level model. - Monetary policy (interacted with openness): - For economies with fully open capital accounts, an interest rate differential of 1 percent appreciates the REER by about 0.6 percent in both models. - Structural features: - Net foreign assets (lagged): ambiguous sign across models—positive in REER-Level model (consistent with need for depreciation to correct negative NFA), opposite sign in REER-Index model. - Population growth (demographics): higher population growth associated with more appreciated currency in both models. - Reserve currency status: tends to be associated with more appreciated currencies; interpretation depends on interactions with capital account openness and global risk aversion. *Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019065.pdf* ### 0.7 percent. For a country at the 75 ### wpiea2019065 - 0.7 percent. For a country at the 75 ### REER determinants — summary findings - General cross-model findings: - At the 75th percentile of the capital controls index, effects reported in the earlier section are "about 0.4-0.5 percent" (relative to a baseline 0.7 percent discussed earlier). - Health spending (lagged): - An increase in health spending of 1 percent of GDP appreciates the REER by 2 percent in the Index model and about 4 percent in the Level model. - Interpretation: insufficient safety nets raise precautionary saving, reduce domestic demand, and lead to more depreciated REER; the positive coefficient is consistent with theory and the current account model. - Foreign exchange intervention (FXI, interacted): - Official reserve purchases lead to a real depreciation, with smaller effects in countries where capital is more mobile. - With the new FXI measure (including operations with FX derivatives), estimated coefficients increase in both REER models relative to earlier versions. - A 1 percent of GDP in official foreign exchange purchases leads to a 0.6 real depreciation in the REER-Index model and a 0.9 depreciation in the REER-Level model for a country in the 75th percentile of the capital controls index. - These effects double for a country at the 90th percentile. - Financial excesses: - Credit booms (private credit-to-GDP ratios above long-term trends) lead to REER appreciation. - Statistically significant only in the REER-Index model: an increase in the credit gap of 10 percent of GDP appreciates the exchange rate by about 1 percent. - Capital controls (interacted): - Effects of capital controls are included indirectly via interactions with global risk aversion and other policy variables (FXI and monetary policy). - Global risk aversion (lagged): - Increased risk aversion tends to weaken most countries' currencies (especially financially integrated ones) and strengthen reserve currencies. This result is visible in both REER models. ### C. REER-Index Determinants (model-specific) - Omitted slow-moving variables: - Country fixed effects in the REER-Index model render some slow-moving variables (institutional quality, demographic indicators such as population age composition and longevity risks) statistically insignificant and hence excluded. - Output gap: - A positive output gap of 1 percent appreciates the REER by about 0.4 percent in the Index model. - Financial home bias (lagged): - Proxied by the share of domestic debt owned by residents; enters with a lag to avoid compositional effects. - An increase in the degree of home bias appreciates the exchange rate; estimated coefficient is 0.2. ### D. REER-Level Determinants (model-specific) - Purpose: - REER-Level measures differences in relative price levels across countries; includes supply-side productivity proxies and slow-moving structural features (demographics, institutional quality, indirect taxation). - Capital-labor ratio (lagged): - A 10 percent higher capital-labor ratio is associated with a 1 percent real exchange rate appreciation (Bhagwati-Kravis-Lipsey effect). - Ratio of traded/non-traded sector productivity (lagged): - An increase in relative productivity of 10 percentage points appreciates the exchange rate by about 2 percent (Balassa-Samuelson effect). - Demographics (old-age dependency ratio): - Higher OAD ratios raise demand for non-tradable old-age services relative to tradable commodities, increasing relative non-traded prices and leading to REER appreciation; estimated effect supports this hypothesis. - Institutional quality: - Higher institutional quality (a higher value = lower institutional risk) is associated with a more appreciated REER; positive estimated coefficient, consistent with the current account model. - VAT revenue: - The estimated coefficient in the REER-Level model has a positive sign (equivalent to about two-thirds) but is not significant. ### E. REER model fit and robustness - Goodness of fit: - REER-Level model R-squared = 0.9. - REER-Index model R-squared = 0.58. - These R-squared values are not directly comparable as the models measure different aspects. - Stationarity and cointegration: - Several unit root and cointegration tests reject hypotheses of non-stationarity of residuals and of no-cointegration among unit-root variables. - Conclusion: regressions using levels of non-stationary variables capture a long-run equilibrating relationship; REER deviations from predicted values are persistent but not permanent. ### V. Constructing current account and REER norms and gaps — approach - EBA current account regression specification (notation as in source): - CACA_GG = α + X_fund' β_fund + X_cyc' β_cyc + P' γ + e (equation (1)) - e is a zero-mean, normally distributed regression residual, assumed to follow an AR(1) process. - P (policy vector) includes: the fiscal balance, measures of capital controls, public health spending, reserve accumulation and financial excesses. - X (fundamentals) and X_cyc (cyclical factors: output gap and terms of trade gap). - Predicted current account (% of GDP): - CACA_GG_hat = α_hat + X_fund'_hat β_fund_hat + X'_hat β_hat + P'_hat γ_hat (equation (2) as presented) - Decomposition into components: - Predicted current account can be decomposed into: - cyclically-adjusted “CA norm” (fundamentals at actual values, policy variables set at desirable medium-term levels P*), - cyclical component (contribution of output gap and terms of trade), - policy gaps (deviations of P from P*). - Expressed in source text (equation (3) format): predicted CA = cyclically-adjusted CA norm + cyclical component + policy gaps. - Definition and interpretation: - The cyclically-adjusted current account is defined as the predicted current account net of cyclical factors: CACA_cyc_adj.P = CACA_GG.P − X_cyc' β_cyc_hat. - Total EBA CA gap (difference between cyclically-adjusted current account and its cyclically-adjusted norm) equals the sum of model-identified policy gaps and the regression residual. - The residual captures other policy distortions or fundamentals not explicitly modeled and regression errors. - Even if overall current account gap is zero (actual = norm), underlying policy distortions may exist if policy gaps offset each other or are offset by the residual. - Policy variables and world averages: - Policy variables are specified as deviations from (GDP-weighted) world averages. - Overall policy gaps appear if a country’s policy distortion differs from the world average. - Example: if all countries have a lower-than-desirable fiscal balance of the same magnitude, the contribution of the fiscal policy gap to each country’s current account would be zero. - Implied: estimated current account gaps reflect domestic policy distortions and respond to policy distortions in other countries, especially large-weight countries. *Italic: Source — content excerpt from wpiea2019065 (IMF PDF chapter).* ### Section VIII, while data limitations prevent a full inclusion in the EBA current account regression model, ### wpiea2019065 - Section VIII, while data limitations prevent a full inclusion in the EBA current account regression model, ### Benchmarks for Policy Variables (P*) - The EBA requires normative policy benchmarks (P*) for: the fiscal balance, public health spending, capital controls, foreign exchange intervention, and monetary policy, and provides guidance for financial excesses. - Fiscal policy: - P* corresponds to levels of the cyclically-adjusted fiscal balance (as a share of potential GDP) desirable from a medium to long-term perspective when output gaps are closed. - Fiscal P* should be anchored around metrics such as the debt-stabilizing primary balance, or long-term adjustment needs given fiscal costs of aging. - Public health spending: - P* guided by a regression benchmark including PPP-based GDP per capita, the current old-age dependency ratio, and income inequality (see Annex VII in source). Staff can choose a different desired level with justification. - Capital controls: - Medium-term desirable benchmark is either the cross-country average of the controls index (0.16 in 2017, out of a potential 0 to 1 range), or a country’s actual level, whichever is smaller. - Asymmetric treatment reflects preference for open capital accounts while recognizing pace considerations for liberalization. - Foreign exchange intervention (FXI): - Desirable medium-term FXI (as a share of GDP) normally set to zero, assuming adequate reserves (including comparable off-balance-sheet FX positions). - Nonzero desirable levels may be set if reserves are significantly below the IMF’s Assessing Reserve Adequacy (ARA) metric, implying extended reserve accumulation may be necessary. - Deviations (policy gap P - P*) in FXI are not automatically interpreted as policy distortions; they may be appropriate responses (e.g., temporary reserve build-up). - Monetary policy: - Desired short-term interest rate = the appropriate monetary policy stance that helps achieve output and inflation objectives (the country desk’s estimate of the “neutral” real rate). - When inflation is close to target, P* typically equals the actual value; otherwise a monetary policy gap is identified and contributes to the REER gap. - Financial excesses / credit gaps: - Measured as credit gaps; desirable credit gap P* should be zero in most cases. - Adjustments considered where credit gap estimates misstate financial imbalances (financial deepening or credit busts). - Presence of a credit policy gap does not necessarily imply a policy distortion—credit is endogenous and may fluctuate for other reasons. - Overarching principle: - Desirable policy settings aim to meet medium-term domestic objectives (fiscal sustainability and intergenerational equity; public health guided by domestic welfare and risk-sharing considerations; credit policies to avoid unsustainable booms and costly contractions). ### Standard Errors of Estimated Norms - The 2018 refinements introduce standard errors for estimated country-specific CA norms to reflect statistical uncertainty. - The CA norm is a linear function of regressors and coefficients; its standard error is computed as a linear combination of the variance-covariance matrix of estimated coefficients (assuming regressors fixed): - Standard error formula reported in source (equation (4)): sqrt(Var( X_a' β̂ + P*_a' γ̂ )) - Resulting standard errors reported with EBA CA norms to guide staff uncertainty ranges. - Staff may use larger uncertainty ranges than estimated standard errors if justified. ### Multilateral Consistency - EBA country sample covers more than 90 percent of global GDP; multilateral consistency is important. - Theoretically, global current accounts sum to zero and REER gaps should average to zero; most variables expressed as deviations from world averages helps ensure this. - Practical adjustments needed because: - Current account balances do not exactly sum to zero over EBA sample due to global statistical discrepancy and incomplete sample coverage (notably omission of some large net commodity exporters). - Some regressors are not expressed as deviations from world averages or have non-linear effects (e.g., lagged NFA-to-GDP and global risk factor). - Consistency procedure: - Uniform GDP-term adjustments to components of the current account decomposition so that across the EBA sample: (i) policy gaps add up to zero; and (ii) residuals add up to zero. - Cyclical component: relative output gaps sum to zero by construction; commodity terms-of-trade gaps do not, reflecting more commodity importers in sample. - EBA sample current account statistical discrepancy was about 0.4 percent of global GDP in 2017 and is mostly attributed to current account norms (except cyclical commodity price parts attributed to cyclical component). - REER models: - Weighted average of residuals must annually add to zero. - Real exchange rates are adjusted by the global weighted average of residuals using trade-weight eigenvector weights. - Consistency adjustment generally small (about 2¼ percent of the global residual in 2017). ### Current Account–REER Elasticities (key inputs and methodologies) - CA-REER semi-elasticity translates CA gaps into consistent REER gaps (allows comparison with REER models). - Semi-elasticity defined as change in (CA/GDP) per percent change in REER = η_TT, decomposed into contributions from nominal trade balance-to-GDP (eta_TT = η_X * s_X − η_M * s_M). - Two methodologies for estimating η_TT: 1. Original CGER approach (calibration-based). 2. CGER-inspired approach (panel regressions for export and import equations). - Staff can use alternative estimates or adjust for country-specific factors with justification. ### Original CGER Approach - Based on Isard et al. (2001) and IMF (1998). - Decomposition: η_TT = η_X * s_X − η_M * s_M, where η_X and η_M are elasticities of nominal exports/GDP and imports/GDP with respect to the REER; s_X and s_M are nominal export and import shares of GDP. - Original CGER calibrated η_X = 0.71 and η_M = 0.92 (assumed common across countries). - Semi-elasticities of nominal trade balance-to-GDP obtained using common calibrated η_X and η_M and country-specific export and import shares (excluding oil), over the 2013-2023 period. - Import and export elasticities common across countries; semi-elasticity varies with openness. ### CGER-Inspired Approach - Builds on original method but estimates η_X and η_M via panel regressions. - Uses an unbalanced panel of most EBA countries with quarterly data 1980Q1–2017Q4. - Dynamic export equation (log of X) relates exports to lags of exports, lags of REER, and trading partners’ real GDP (world demand); includes time and country fixed effects and up to eight lags. - Dynamic import equation (log of M) relates imports to lags of imports, lags of REER, and domestic real GDP (domestic demand); also includes fixed effects and up to eight lags. - Long-run elasticities computed as sums of β coefficients divided by (1 − sum of δ coefficients). - Panel estimation yields η_X = 0.11 and η_M = 0.57 (volume elasticities). - These volume elasticities are used in decomposition (equation (6)) with country-specific openness ratios s_X and s_M to calculate η_TT. - Trade shares computed on aggregate imports and exports (not excluding oil) in CGER-inspired approach to reflect increased substitutability between energy sources. ### Elasticity Estimates — Comparison and Use - Table 11 (in source) compares updated semi-elasticities from both approaches. - Semi-elasticity estimates from original CGER and CGER-inspired approaches are very similar—correlation of 0.9. - Suggested elasticities computed using same methodology for all countries; do not automatically correct for country-specific features (commodity export share, value-added trade, capacity constraints). - IMF country teams encouraged to explore more granular data and adjust estimates for country-specific factors with justification. ### The External Sustainability (ES) Approach - ES approach determines the CA-to-GDP ratio that stabilizes the NFA-to-GDP position over the medium term at a benchmark/desired level. - Complementary to CA and REER models; useful where debtor/creditor positions are excessively large and sustainability is central. - Strength: limited required variables and assumptions (medium-term GDP growth, inflation, rates of return). - Limitation: requires a normative benchmark for desirable NFA-to-GDP level; normative choice difficult except when large negative NFA suggests unambiguous need for higher NFA. - ES does not provide specific policy gaps (unlike EBA CA model) but can take precedence when external sustainability is primary concern. ### ES: Link Between Stock and Flow Positions (formulation and implications) - Law of motion of NFA: change in NFA = CA + valuation changes (VC). ES typically assumes VC = 0 for forward-looking stabilization. - Expressed in nominal GDP ratios: x_t − x_{t-1}/[(1+π_t)(1+g_t)] = ca_t (notation per source). - Decomposing CA into net exports and transfers (nxt) and net interest payments (assuming same real return r on assets and liabilities) leads to steady-state expressions for ca and nxt that stabilize NFA at a given lnNFA level (equation (11) in source). - Key implications: - For debtor economies (lnNFA < 0), lower growth requires higher CA or higher nxt to stabilize NFA at benchmark. - For debtor economies, higher rate of return on external positions (r > g) requires higher nxt to stabilize NFA. ### ES Implementation Steps in EBA Framework 1. Choose benchmark/desirable NFA-to-GDP level (often current/last year’s level; for large debtors select stronger external position informed by regional/country-group averages). 2. Derive CA that stabilizes NFA at benchmark using medium-term inflation (π) and potential growth (g) per equation (11). 3. Produce staff forecast of medium-term CA under current policies and REER; adjust for cyclical components using same methodology as EBA CA model. 4. Determine CA gap = staff medium-term CA under current policies − CA that stabilizes NFA at benchmark. 5. Calculate corresponding REER gap via staff-assessed REER–CA semi-elasticities (benchmarks in Section VI). - Note: CA that stabilizes NFA at benchmark is not a general CA norm; it is relevant when external sustainability is the central concern and may override EBA CA model results in such cases. ### Complementary Tools (addressing model residuals and measurement / structural issues) - Rationale: EBA models leave sizable residuals in some cases; complementary tools help disentangle residuals due to policy distortions from those reflecting country characteristics or measurement issues. - Two tools developed: 1. Measurement-bias adjustor tool to estimate potential biases in the measurement of the current account and inform outside-the-model adjustments. 2. Structural-policy tool to explore role of product and labor market distortions (for a subset of countries and years where data available). #### Measurement Issues - Increased international integration and activities of multinational corporations blur resident/non-resident boundaries and income attribution, raising potential biases in external statistics. - Two definitional departures noted (BPM6 transaction-based nominal recording vs. economic concept of accumulation of real net external wealth): - Retained earnings on equity investments: - BPM6 records retained earnings differently for direct vs. portfolio investment. For direct investment equity, retained earnings recorded in income; for portfolio equity, only paid dividends recorded in current account income and retained earnings reflected in IIP valuation changes. - Economic perspective could treat retained earnings as income in both cases. - Inflation component in interest income: - Investment income recorded in nominal terms; nominal interest includes inflation compensation, deviating from real income accumulation—leading to systematic bias in NIIP valuation changes. - Earlier EBA included a financial center dummy to capture measurement biases for a few economies; this was too restrictive because biases vary in sign and magnitude across countries and time. - 2018 refinements: - Removed financial center dummy. - Introduced outside-of-model adjustors for measurement issues when criteria met: - Empirical estimates of retained earnings and inflation measurement biases indicate sizable mismeasurement. - Evidence of systematic and persistent differences between a country’s financial account and changes in its NFA position. - Both point in same direction. - Magnitude of adjustment determined case-by-case considering data availability and that EBA NFA-to-GDP coefficient partially captures biases. - Data constraints remain important: stock-flow reconciliation and IIP currency exposure data limitations hinder granular estimation of measurement biases. #### Structural Tools - Structural distortions (product and labor markets) are important determinants of persistent CA imbalances but cannot be fully included in EBA regressions due to data limits. - Two-stage model-based complementary tool developed using subset of countries/years: - First stage: estimated EBA CA residuals regressed on structural policy variables S̃_{a j} (deviations from GDP-weighted world average) per equation (12): μ̂_{a j} = α + S̃_{a j} γ + ε_{a j}. - Structural indicators used: - For OECD countries: burdens in licenses and permits system (LPS) and employment protection laws (EPL). - For others: World Economic Forum (WEF) indicators — number of starting-a-business procedures (SBP) and degree of cooperation in labor-employer relations (CLER). - Selection criteria for indicators: robust empirical results and conceptual clarity on how indicator affects CA. - Empirical findings (consistent with literature): - Reducing burdens in LPS can lower the current account through higher investment by new firms and upward wage pressure reducing competitiveness. - Easing certain labor market rigidities (reducing EPL) can improve the current account by lowering labor costs and boosting competitiveness. - Normative exercise: - Assumes desired policy settings S* correspond to best-practice frontier (with some flexibility, especially for labor market benchmarks). - Tool estimates extent to which reforms could explain residuals and address external imbalances via (S − S*) γ. ### Concluding Remarks (limited explanatory power, data needs, and research directions) - Latest EBA refinements improved conceptual rigor but models still leave large residuals in many cases, reflecting structural distortions, measurement biases, and unmodelled fundamentals. - Continued methodological and data advances needed; complementary tools and staff judgment remain essential while data limitations persist. - Key data collection priorities to better incorporate aspects into EBA: - Labor and product market policies and distortions. - External sector statistics: currency composition of cross-border positions and stock-flow reconciliation. - Areas needing deeper exploration (future research directions): - Sectoral distribution of net saving and role of corporates in persistent CA surpluses in key advanced and some emerging economies. - Measurement biases from expansion of multinational activities. - Risks from stock imbalances (size and composition) and their effect on external assessments. - Interactions between aging and design/sustainability of public pension systems. - Role of exchange rates as adjustment mechanism given U.S. dollar dominance and changing trade landscape (increased intermediate goods trade and global value chains). *Italic: Source — wpiea2019065, Section VIII and related sections, as provided in the supplied content.* ### REFERENCES ### wpiea2019065 - REFERENCES (Tables and Empirical Findings) ### EBA Current Account Regression — Main Results (Table 1) - Dependent variable: CA/GDP. - Cyclical factors: - Output gap #: -0.356*** (0.000). - Commodity terms of trade gap (interacted with trade openness): 0.161*** (0.000). - Macroeconomic fundamentals: - Net foreign asset (NFA) position (lagged): 0.023*** (0.000). - Net foreign asset (NFA) position times dummy if NFA/GDP < -60 percent (lagged): -0.006 (0.593). - Output per worker (lagged): 0.023 (0.259). - Expected real GDP growth 5 years ahead #: -0.302*** (0.004). - Reserve currency status: -0.030*** (0.009). - Structural fundamentals: - Old-age dependency ratio (OAD) #: -0.069 (0.109). - Population growth #: -0.692* (0.061). - Share of prime-aged savers #: 0.138** (0.013). - Life expectancy #: -0.005*** (0.000). - Interaction between life expectancy #, and future OAD: 0.013*** (0.005). - Institutional quality #: -0.047** (0.015). - Exhaustible resources of oil and natural gas: 0.310*** (0.000). - Policy variables: - Fiscal policy #: 0.329*** (0.000). - Health spending (lagged) #: -0.399*** (0.003). - Foreign exchange intervention (interacted with capital controls) #: 0.754*** (0.001). - Credit gap #: -0.104*** (0.000). - Capital controls interactions: - Output per worker (lagged, interacted with capital openness): 0.041* (0.051). - Global risk aversion (lagged, interacted with capital openness): 0.020 (0.190). - Global risk aversion (lagged, interacted with capital openness and reserve currency): 0.002 (0.971). - Constant: -0.009*** (0.002). - Sample and fit: - Observations: 1,367. - Number of countries: 49. - R-squared IV: 0.524. - R-squared Fit: 0.550. - Root MSE: 0.031. - Notes: - Standard errors in parenthesis. * significant at 10%; ** significant at 5%; *** significant at 1%. - # sign means variable is included in differences from (GDP-weighted) world counterpart. - Capital account openness is calculated as one minus capital control index. ### EBA Current Account Regression — 2013 and 2015 Model Comparisons (Table 2) - 2013 Model (Original Data / Updated Data) and 2015 Model reported key coefficients and significance: - Output gap #: -0.400***, -0.385***, -0.392*** (p-values all (0.000)). - Commodity terms of trade gap (interacted with trade openness): 0.230***, 0.197***, 0.139*** (all (0.000)). - Net foreign asset (NFA) position (lagged): 0.016**, 0.015**, 0.014** ((0.019)(0.016)(0.016)). - Expected real GDP growth 5 years ahead #: -0.471***, -0.425***, -0.272*** ((0.000)(0.000)(0.009)). - Reserve currency status: -0.045***, -0.041***, -0.038*** ((0.000)(0.000)(0.001)). - Financial center dummy (where present): 0.033***, 0.027***, 0.028*** ((0.000)(0.000)(0.000)). - Exhaustible resources of oil and natural gas: 0.615***, 0.410***, 0.398*** ((0.000)(0.000)(0.000)). - Fiscal policy #: 0.324***, 0.470***, 0.543*** ((0.001)(0.000)(0.000)). - Health spending (lagged) #: -0.551***, -0.503***, -0.310** ((0.000)(0.000)(0.022)). - Foreign exchange intervention (interacted with capital controls) #: 0.346**, 0.449**, 0.261 ((0.040)(0.024)(0.173)). - Demeaned credit #: -0.026***, -0.021***, -0.038*** ((0.002)(0.005)(0.000)). - Sample sizes and fit: - Observations reported across specifications (e.g., 1080/1,197/1,340). - Number of countries: 49 in each specification. - R-squared IV values: 0.52 / 0.544 / 0.511. - Root MSE around 0.032–0.033. ### Robustness Checks — Institutions, Credit, Demographics, FX Intervention (Tables 3–6) - Robustness on Institutions (Table 3): - Coefficients remain broadly stable relative to baseline (e.g., Output gap #: -0.356*** to -0.363***). - Institutional quality variations yield coefficients: Full ICRG #: -0.047** (0.015); Full ICRG through 2001, zero after #: -0.059*** (0.005); WGI after 2001, zero before 2002 #: -0.033* (0.052); WGI and ICRG average #: -0.045* (0.077); WGI from 2002; base pre 2002 on ICRG/WGI relation #: -0.038** (0.049). - Observations: 1,367; Number of countries: 49; R-squared IV: 0.524–0.528; R-squared Fit: 0.546–0.550; Root MSE: 0.031. - Robustness on Credit (Table 4): - Credit specifications (Credit gap Lambda 1600/25000, demeaned, detrended, growth) influence coefficients but key signs persist. - Credit gap (Lambda 1600) #: -0.104*** (0.000); Credit gap (Lambda 25000) #: -0.092*** (0.000). - Demeaned credit #: -0.035*** (0.000); Credit-to-GDP change, current #: -0.107*** (0.000). - Observations: 1,367 across variants; Number of countries: 49; R-squared IV: 0.485–0.525; Root MSE 0.031–0.033. - Robustness on Demographics (Table 5): - Variants replacing life expectancy (LE) with aging speed or including LE squared produce modest changes. - Life expectancy #: -0.005*** (0.000) in baseline; alternative specifications show varied significance (e.g., LE squared 0.0003* (0.079)). - Population growth #: baseline -0.692* (0.061); other specs show similar negative signs with varying significance. - Observations: 1,367; Number of countries: 49; R-squared IV: 0.511–0.524. - Robustness on Foreign Exchange Intervention (Table 6): - Baseline FX intervention (interacted with capital controls) #: 0.754*** (0.001). - Alternative FXI definitions and IV specifications produce a range: - No FXI refinements: FXI (interacted) 0.276*** (0.000). - Narrow FXI: 0.964*** (0.001). - Narrow FXI with old IV: 0.195 (0.295). - Observations: mainly 1,367; Number of countries: 49; R-squared IV: 0.524–0.534; Root MSE: 0.031. ### REER Models — Regression Results and Diagnostics (Tables 7–10) - REER Models (Table 7) — Index and Level specifications: - Dependent variable: REER (+=appreciation). - Index model highlights: - Commodity terms of trade (log index): 0.184*** (0.000). - Net foreign assets (NFA) position (lagged): -0.109*** (0.000). - Output per worker (log, lagged): 0.217*** (0.000). - Expected real GDP growth 5 years ahead †: 2.012*** (0.003). - Institutional quality †: 0.653*** (0.000). - Trade openness (lagged) †: -0.208** (0.030). - Share of administered prices in CPI †: -1.713*** (0.000). - Real interest rates * capital account openness †: 0.697*** (0.004). - Health spending (lagged) †: 2.040** (0.015). - Foreign exchange intervention (interacted with capital control) †: -2.479* (0.078). - Level model highlights: - Commodity terms of trade (log index) not applicable; commodity log level interacted with trade openness: 0.064*** (0.002). - Net foreign assets (NFA) position (lagged): 0.056*** (0.003). - Output per worker (log, lagged): 0.171*** (0.000). - Expected real GDP growth 5 years ahead †: 1.961** (0.048). - Reserve currency status: -0.357*** (0.000). - Institutional quality †: 0.653*** (0.000) (index and level both show strong institutional quality effects). - Sample and fit: - Observations: 1,004 (Index), 990 (Level). - R-squared: 0.548 (Index), 0.895 (Level). - RMSE: 0.089 (Index), 0.146 (Level). - Number of countries: 40 (Index), 39 (Level). - † sign means variable is included in differences from (trade weighted) world counterpart. - REER Models — 2015 Models (Table 8): - Index and Level 2015 (Original Data / Updated Data) show: - Commodity terms of trade (log index): 0.092* / 0.128*** ((0.063)(0.005)). - Net foreign assets (NFA) position (lagged): 0.105*** / 0.069*** ((0.000)(0.000)). - Output per worker (lagged): 0.698*** / 0.614*** (Index original/updated) and 0.164*** / 0.173*** (Level original/updated). - Old-age dependency ratio (OAD) †: 0.911*** / 0.720*** (Index originals) and positive significant OAD in Level models. - Share of administered prices in CPI †: consistently large negative coefficients (e.g., -2.124***, -2.142***, -2.543***, -2.644***). - Foreign exchange intervention (interacted with capital control) †: negative and significant in several specifications (e.g., -1.731***, -1.204**, -2.097*, -2.104 (p-values vary)). - High fits reported for Level models: R-squared up to 0.905 and 0.897; RMSE 0.140–0.144. - Unit root and cointegration diagnostics (Tables 9–10): - Fisher test and Hadri test reported for panels; mixed results across variables (examples): - REER-Index: Fisher test -2.29**; Hadri test 26.30***; unit root diagnostic described as "inconclusive." - Reserve currency status: Fisher 0.55; Hadri 9.12***; unit root? "yes". - FXI*capital controls (instrumented, rel. TRD PTR): Fisher -3.56***; Hadri 27.17***; unit root diagnostic "inconclusive". - Cointegration tests (Table 10): - Model residual panel unit root tests (based on fitted residuals) show Dickey-Fuller and modified Dickey-Fuller statistics that reject unit roots: - Dickey-Fuller (REER-Index model): -5.35***. - Modified Dickey-Fuller (REER-Index model): -6.43***. - REER-Level model Dickey-Fuller/statistics likewise show significance (e.g., -3.52***, -3.33***). - Fisher unit root test for residual: -3.20*** (REER-Index residual) and -4.10*** (REER-Level residual). - The alternative hypothesis is that all panels are cointegrated. ### CA to REER Elasticities (Table 11) - Country-specific CA-to-REER elasticities reported for two approaches: - Original CGER and CGER-inspired Approach 2/. - Select values (Original CGER / CGER-inspired Approach 2/): - Argentina: -0.13 / -0.11. - Australia: -0.20 / -0.15. - Austria: -0.45 / -0.36. - Belgium: -0.72 / -0.58. - China: -0.23 / -0.12. - Germany: -0.38 / -0.28. - Ireland: -0.92 / -0.65. - Netherlands: -0.74 / -0.53. - Switzerland: -0.53 / -0.39. - Thailand: -0.64 / -0.41. - United States: -0.12 / -0.10. - Notes: - 1/ Figures shown are those previously used in the CGER exercise, in most cases using a common elasticity assumption of 0.71 for exports and 0.92 for imports. - 2/ Based on newly estimated common elasticities using quarterly data (1980Q1-2017Q4), adjusted by the size of exports and imports in GDP; 0.11 for exports and 0.57 for imports. ### Effect of Product and Labor Market Policies on EBA Current Account Model Residuals (Table 12) - Regressions using OECD and WEF indicators: - Specification (1) (OECD): - PMRs: LPS (+ = more burdens): 0.0049**. - Observations: 374. - R-squared: 0.026. - Number of countries: 24. - Specification (2) (WEF): - LMRs: EPL (+ = stricter regulations): -0.0048**. - PMRs: SBP (+ = more procedures): 0.0242**. - LMRs: CLER (+ = more cooperation): 0.0508***. - Observations: 533. - R-squared: 0.053. - Number of countries: 49. - Significance codes: * significant at 10%; ** significant at 5%; *** significant at 1%. - Source for this table: OECD, WEF, and Staff Estimates. *Italic source attribution: wpiea2019065 - REFERENCES (IMF Working Paper tables and bibliographic references).* ### Annex I: Comparison of EBA Current Account Model Specifications Across Vintages ### Annex I: Comparison of EBA Current Account Model Specifications Across Vintages ### Model specification changes (Annex I) - Variables whose specification changed since 2013 include: - Aging speed: defined as the 20 year ahead change in old age dependency ratio (OADR). - Prime-age savers share: used in updated specifications. - Life expectancy and interactions: life expectancy * future OADR and life expectancy * future OADR included in later vintages. - Institutional controls expanded from ICRG-5 subindices to ICRG-All 12 subindices. - Financial center dummy: applied only to Switzerland and the Netherlands. - Credit measure evolution: from demeaned Credit-to-GDP to detrended Credit-to-GDP with HP filter. - FX intervention: from spot FX intervention to spot and derivative FX intervention. - Complementary tools: - One tool estimates magnitude of retained earnings and inflation biases. - Another tool estimates whether product and labor market distortions explain model residuals. - Instrumentation and model form: moved toward simplified instrumentation and more structural approaches in later vintages. *Note: Aging speed is defined as the 20 year ahead change in old age dependency ratio (OADR). Financial center dummy applied only to Switzerland and the Netherlands.* ### Data sources and dependent variables (Annex II) - Current account and related regressors: - Current Account: IMF WEO. - Net Foreign Assets (NFA) position: EWN: Lane, Milesi-Ferretti. - Output per worker, relative to top 3 economies: IMF WEO. - Output per worker interacted with capital openness: IMF WEO and Quinn Database. - Expected real GDP growth 5 years ahead: IMF WEO. - Output Gap: IMF WEO. - Commodity Terms of Trade Gap (interacted with trade openness): IMF WEO and World Bank, WITS. - Reserve Currency Status: IMF, COFER. - Global Financial Conditions (interacted with capital controls): Haver, and Quinn Database. - Old-age dependency ratio (OAD), Population Growth, Share of prime-aged savers, Life expectancy, Life expectancy at prime age (interacted with future OAD): UN World Population Prospects, 2017 Revision. - Institutional quality: PRS Group, International Country Risk Guide (ICRG). - Exhaustible resources of oil and natural gas: IMF WEO, World Bank WITS, and British Petroleum Statistical Review of World Energy. - Fiscal Policy: IMF WEO. - Health Spending: OECD and WDI; ADB, CEPAL, and IMF Fiscal Affairs Department database in some cases. - Foreign Exchange Intervention: IMF WEO; EWN: Lane, Milesi-Ferretti; IMF Data Template on International Reserves and Foreign Currency Liquidity; Quinn Database. - Credit Gap: BIS Credit Statistics; World Bank Global Financial Development Database. - REER and level inputs: - Real Effective Exchange Rate: IMF INS. - Share of administered prices: EBRD Structural Change Indicators. - Financial home bias: BIS Debt Securities Statistics. - Price Level: World Bank, International Comparison Program, 2011. - Capital stock per employed person: Penn World Table sources. - VAT revenue: OECD's Revenue Statistics Dataset, Council of State Governments (USA), Bureau of Economic Analysis. *Note: Dependent variables in each model are in bold in source tables.* ### Variable definitions and measurement conventions (Annex III) - NFA-to-GDP: - Enters directly (not relative) in level terms. - Interacted with dummy equal to one if NFA position < negative 60 percent of GDP. - NFA data: updated and extended Lane and Milesi-Ferretti (2007) EWN dataset. - Output per worker, relative to top 3 economies: - Ratio of PPP GDP to working age population relative to Germany, Japan, and U.S., demeaned. - Interacted with capital account openness. - Oil and gas trade balance, adjusted for “temporariness”: - Net oil and gas balance (five-year moving average, percent of GDP) multiplied by relative temporariness = (current extraction / proven reserves) relative to Norway in 2010. - Proven reserves: British Petroleum Statistical Review of World Energy (location based). - Population growth: annual growth rate of total population. - Old age dependency ratio: population aged over 65 divided by population aged 30-64. - Share of prime-aged savers: current share of ages 45-64 as proportion of total working-age population (ages 30-64). - Life expectancy: life expectancy of a current prime-aged saver. - Interaction: life expectancy × future OAD, where future OAD = average OAD 15, 20 and 25 years ahead. - 5-year growth forecast: IMF WEO 5-years ahead real GDP growth (proxy for potential output growth). - Public health spending/GDP: OECD definition when available; cross-source definitional differences noted. - Global risk aversion proxy (VXO): index of implied U.S. stock market volatility based on S&P 100; annual average during the sample period is 0.2. - Own currency share in world reserves (exorbitant privilege): - 2017 shares: US dollar 67 percent; euro 20 percent; yen 4.7 percent; sterling 4.4 percent. - Enters alone and interacted with VIX and capital account openness index. - Output gap: IMF staff estimates in WEO; where unavailable, HP filtered WEO actual and projected data for 1980-2023 used. - Commodity terms of trade gap (CA model): - Ratio of geometric weighted-average price of 43 commodity export categories to 43 commodity import categories, each relative to manufactured goods prices in advanced economies; weights by commodity shares. - Extended using IMF commodity projections, HP filtered country-specific, zero country-specific mean; interacted with trade openness in CA model. - Cyclically-adjusted fiscal balance (instrumented): - Based on IMF staff cyclical-adjustment estimates or country-specific regression residuals. - Instrumented with lagged global variables and country-specific factors; first-stage controls include GDP per capita, exchange rate regime, democracy ranking, lagged output gap, and average cross-sectional fiscal balance. - Capital controls index: - Quinn index on overall private-sector capital controls, scaled 0 (no controls) to 1 (full controls). - 2017 mean across countries: 0.16; range: 0 to 0.75. - Foreign exchange intervention (instrumented) (FXI): - Change in central bank foreign exchange reserves including off-balance sheet FX intervention during the year, scaled by nominal GDP in U.S. dollars. - Instrumented with difference between M2/GDP and reserves/GDP, reserve accumulation and emerging market/developing economies indicator, all interacted with capital controls index. - Real interest rate: nominal short-term interest rate minus annual inflation rate. - Private credit-to-GDP gap: - Credit-to-GDP ratio detrended using one-sided Hodrick-Prescott filter. - Total credit to non-financial private sector excluding non-bank cross-border flows from BIS. - Institutional quality: - 12 ICRG sub-indicators normalized between 0 and 1 (higher = less risk): government stability; internal conflict; external conflict; military in politics; law and order; ethnic tensions; bureaucracy quality; socioeconomic conditions; investment profile; corruption; religious tensions; democratic accountability. - Financial home bias: share of domestic debt owned by residents (BIS). - REER Levels transformation: - Log REER Index rebased to year 2011; add log of PPP relative prices where 2011 US price level equals 1. - Capital stock per employed person: capital stock at constant 2005 national prices divided by total employed population. - VAT revenue: VAT revenue as percent of GDP. - Ratio of Traded/Non-Traded Sector Productivity: - Estimated using sectoral labor productivities at 2005 USD PPP from Mano and Castillo (2015); missing values imputed via OLS fitted values or extended via WDI sectoral productivity changes. - Share of administered prices: - Relevant for few transition economies; zero assigned for most countries. ### Country coverage (Annex IV) - Countries included in the current account model (full list in source): Argentina*, Australia, Austria, Belgium, Brazil, Canada, Chile, China, Colombia, Costa Rica*, Czech Republic, Denmark, Egypt*, Finland, France, Germany, Greece, Guatemala*, Hungary, India, Indonesia, Ireland, Israel*, Italy, Japan, Korea, Malaysia, Mexico, Morocco*, Netherlands, New Zealand, Norway, Pakistan#, Peru, Philippines, Poland, Portugal, Russia, South Africa, Spain, Sri Lanka*, Sweden, Switzerland, Thailand, Tunisia*, Turkey, United Kingdom, United States, Uruguay*. - Notation: - Asterisks (*) denote countries not included in REER index regression. - Sharps (#) denote countries not in REER levels regression for data availability reasons. - Note: All countries above are included in the current account model. ### Demographic block and implications (Annex V) - Structural modeling approach: - Multi-country overlapping generations model with population growth (fertility and migration) and age-specific mortality. - Two-country extension to analyze capital flows and unfunded old-age transfer schemes. - Key mechanisms and model predictions: - Static effect: age composition (share young, OAD, share of prime savers) affects aggregate savings. - Dynamic effect: old-age survival risk (life expectancy of prime-aged saver) drives household savings; higher longevity ⇒ higher savings. - In open economy, higher longevity relative to world average and higher share of prime-aged savers relative to world average ⇒ net capital export (saving more and investing abroad). - Current EBA demographic specification (Table A.1 comparison): - 2015 EBA static effects: Old age dependency (OAD) ratio (ages 65+/30-64); Population growth. - 2018 EBA static effects: OAD (ages 65+/30-64); Population growth; Current share of prime savers (ages 45-64) as proportion of working-age population (30-64). - 2015 EBA dynamic effects: Interaction of relative aging speed (20-year ahead change in OAD) with current OAD; Interaction of relative current OAD with aging speed. - 2018 EBA dynamic effects: Life expectancy of a current prime-aged saver; Interaction of life expectancy with future old-age dependency. - Empirical findings and magnitudes: - The combined impact of demographic variables explains about 15 percent of the cross-country variation in the current account over the last five years. - The revised demographic specification explains a larger fraction of the unexplained component of the current account after accounting for non-demographic regressors. *Annex references: aging speed defined as 20 year ahead change in OADR; future OAD is moving average of OAD 15 to 25 years ahead.* *Source: IMF working paper annexes (Annex I–V).* ### Annex VI: Estimation of Exhaustible Resources in the EBA Models ### Annex VI: Estimation of Exhaustible Resources in the EBA Models ### Purpose and conceptual approach - The EBA model treats income from exhaustible resources as potentially temporary; intergenerational equity implies saving a fraction of such income for future generations, with the saved fraction increasing with the temporariness of the resource. - The exhaustible-resources income variable applies only to net exporters of oil and gas and captures: - (i) the temporariness of the resource; and - (ii) the extraction size (relative to GDP). - Temporary movements in oil and gas prices are controlled for by the detrended commodity terms of trade. ### Variable construction and interpretation - The oil and natural gas trade balance variable is defined as the 5-year moving average of net exports for oil and natural gas, relative to GDP, multiplied by a temporariness index for oil and gas. - The first component captures the average oil and gas extraction size (X_{k,i,s} / Y_{i,s} as expressed in the source). - The second component captures the temporariness of the resource relative to this measure for Norway in 2010. - The temporariness term (“temp”) is defined as the ratio of current extraction to proven reserves (or the inverse of “years-till-exhaustion” from the British Petroleum Statistical Review of World Energy). - Higher values for the temporariness term indicate the resource is more temporary (expected to be exhausted sooner). - Implication: the more temporary the resource revenues and the larger the extraction (relative to GDP), the larger the positive effect on the current account balance. ### Key technical points - The oil and natural gas trade balance uses a 5-year moving average of net exports (relative to GDP). - Temporariness is benchmarked relative to Norway in 2010. - Short-term price fluctuations are isolated via detrended commodity terms of trade to focus the variable on structural/exhaustible aspects rather than transitory price shocks. --- ### Annex VII: Benchmark for Public Health Spending in the EBA Models ### Rationale - Public expenditure on health (share of GDP) is treated as a social protection policy that can influence the national saving rate. - Higher public health spending reduces households’ need for precautionary saving; the expected coefficient in the current account regression is negative (higher public health expenditure → decline in current account via decrease in national saving). ### Benchmark (P*) specification - The desired level of public health spending to GDP (P*) is based on a benchmark that varies with country characteristics. - A cross-sectional regression (average of 2005-2016) links public health spending (share of GDP) to structural determinants: - Income: log difference between countries’ PPP-based GDP per capita and the world average. - Demographics: old age dependency ratio (65+/30-64). - Inequality: gross income concept Gini coefficient (from the Standardized World Income Inequality Database for 2013). - The fitted values from the regression serve as indicative benchmarks; IMF staff may set different desired policy levels based on country-specific considerations. ### Regression results (Table A.2) - Relative Income: 0.018*** - Old age dependency ratio: 0.096*** - Gini coefficient: 0.0175 - Constant: 0.021* - Observations: 49 - R-squared: 0.835 - Robust standard errors in parentheses - Significance notation: *** p<0.01, ** p<0.05, * p<0.1 ### Interpretation and policy guidance - All variables have the expected sign and are statistically significant, with the exception of the inequality proxy which lost significance following data revisions and sample extension. - The benchmark provides an evidence-based reference for desired public health spending, but staff judgment may adjust P* for well-justified country-specific reasons. *Source: Annex VI and Annex VII from the provided IMF content unit.* --- _Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019065.pdf_