## Annex I.

## Source details

**Canonical URL:** [Annex I.](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022106-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022106-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022106-print-pdf.pdf.json)

---

### A. Stylized facts and component importance
- CAB identity: CAB = TB + IB; IB = PIB + SIB.
- Relative sizes and medians (2017–2019 averages, sample statistics):
  - Median gross trade = 79 percent of GDP.
  - Median gross income = 16 percent of GDP.
  - Global annual income credits (2017–2019 average, sample of 168 countries) = $5.7 trillion.
  - Exports (goods and services) = $23.3 trillion.
  - Median TB = -3 percent of GDP.
  - Median IB = +0.5 percent of GDP.
- Cross-country IB vs TB:
  - In 37 countries, |IB| / |TB| > 1.
  - In 57 countries, 1 > |IB| / |TB| > 0.5.
- Table 1 selected summary statistics (2017–2019 averages, percent of GDP):
  - Income Balance (IB): Obs 169; Mean 3.54; Std 13.63; 25% Prct. -2.22; Median 0.54; 75% Prct. 6.42.
  - IB Debit: Obs 169; Mean 15.78; Std 64.58; 25% Prct. 4.02; Median 6.94; 75% Prct. 10.54.
  - IB Credit: Obs 168; Mean 20.25; Std 71.57; 25% Prct. 4.54; Median 8.03; 75% Prct. 15.06.
  - PIB: Obs 169; Mean -2.99; Std 20.94; 25% Prct. -4.34; Median -2.32; 75% Prct. 0.1.
  - IIB (Investment PIB): Obs 166; Mean -4.52; Std 24.33; 25% Prct. -5.00; Median -2.71; 75% Prct. -0.7.
  - SIB: Obs 169; Mean 6.53; Std 22.15; 25% Prct. -0.39; Median 2.37; 75% Prct. 7.57.
  - Trade Balance (TB): Obs 169; Mean -8.38; Std 28.97; 25% Prct. -13.94; Median -3.1; 75% Prct. 1.98.
  - CA: Obs 169; Mean -5.38; Std 35.64; 25% Prct. -6.4; Median -2.65; 75% Prct. 1.17.
  - Gross Income (debit + credit): Obs 168; Mean 36; Std 135.44; 25% Prct. 9.97; Median 16.48; 75% Prct. 22.74.
  - Gross Trade (Export + Import): Obs 169; Mean 139.19; Std 614.05; 25% Prct. 57.03; Median 78.59; 75% Prct. 107.
- Additional stylized facts:
  - Primary income flows are four times larger than secondary income flows in global dollar terms; median PIB = -2.3 percent of GDP; median SIB = +2.4 percent of GDP.
  - AEs on average net payers of both primary and secondary income; Japan and some other G7 countries have positive PIBs.
  - Developed country average gross investment income = 14.7 percent of GDP; EMs = 5.5 percent of GDP; LICs = 4.3 percent of GDP.
  - Only 25 percent of countries are net receivers of primary income; about 70 percent are net receivers of secondary income.
  - Remittances make up 80% of secondary income inflows in the median country.

### B. NIIP (IIP) links and composition implications
- Empirical link: a 10 percentage-point rise in net foreign assets as a share of GDP associated with a rise in the income balance of 0.2 percent of GDP (Colacelli et al., 2021).
- Typical IIP patterns:
  - Sample of 50 AEs and EMs (Bénétrix et al., 2019): median NFA = –21 percent; mean NFA = –4 percent.
  - EMs: liabilities exceed assets by about one-third.
  - AEs more globally financially integrated in gross terms.
- Crisis risk by liability type:
  - Higher net foreign liabilities raise external crisis probability, especially when liabilities are debt-heavy.
  - Net foreign portfolio equity liabilities have weaker effects on external crisis probabilities; FDI liabilities tend to reduce crisis risk.
  - Foreign-currency debt liabilities increase likelihood of stress episodes; equity liabilities do not.
- Currency composition (reported stock FC shares):
  - Sample assets FC share: mean 83%; median 96%.
  - Sample liabilities FC share: mean 36%; median 33%.
  - Debt stock FC share: assets mean 79%; median 95%; liabilities mean 66%; median 77%.
  - AE median FC share of foreign assets ≈ 90%; FC share of foreign liabilities ≈ 30 percent.
  - EMs: FC share of foreign assets ≈ 100 percent; FC share of foreign liabilities ≈ 40 percent.
- Depreciation implications:
  - For many countries, depreciation mechanically increases local-currency income receipts from foreign assets more than payments on foreign liabilities, implying an IIB improvement following a depreciation.
  - AE median FC NFA = 74 percent of GDP; EM median FC NFA = 16 percent of GDP.
  - For 80 percent of countries in the sample, the NFA would mechanically improve following a depreciation because they have a positive FC NFA.

### III. Correlation and variance decomposition of TB and IB
- Within-country correlations (time series):
  - Aggregate sample: Cor(tb,ib) mean = -0.35; median = -0.42.
  - Correlations statistically significant and negative in 60 percent of countries.
  - By income group (Cor(tb,ib) mean / median):
    - AEs: -0.10 / -0.11.
    - EMs: -0.41 / -0.48.
    - LICs: -0.43 / -0.49.
  - Partial and first-difference correlations indicate AEs have near-zero pretrend-adjusted correlations (partial Cor(tb,ib)|trend: AEs 0.01 mean / -0.06 median).
- Variance decomposition (Var terms expressed as percent of Var(CA)):
  - Var(ib)/Var(ca)*100: All mean = 59; median = 28. Advanced mean = 46; median = 19. Emerging mean = 49; median = 27. Low-Income mean = 87; median = 58.
  - Var(tb)/Var(ca)*100: All mean = 143; median = 113.
  - 2Cov(tb,ib)/Var(ca)*100 (covariance share): All mean = -98; median = -37. Advanced mean = -49; median = -11. Emerging mean = -93; median = -43. Low-Income mean = -145; median = -70.
  - Aggregate interpretation:
    - Average covariance share = -98 percent implies CA variance is cut in half relative to zero-correlation counterfactual.
    - Median country covariance share = -37 percent implies CA variance is cut by 27 percent.
    - In 80 percent of countries the covariance term is negative.
- Gross-component covariance decomposition (values as "% Var(CA)"):
  - Cov(cr,ex): All 7.18 (58.4%); Advanced 14.79 (115.6%); Emerging 3.82 (28.3%); Low-income 8.29 (70.3%).
  - Cov(db,im): All 8.17 (55.9%); Advanced 14.53 (128.6%); Emerging 5.20 (26.0%); Low-income 8.56 (62.5%).
  - Cov(db,ex): All 11.99 (66.4%); Advanced 19.77 (125.9%); Emerging 11.12 (47.9%); Low-income 8.83 (59.6%) — sign listed as negative in source decomposition.
  - Cov(cr,im): All 16.02 (95.3%); Advanced 16.04 (135.2%); Emerging 10.64 (41.4%); Low-income 24.54 (157.0%) — sign listed as negative in source decomposition.
  - Cov(ib,tb) = -12.67 (-47.5%) All; Advanced -6.50 (-16.9%); Emerging -12.74 (-35.0%); Low-income -16.51 (-83.8%).
- Interpretation:
  - Negative contributors Cov(cr,im) and Cov(db,ex) drive overall negative Cov(ib,tb).
  - Cov(cr,im) particularly important in LICs (driven by secondary income); Cov(db,ex) important in AEs and EMs (driven by primary income).
  - Positive Cov(cr,ex) and Cov(db,im) consistent with receipts financing consumption/imports and profits from exports leading to income debits.

### IV. IB behavior in crisis episodes — methodology and conditional means
- Crisis definitions and sample:
  - Crisis types: external, fiscal, financial, real (Vulnerability Exercise definitions).
  - Sample: 177 countries; average of 25 new crises per year; crises last 2 years on average.
- Estimator and identification:
  - Use FEct (two-way fixed effects counterfactual estimator) with diagnostics and factor-augmented FEct when parallel trends rejected.
  - Placebo tests hide P = 3 pre-crisis years to test pretrends.
- Conditional mean results (All countries, any crisis):
  - IB relative to tranquil benchmark:
    - Before crisis: -0.709 percent of GDP lower than normal (p-value 0.007).
    - During crisis: -1.017 percent of GDP lower than normal (p-value 0.000).
    - After crisis (three years after): -0.309 percent of GDP lower than normal (p-value 0.198).
  - TB relative to tranquil benchmark:
    - Before crisis: -0.89 percent of GDP lower than normal (p-value 0.028).
    - During crisis: 0.078 percent of GDP (p-value 0.853).
    - After crisis: -0.116 percent of GDP (p-value 0.804).
- Dynamics and component responses:
  - Net FDI income (All Any): Before 0.504 (p-value 0.111); During 0.53* (p-value 0.072); After 0.379** (p-value 0.050).
  - Net debt income (All Any): Before -0.153 (p-value 0.213); During -0.35* (p-value 0.062); After -0.098 (p-value 0.187).
  - Interpretation:
    - IB already lower pre-crisis, deteriorates during crisis, and remains no better in aftermath — IB often destabilizing during crises.
    - Net debt income deteriorates before and during crises; net FDI income increases on the eve of crises and after, consistent with FDI’s shock-absorbing properties.
    - Portfolio equity income changes small and insignificant in reported results.

### IV. Crisis heterogeneity and causal (factor-augmented ATT) estimates
- Crisis-type heterogeneity (selected estimates):
  - Financial crises (All sample):
    - IB: Before -0.729* (p-value 0.073); During -1.521*** (p-value 0.002); After -1.32*** (p-value 0.005).
    - TB: Before -1.205** (p-value 0.046); During 1.131* (p-value 0.053); After 1.093* (p-value 0.075).
  - External crises (All sample):
    - IB: Before -0.856** (p-value 0.026); During -1.038* (p-value 0.086); After -0.144 (p-value 0.649).
- Country group differences:
  - AEs and EMs (Any crisis):
    - IB: Before -1.193*** (p-value 0.000); During -1.595*** (p-value 0.000); After -0.372 (p-value 0.238).
    - TB: Before -1.404** (p-value 0.011); During 0.355 (p-value 0.464); After -0.03 (p-value 0.950).
  - LICs (Any crisis):
    - IB: Before -0.517 (p-value 0.389); During -0.476 (p-value 0.491); After 0.023 (p-value 0.976).
    - TB: Before -0.489 (p-value 0.288); During -0.133 (p-value 0.762); After -0.372 (p-value 0.379).
- Factor-augmented ATT estimates (Table 8, p-values from block bootstrap with 500 iterations):
  - All countries (Any crisis):
    - IB ATT: 0.303 (p-value 0.641).
    - Net FDI Income ATT: 0.53* (p-value 0.072).
    - Net Debt Income ATT: -0.35* (p-value 0.062).
  - AEs and EMs (Any crisis):
    - IB ATT: -1.519*** (p-value 0.004).
    - Net Debt Income ATT: -0.493** (p-value 0.037).
  - LICs (Any crisis):
    - IB ATT: 0.915* (p-value 0.090).
    - SIB ATT: 0.581 (p-value 0.328) (net debt income: insufficient observations in some rows).
- Interpretation:
  - Causal effects differ by income group: AEs and EMs show large negative causal IB effect during crises (about -1.5 percent of GDP relative to no-crisis counterfactual); LICs show positive IB ATT (0.915*), possibly due to secondary income dynamics.
  - Net FDI income ATTs often positive, consistent with stabilizing role of FDI; net debt income ATTs negative and concentrated in AEs and EMs.

### V. Exchange-rate semi-elasticities and REER-gap implications
- Semi-elasticity construction:
  - IB semi-elasticity η_II obtained by estimating elasticities for income receipts and payments separately and combining panel-wide estimates with country-specific income shares to derive country-specific semi-elasticities.
  - CAB semi-elasticity η_CA = η_TI + η_II, where η_TI is TB semi-elasticity.
- Panel regression highlights (sample: unbalanced panel of 145 countries, 1985–2019):
  - Lagged dependent-variable dynamics (Y(t-1) coefficients):
    - All Countries IB Credit: 0.84*** (0.05).
    - All Countries IB Debit: 0.77*** (0.04).
  - log(REER(t)) contemporaneous coefficients (selected):
    - All Countries IB Credit: -0.89*** (0.11).
    - All Countries IB Debit: -0.87*** (0.1).
    - EMs IB Debit: -0.47*** (0.13).
    - Primary Income Credit: -0.87*** (0.16).
- Elasticity estimates (Table 10, selected):
  - Panel A All Countries:
    - μ_credit = -0.63 (0.74).
    - μ_debit = -0.74*** (0.28).
  - Panel B by income group:
    - Advanced: μ_credit 0.04 (0.45); μ_debit -0.29 (0.83).
    - Emerging: μ_credit -1.28*** (0.36); μ_debit -0.96*** (0.28).
    - Low-income: μ_credit -0.28 (0.46); μ_debit -0.65** (0.29).
  - Panel C by income type:
    - Primary: μ_credit -1.62*** (0.46); μ_debit -0.64*** (0.25).
    - Secondary: μ_credit -0.82*** (0.33); μ_debit -0.69* (0.36).
- Mechanical vs regression elasticities for investment income (Table 11, 50-country sample):
  - Regression estimates (Investment IB): -0.77 (credits), -0.32 (debits).
  - Mechanical estimates (based on FC shares): -0.82 (credits), -0.45 (debits).
  - Conclusion: mechanical valuation effects based on FC shares account for a large part of investment income elasticity; differences by income group exist.
- Cross-sectional distribution of income-balance semi-elasticities (Table 12 summary):
  - η_II: Mean -0.002; Std. Dev. 0.078; p10 -0.095; p25 -0.025; Median 0.009; p75 0.030; p90 0.062.
  - η_TI: Mean -0.319; Std. Dev. 0.163; Median -0.291.
  - η_CA = η_TI + η_II: Mean -0.345; Std. Dev. 0.203; Median -0.304.
- ER gap multipliers (inverses, distribution):
  - 1/η_TI (baseline) median -3.430; mean -3.982; Std. Dev. 2.247.
  - 1/η_CA (adjusted) median -3.292; mean -4.032; Std. Dev. 2.673.
  - Ratio η_TI / η_CA: Mean 0.995; Median 0.996; p75 1.114; p25 0.849.
  - η_II / η_TI: Mean 0.062; Median 0.004; p75 0.177; p25 -0.103.
- Country examples (illustrative effects reported):
  - United States: η_TI = −0.1; η_II = +0.02; η_CA = −0.08 ⇒ multiplier ≈ 12.5. With CA gap -1.3 percent of GDP, implied overvaluation 16 percent instead of 13 percent (trade-only).
  - United Kingdom: CA gap -2.9 percent of GDP implies overvaluation 17 percent instead of 14 percent when IB included.
  - Netherlands: η_II = +0.13; η_TI = −0.49; large CA gap 4.9 percent of GDP changes undervaluation from 10 percent to 13 percent.
  - Germany: gap 4.3 percent of GDP: undervaluation rises from about 15 percent to 17 percent when IB included.
- Distributional interpretation:
  - η_II values close to zero for most countries imply IB channel generally not essential for ER-based external adjustment.
  - However, tails include countries with non-trivial η_II; for about a quarter of countries, imposing η_II = 0 yields multiplicative bias ≥ 11 percent in ER-gap multipliers (p75), and p25 implies -15 percent or lower.

### VI. Crisis-financing and policy implications
- Event-study and causal findings:
  - IB is lower before crises, deteriorates during crises, and remains worse in aftermath — unlike TB which typically improves during crises.
  - IB deterioration stronger for financial crises and more robust among AEs and EMs than LICs.
  - Net FDI income tends to improve on the eve of crises and after, consistent with FDI’s shock-absorbing role.
  - Factor-augmented ATT: AEs and EMs IB ATT = -1.519*** (p-value 0.004); LICs IB ATT = 0.915* (p-value 0.090).
- Policy-relevant implications:
  - Crisis contingency planning and external-financing needs:
    - Not accounting for IB deterioration risks underestimating external financing gaps and foreign-exchange reserve losses, especially in financial crises and for AEs and EMs.
  - External-assessment and REER valuation:
    - For most countries, η_II ≈ 0 implies omitting IB from CAB semi-elasticity computations does not materially bias REER-gap estimates.
    - For countries in distribution tails or with large CA gaps (examples above), including IB can materially alter REER gap estimates and should be evaluated.
  - Additional research and monitoring priorities:
    - Decompose primary income into debt and non-debt income.
    - Distinguish income from domestic vs foreign yields/dividends.
    - Investigate country characteristics or policies (financial account openness, exchange rate regime) that strengthen IB shock-dampening properties.
  - Specific caution:
    - Many η_II and η_TI estimates not statistically different from zero; REER multipliers are inverses of often small η_CA, so confidence bands can be wide.

*Italic: Annex I, The Current Account Income Balance: Adjustment Channel or Vulnerability Amplifier? IMF Working Paper No. WP/2022/106 (excerpts provided).*

### Annex I. ...............................................................................................................

### Annex I.

### A. Stylized Facts
- The income balance (IB) is an important component of the current account balance (CAB):
  - The IB is at least half of the trade balance (TB) in half of the world’s countries.
  - The IB exceeds the TB in 20 percent of countries.
  - Within income flows, primary income (dominated by investment income) is four times larger than secondary income.
- Typical country patterns:
  - Most emerging markets (EMs) and low-income countries (LICs) are net receivers of secondary income and net payers of primary income.
  - Advanced Economies (AEs) are on average net payers of both primary and secondary income.
- IB–TB comovement and volatility:
  - Within-country correlations between the TB and IB are large and negative.
  - Covariance share of volatility: mean value is -98 percent (relative to counterfactual of zero correlation, the IB dampens CAB variance by half).
  - Median country’s covariance share is -37 percent.
  - The dampening effect is more substantial in EMs and LICs; absent in a significant minority of countries, including some with sizeable current account gaps.

### B. Definition of Crises
- Crisis types analyzed (from the Vulnerability Exercise dataset):
  - External crises
  - Fiscal crises
  - Financial crises
  - Real (growth) crises

### C. Assumptions Underpinning FEct Methodology
- The analysis uses a two-way fixed effects counterfactual estimator (FEct) (Liu et al., 2021) to:
  - Account for heterogeneity in IB crisis behavior.
  - Test the parallel trends assumption by comparing trends in IB for crisis versus non-crisis countries in the years leading up to a crisis.
- Diagnostics inform a factor-augmented FEct to address violations of parallel trends and to provide complementary causal-effect estimates.

### D. Estimation Procedure of FEct
- Event-study approach in the tradition of Catão & Milesi-Ferretti (2014), Gourinchas & Obstfeld (2012), and Kolerus (2021):
  - Estimates conditional means during a crisis window benchmarked against tranquil times.
- Attention to recent critiques of traditional two-way fixed effects (TWFE) estimators:
  - Heterogeneous treatment effects can bias TWFE estimates.
  - FEct chosen to mitigate such biases and permit causal diagnostics.

### E. Causal Inference under FEct
- Parallel trends diagnostics used to assess whether a causal treatment effect (crisis) can be identified relative to an unobserved counterfactual of no crisis.
- When parallel trends are rejected, factor-augmented FEct is applied to improve identification.
- Causal findings:
  - Crises worsen the IB relative to an unobserved counterfactual of no crisis among AEs and EMs.
  - Some evidence that crises improve the IB for LICs.

### F. Crisis Response — Key Results
- General patterns around crises:
  - Before a crisis, the IB is lower than in tranquil times.
  - During the crisis, the IB, if anything, worsens (in contrast to the TB which typically improves).
  - In the crisis aftermath, the IB is still no better than in tranquil times—suggesting a destabilizing role for the IB during crisis episodes.
- Heterogeneity:
  - Destabilizing behavior is especially strong for financial crises.
  - Destabilizing effects are more robust among AEs and EMs than among LICs.
- Component dynamics:
  - FDI and other equity income start to improve on the eve of the crisis.
- Robustness:
  - The lack of statistically significant improvement in the IB is robust across crisis types and income groups.
- Interpretation:
  - The IB generally does not contribute to crisis-period current account correction and may amplify vulnerability during crises for many countries.

### G. Current Account REER Elasticity — Additional Results and Implications
- Estimation approach:
  - IB semi-elasticities with respect to the exchange rate (ER) are obtained by estimating elasticities for income receipts and payments separately in panel data and combining panel-wide estimates with country-specific shares of income receipts and payments as a share of GDP to derive country-specific semi-elasticities.
  - IB semi-elasticities combined with TB semi-elasticities to derive CAB semi-elasticities and assess implications for ER misalignment estimates.
- Key quantitative findings:
  - IB semi-elasticities are broadly symmetric around zero.
  - IB semi-elasticities are between -0.1 and +0.1 in about 90 percent of countries.
  - TB semi-elasticities have a median of -0.3 and can be very small in some countries.
- Interpretation and policy relevance:
  - Income debits and credits tend to act against one another in response to ER changes, which helps explain the narrow distribution of IB semi-elasticities.
  - The IB is generally not an essential channel through which the ER stabilizes the current account; omitting the IB channel often does not seriously bias CAB semi-elasticities.
  - However, because ER misalignment estimates depend on the inverse of the CAB semi-elasticity, omitting the IB can be arithmetically important in some cases—ER misalignments can be severely biased in important cases, and examples are presented where ER misalignments are underestimated in countries with sizeable CAB gaps.
  - Caution is noted that CAB gaps can be subject to large statistical uncertainty.

*IMF Working Paper — Annex I.*

### section III discusses the covariance of the IB  and TB over the cycle, section IV uncovers the behavior of the IB

### wpiea2022106-print-pdf - section III discusses the covariance of the IB  and TB over the cycle, section IV uncovers the behavior of the IB

### II. Stylized facts — overview
- The current account (CAB) equals the trade balance (TB) plus the income balance (IB): CAB = TB + IB.
- The IB is the net of receipts/credits and payments/debits and is the sum of the primary income balance (PIB) and secondary income balance (SIB).
- Primary income receipts include Investment Income, Employee Compensation, and returns from renting natural resources; investment income (IIB) includes dividends, earning reinvestments or withdrawals, and interest on debt.
- Secondary income consists of current transfers (no quid pro quo), e.g., workers’ remittances and inter-government official transfers.

### A. The importance of the income balance and its components
- In gross terms during 2017 to 2019:
  - Median gross trade (exports + imports) = 79 percent of GDP.
  - Median gross income (receipts + payments) = 16 percent of GDP.
  - Global annual income credits (2017–2019 average for sample of 168 countries) = $5.7 trillion.
  - Exports (of goods and services) for same period = $23.3 trillion.
- In net terms (2017–2019 averages):
  - Median TB = -3 percent of GDP.
  - Median IB = +0.5 percent of GDP.
- Relative importance of IB vs TB across countries:
  - In 37 countries, |IB| / |TB| > 1.
  - In 57 countries, 1 > |IB| / |TB| > 0.5.
- Table 1 summary statistics (all variables are average of 2017 to 2019):
  - Income Balance (IB): Obs 169; Mean 3.54; Std 13.63; 25% Prct. -2.22; Median 0.54; 75% Prct. 6.42 (% of GDP).
  - IB Debit: Obs 169; Mean 15.78; Std 64.58; 25% Prct. 4.02; Median 6.94; 75% Prct. 10.54 (% of GDP).
  - IB Credit: Obs 168; Mean 20.25; Std 71.57; 25% Prct. 4.54; Median 8.03; 75% Prct. 15.06 (% of GDP).
  - PIB: Obs 169; Mean -2.99; Std 20.94; 25% Prct. -4.34; Median -2.32; 75% Prct. 0.1 (% of GDP).
  - Investment PIB (IIB): Obs 166; Mean -4.52; Std 24.33; 25% Prct. -5.00; Median -2.71; 75% Prct. -0.7 (% of GDP).
  - non-FDI IIB: Obs 153; Mean -1.01; Std 6.81; 25% Prct. -1.39; Median -0.62; 75% Prct. -0.07 (% of GDP).
  - FDI IIB: Obs 153; Mean -2.49; Std 4.04; 25% Prct. -3.91; Median -1.74; 75% Prct. -0.43 (% of GDP).
  - Portfolio IIB: Obs 141; Mean -0.17; Std 1.66; 25% Prct. -0.79; Median -0.24; 75% Prct. 0.03 (% of GDP).
  - SIB: Obs 169; Mean 6.53; Std 22.15; 25% Prct. -0.39; Median 2.37; 75% Prct. 7.57 (% of GDP).
  - Trade Balance (TB): Obs 169; Mean -8.38; Std 28.97; 25% Prct. -13.94; Median -3.1; 75% Prct. 1.98 (% of GDP).
  - CA: Obs 169; Mean -5.38; Std 35.64; 25% Prct. -6.4; Median -2.65; 75% Prct. 1.17 (% of GDP).
  - IB/TB ratio: Obs 169; Mean -0.57; Std 5.06; 25% Prct. -0.86; Median -0.42; 75% Prct. -0.02.
  - |IB/TB| ratio: Obs 169; Mean 1.49; Std 4.87; 25% Prct. 0.26; Median 0.59; 75% Prct. 0.91.
  - Gross Income (debit + credit): Obs 168; Mean 36; Std 135.44; 25% Prct. 9.97; Median 16.48; 75% Prct. 22.74 (% of GDP).
  - Gross Trade (Export + Import): Obs 169; Mean 139.19; Std 614.05; 25% Prct. 57.03; Median 78.59; 75% Prct. 107 (% of GDP).
  - Gross Income / Gross Trade ratio: Obs 168; Mean 0.3; Std 0.49; 25% Prct. 0.14; Median 0.21; 75% Prct. 0.3.
- Additional stylized facts:
  - Primary income flows are four times larger than secondary income flows in global dollar terms, but median PIB (-2.3 percent of GDP) is offset by median SIB (+2.4 percent of GDP).
  - Advanced economies (AEs) are on average net payers of both primary and secondary income; Japan and some other G7 countries have positive PIBs.
  - In 2017–2019, average developed country had gross investment income = 14.7 percent of GDP; EMs = 5.5 percent of GDP; LICs = 4.3 percent of GDP.
  - Only 25 percent of countries are net receivers of primary income; about 70 percent are net receivers of secondary income.
  - Within PIB, IIB is the predominant component; FDI income is the biggest component of investment income for most countries.
  - Remittances make up 80% of secondary income inflows in the median country.
  - Many countries with highest IB/GDP ratios are EMs or LICs and rely on remittances to finance imports; countries with lowest IB/GDP ratios are AEs or EMs with large negative primary income outflows.

### B. NIIP (IIP) and the income balance
- Investment income is closely linked to the international investment position (IIP); Colacelli et al. (2021) estimate that a 10 percentage-point rise in net foreign assets as a share of GDP is associated with a rise in the income balance of 0.2 percent of GDP.
- Many countries pay a higher return on liabilities than the return on their assets; a country with a neutral IIP is predicted to have a negative IIB.
- Feedback loop risk:
  - Larger primary/investment income outflows imply larger deficits that need financing → more negative IIP → larger income outflows.
  - COVID-19 related foreign financing has weakened many countries’ IIPs.
  - For a debtor country, lower GDP growth or higher interest rates lead to faster deterioration in net foreign assets as a share of GDP.
  - Offsetting channels: wealth effects and exchange rate adjustment can stabilize (creditor countries may consume/import more due to positive wealth effects; debtor countries may see currency depreciation that increases net exports).
  - Empirical asymmetry: creditors appear to exhibit destabilizing dynamics despite currency appreciation effects, while debtors exhibit stabilizing dynamics mainly through wealth effects.
- AE vs EM IIP patterns (sample and literature findings):
  - In a sample of 50 AEs and EMs (Bénétrix et al., 2019): median NFA = –21 percent; mean NFA = –4 percent.
  - EMs: liabilities exceed assets by about one-third.
  - In gross terms, AEs are much more globally financially integrated than EMs.
- Crisis risk and IIP composition:
  - Higher net foreign liabilities increase probability of external crisis, especially if liabilities are tilted towards net debt.
  - Net foreign portfolio equity liabilities have weaker estimated effects on external crisis probabilities.
  - FDI liabilities tend to reduce crisis risk.
  - Foreign-currency debt liabilities increase likelihood of stress episodes; equity liabilities do not.
  - Debt liabilities incur interest payments largely invariant to domestic cycle; FDI exhibits risk-sharing properties.
- IIP composition summary (selected reported values and distinctions):
  - For AEs, debt contributes just over half of both assets and liabilities; NFA net of debt is only 5 percent of GDP on average.
  - For EMs, debt securities account for about 60 percent of assets; debt liabilities account for about half of all liabilities.
  - For EMs in sample, NFA ≈ 30 percent of GDP; excluding net debt would reduce the liability closer to 20 percent of GDP.
- Currency composition and exchange rate channel implications:
  - Most countries have negative IIBs; other things equal, a negative IIB tends to deteriorate following a depreciation.
  - Reported FC shares (from Table 2 and related discussion):
    - Overall sample stock’s FC share: 83% (mean) and 96% (median) for assets; 36% (mean) and 33% (median) for liabilities.
    - Debt stock’s FC share: 79% (mean) and 95% (median) for assets; 66% (mean) and 77% (median) for liabilities.
    - For AEs: median foreign-currency (FC) share of foreign assets ≈ 90% (mean lower); FC share of foreign liabilities ≈ 30 percent.
    - For EMs: FC share of foreign assets ≈ 100 percent; FC share of foreign liabilities ≈ 40 percent.
  - Implications of a depreciation:
    - Depreciation mechanically increases local-currency income receipts from foreign assets (most assets denominated in FC) more than it increases local-currency income payments on foreign liabilities (smaller FC share of liabilities), implying, other things equal, the IIB would improve following a depreciation for many countries.
    - Regarding stock positions, AE median FC NFA = 74 percent of GDP (almost all AEs have positive FC NFAs); EM median FC NFA = 16 percent of GDP.
    - For 80 percent of countries in the sample, the NFA would mechanically improve following a depreciation because they have a positive FC NFA.

*Italic source: wpiea2022106-print-pdf (sections II.A and II.B content provided).*

### 1.5 percent of GDP in AEs and -0.5 percent of  GDP in EMs.  Therefore, a depreciation would mechanically

### wpiea2022106-print-pdf - 1.5 percent of GDP in AEs and -0.5 percent of  GDP in EMs.  Therefore, a depreciation would mechanically

### III. Correlation and variance decomposition
- Within-country correlations of TB and IB over time:
  - Aggregate sample: Cor(tb,ib) mean = -0.35, median = -0.42.
  - Correlations are statistically significant and negative in 60 percent of countries (results available on request).
  - Negative correlation present for all components of the income balance; TB is more negatively correlated with total IB than with any individual IB elements.
  - By income group:
    - AEs: Cor(tb,ib) mean = -0.10, median = -0.11.
    - EMs: Cor(tb,ib) mean = -0.41, median = -0.48.
    - LICs: Cor(tb,ib) mean = -0.43, median = -0.49.
  - Component-level correlations (mean / median):
    - Cor(tb,pib): -0.27 / -0.35 (All); AEs -0.08 / -0.07; EMs -0.35 / -0.44; LICs -0.26 / -0.35.
    - Cor(tb,sib): -0.14 / -0.14 (All); AEs 0.07 / 0.09; EMs -0.13 / -0.13; LICs -0.30 / -0.35.
    - Cor(tb,iib non-FDI): -0.15 / -0.13 (All); AEs 0.04 / 0.17; EMs -0.18 / -0.18; LICs -0.26 / -0.17.
    - Cor(tb,iib FDI): -0.09 / -0.16 (All); AEs -0.06 / -0.10; EMs -0.21 / -0.24; LICs 0.12 / 0.15.
    - Partial Cor(tb,ib)|trend: -0.27 / -0.32 (All); AEs 0.01 / -0.06; EMs -0.31 / -0.38; LICs -0.38 / -0.43.
    - Cor(tb,ib) in first differences: -0.21 / -0.23 (All); AEs 0.00 / -0.00; EMs -0.22 / -0.24; LICs -0.31 / -0.30.
  - Interpretation:
    - Negative correlations stronger in EMs and LICs; AEs show much smaller correlation.
    - In EMs, negative correlation mainly from IIB; in LICs, both PIB and SIB are responsible.
    - For AEs, partial and first-difference correlations near zero imply diverging trends drive small correlations.

- Variance decomposition of the current account (Panel B, Table 3):
  - Var(ib)/Var(ca)*100: All mean = 59, median = 28; Advanced mean = 46, median = 19; Emerging mean = 49, median = 27; Low-Income mean = 87, median = 58.
  - Var(tb)/Var(ca)*100: All mean = 143, median = 113; Advanced mean = 106, median = 81; Emerging mean = 143, median = 117; Low-Income mean = 172, median = 125.
  - 2Cov(tb,ib)/Var(ca)*100: All mean = -98, median = -37; Advanced mean = -49, median = -11; Emerging mean = -93, median = -43; Low-Income mean = -145, median = -70.
  - Aggregate interpretation:
    - Average covariance share is -98% (All countries): a covariance share of -100 percent implies CA variance is cut in half relative to a zero-correlation counterfactual.
    - Median country covariance share = -37% (All): implies CA variance is cut by 27%.
    - Across groups: median covariance share in AEs = -11%; EMs = -43 percent; LICs = -70 percent.
    - In 80 percent of countries the covariance term is negative.
    - Notable contributors to global CA imbalances (mentioned): USA, Netherlands, Germany, the UK, Russia show positive or negligible covariance; Canada, Switzerland, Poland, Mexico show covariance shares near or below -100%.

- Decomposition into gross components (equation (2) and Table 4):
  - Covariance decomposition identifies four gross covariances contributing to Cov(ib,tb):
    - Cov(cr,ex) (income credits with exports): positive.
    - Cov(db,im) (income debits with imports): positive.
    - Cov(db,ex) (income debits with exports): negative.
    - Cov(cr,im) (income credits with imports): negative.
  - Table 4: values expressed as "% Var (CA)" (All / Advanced / Emerging / Low-income):
    - Cov(cr,ex): All 7.18 (58.4%); Advanced 14.79 (115.6%); Emerging 3.82 (28.3%); Low-income 8.29 (70.3%).
    - Cov(db,im): All 8.17 (55.9%); Advanced 14.53 (128.6%); Emerging 5.20 (26.0%); Low-income 8.56 (62.5%).
    - Cov(db,ex): All 11.99 (66.4%); Advanced 19.77 (125.9%); Emerging 11.12 (47.9%); Low-income 8.83 (59.6%). (Sign listed as "-")
    - Cov(cr,im): All 16.02 (95.3%); Advanced 16.04 (135.2%); Emerging 10.64 (41.4%); Low-income 24.54 (157.0%). (Sign listed as "-")
    - Cov(ib,tb) = -12.67 (-47.5%) All; Advanced -6.50 (-16.9%); Emerging -12.74 (-35.0%); Low-income -16.51 (-83.8%).
  - Interpretation:
    - On aggregate, Cov(cr,im) and Cov(db,ex) are important negative contributors to Cov(ib,tb).
    - Cov(cr,im) particularly important in LICs (driven by secondary income).
    - Cov(db,ex) important in AEs and EMs (much driven by primary income).
    - Positive Cov(cr,ex) and Cov(db,im) suggest links like secondary receipts financing consumption (including imports) and profits from exports leading to higher repatriation of investment income outflows.

### IV. The Behavior of the IB During Crisis Episodes
A. Methodology
- Crisis definitions and sample:
  - Crises types: external, fiscal, financial, real (detailed definitions in Appendix B).
  - Sample: 177 countries; average of 25 new crises per year; crises last 2 years on average.
- Empirical approach:
  - Objective: estimate conditional dynamic means of IB shortly before, during, and after crises relative to tranquil times.
  - Methodological choice: use the Fixed-Effect Counterfactual (FEct) estimator (Gobillon & Magnac (2016); Liu et al. (2021a)).
  - Motivation:
    - TWFE estimators can be biased when treatment effects are heterogeneous or when treatment and outcome correlate (heterogeneous treatment effects; divergent pretrends).
    - FEct is robust to heterogeneity in behavior during crises and facilitates testing for pretrends.
  - Placebo tests and pretrend assessment:
    - Use placebo tests recommended by Liu et al. (2021) and Hartman & Hidalgo (2018).
    - In placebo, P periods before actual crisis are hidden; same estimators applied to periods -P to 0.
    - P is set to 3 years.
    - Rejection of parallel trends in placebo interpreted as evidence that the variable is a significant pre-crisis predictor and that further refinements are needed for causal identification.
  - Appendices:
    - Appendix C: technical assumptions of FEct.
    - Appendix D: estimation procedure.
    - Appendix E: augmentations to remove pretrends and estimate treatment effects relative to an unobserved no-crisis counterfactual.

B. Conditional Means Results
- Table 5 (IB and TB conditional means around crises):
  - Reported result for all countries and any type of crisis, first row:
    - The IB is 0.709 percent of GDP lower than normal in the years leading up to the crisis. 

*Italic line: Content extracted from wpiea2022106-print-pdf - 1.5 percent of GDP in AEs and -0.5 percent of  GDP in EMs.  Therefore, a depreciation would mechanically*

### 1.017 percent of GDP lower than normal during  crisis years. In  the three years after a  crisis, the IB  is 0.309

### wpiea2022106-print-pdf - 1.017 percent of GDP lower than normal during  crisis years. In  the three years after a  crisis, the IB  is 0.309 percent of GDP lower than normal, albeit insignificantly so, and this value represents a small improvement relative to the pre-crisis years.

### Key findings on IB and TB during crises
- Across all crises (All Any), the IB is estimated:
  - Before crisis: -0.709*** (p-value 0.007)
  - During crisis: -1.017*** (p-value 0.000)
  - After crisis: -0.309 (p-value 0.198)
- Across all crises (All Any), the TB is estimated:
  - Before crisis: -0.89** (p-value 0.028)
  - During crisis: 0.078 (p-value 0.853)
  - After crisis: -0.116 (p-value 0.804)
- Interpretation: the IB is lower than the no-crisis benchmark in the three years before a crisis, deteriorates further during crisis years, and improves after the crisis but remains below benchmark; the TB deteriorates before crises but improves sharply during crises and is broadly in line with normal years thereafter.

### Dynamics and component responses (Figure V-1 and Tables 5–7)
- Dynamic pattern (first row description):
  - IB: lower than the no-crisis benchmark in years “-3”, “-2”, “-1”; small negative change “During Crisis”; improves in years “1”, “2”, “3” but remains below benchmark.
  - TB: deteriorates relative to benchmark before crisis; improves sharply during crisis; remains close to no-crisis benchmark 3 years after crisis.
- FDI, debt, and portfolio equity (IIB components) summary:
  - Net FDI income (All Any):
    - Before: 0.504 (p-value 0.111)
    - During: 0.53* (p-value 0.072)
    - After: 0.379** (p-value 0.050)
  - Net Debt Income (All Any):
    - Before: -0.153 (p-value 0.213)
    - During: -0.35* (p-value 0.062)
    - After: -0.098 (p-value 0.187)
- Interpretation of components:
  - Net debt income shows a negative and robust decrease before and during crises relative to tranquil times; debt-income deterioration is more persistent after crises.
  - Net FDI income increases before and during crises and remains higher after crises; FDI exhibits shock-absorbing and risk-sharing properties that stabilize the current account.
  - Portfolio equity income changes are small and insignificant (available on request).

### Crisis-type heterogeneity (selected estimates from Table 5 and Table 7)
- Financial crises (All sample):
  - IB: Before -0.729* (p-value 0.073); During -1.521*** (p-value 0.002); After -1.32*** (p-value 0.005)
  - TB: Before -1.205** (p-value 0.046); During 1.131* (p-value 0.053); After 1.093* (p-value 0.075)
- External crises (All sample):
  - IB: Before -0.856** (p-value 0.026); During -1.038* (p-value 0.086); After -0.144 (p-value 0.649)
  - TB: Before -1.182*** (p-value 0.009); During -0.42 (p-value 0.603); After 0.328 (p-value 0.562)
- Real crises (All sample):
  - IB: During -0.892** (p-value 0.033)
  - TB: During 0.516 (p-value 0.357); After 0.939 (p-value 0.104)
- Interpretation: IB does not play a stabilizing role for any crisis type; destabilization is particularly pronounced in financial crises. TB responses contrast sharply with IB in financial crises.

### Country group differences (AEs and EMs vs LICs)
- AEs and EMs (Any crisis, Table 5):
  - IB: Before -1.193*** (p-value 0.000); During -1.595*** (p-value 0.000); After -0.372 (p-value 0.238)
  - TB: Before -1.404** (p-value 0.011); During 0.355 (p-value 0.464); After -0.03 (p-value 0.950)
- LICs (Any crisis, Table 5):
  - IB: Before -0.517 (p-value 0.389); During -0.476 (p-value 0.491); After 0.023 (p-value 0.976)
  - TB: Before -0.489 (p-value 0.288); During -0.133 (p-value 0.762); After -0.372 (p-value 0.379)
- Interpretation:
  - Destabilization (IB deterioration) is more robust among AEs and EMs, likely reflecting deeper financial integration.
  - In LICs, IB coefficients are closer to zero and often insignificant; destabilization is more limited and concentrated in external crises.

### Primary and secondary income behavior (Table 6)
- Primary Income Balance (PIB) (All Any):
  - Before: 0.000 (p-value 0.999)
  - During: -0.579* (p-value 0.076)
  - After: -0.191 (p-value 0.289)
- Secondary Income Balance (SIB) (All Any):
  - Before: -0.59*** (p-value 0.004)
  - During: -0.616*** (p-value 0.008)
  - After: 0.01 (p-value 0.954)
- Interpretation:
  - PIB deteriorates during crises, especially in external and financial crises and in AEs and EMs; recovery is slow and persistent after external and financial crises.
  - SIB is already lower before crises and does not deteriorate further during crises; SIB often rises to reach the tranquil benchmark after crises, with LICs sometimes reaching the benchmark during crises (possibly due to countercyclical remittances).
  - SIB decreases are larger and more significant in fiscal crises and among AEs and EMs; SIB is significantly positive after external crises and significantly negative after financial crises.

### Causal inference: factor-augmented ATT estimates (Table 8)
- Method: factor-augmented ATT estimation addressing divergent pretrends; placebo interval is three years; p-values from block bootstrap with 500 iterations.
- All countries (Any crisis):
  - IB ATT: 0.303 (p-value 0.641)
  - Net FDI Income ATT: 0.53* (p-value 0.072)
  - Net Debt Income ATT: -0.35* (p-value 0.062)
  - SIB ATT: -0.289 (p-value 0.507)
- AEs and EMs (Any crisis):
  - IB ATT: -1.519*** (p-value 0.004)
  - Net FDI Income ATT: 0.465 (p-value 0.202)
  - Net Debt Income ATT: -0.493** (p-value 0.037)
  - SIB ATT: 0.28 (p-value 0.418)
- LICs (Any crisis):
  - IB ATT: 0.915* (p-value 0.090)
  - Net FDI Income ATT: 0.361 (p-value 0.465)
  - SIB ATT: 0.581 (p-value 0.328); net debt income: "insufficient observations" in some rows
- Interpretation:
  - Stark difference by income group: AEs and EMs experience a large negative causal IB effect during crises (IB ATT -1.519***, about -1.5 percent of GDP relative to a no-crisis counterfactual).
  - LICs show a positive and significant IB ATT (0.915*), implying IB is better than the counterfactual of no crisis—potentially linked to larger secondary income shares and smaller non-FDI investment income shares.
  - Net FDI income ATTs are often positive across groups and crisis types, indicating FDI's stabilizing role.
  - Net debt income ATTs are negative and concentrated in AEs and EMs.
  - For LICs, secondary income ATT is sizably positive and significant in some cases (e.g., fiscal crises), suggesting crises may increase net secondary income receipts to LICs.

### Exchange rate adjustment and the role of the IB (conceptual and empirical implications)
- Current account gap to REER gap translation uses semielasticity ηCA, where ηCA = ηTI + ηII (ηTI for trade balance, ηII for income balance).
- If ηII ≠ 0, excluding the income-balance channel (ηII = 0 assumption) can bias REER gap estimates; practice often assumes all adjustment occurs through the TB (i.e., ηII = 0).
- Theoretical and empirical considerations:
  - Both income receipts (credits) and income payments (debits) are expected to rise after a depreciation due to mechanical valuation effects and other channels (e.g., increased local-currency profits leading to more repatriation).
  - Prior panel evidence cited: income credit elasticities ≈ -1/2; income debit elasticities ≈ -1/4; because receipts and payments offset, income-balance semi-elasticities are generally small.
- Implication: the income balance can respond to exchange rate movements; whether the exchange rate aids external adjustment through the IB depends on the net of credit and debit responses and their relative sizes—an empirical question requiring estimation of ηII.

*Source: IMF Working Paper (wpiea2022106-print-pdf) — excerpts provided in the supplied content.*

### 0.03 in Japan. For Colombia, the implied estimate is up to -0.01 (Behar, 2021). These results are consistent with

### wpiea2022106-print-pdf - 0.03 in Japan. For Colombia, the implied estimate is up to -0.01 (Behar, 2021). These results are consistent with

### Estimation methodology
- Sample: unbalanced panel of 145 countries between 1985 and 2019.
- Key parameters:
  - 휂 shows a semi-elasticity and 휇 an elasticity to the REER.
  - Exchange rate semi-elasticity of the nominal income balance: 휂휂
퐼퐼퐼퐼 ≡ Δ(퐼퐼푇푇/푌푌) / Δ(퐸퐸푅푅/퐸퐸푅푅).
- Decomposition (equation (3)):
  - 휂휂
퐼퐼퐼퐼 = 휇휇
퐶퐶퐶퐶퐶퐶퐶퐶_푖푖퐶퐶 × 푆푆
퐶퐶퐶퐶퐶퐶퐶퐶_푖푖퐶퐶 − 휇휇
퐺퐺퐶퐶퐷퐷_푖푖퐶퐶 × 푆푆
퐺퐺퐶퐶퐷퐷_푖푖퐶퐶
  - Where 휇휇 terms are exchange rate elasticities of nominal income credits (debits) over GDP; 푆 denotes nominal share of GDP.
- Regression specification for income credits (equation (4)):
  - Includes country and year fixed effects.
  - Preferred specification includes one lag of the dependent variable and the exchange rate, and controls for the size of financial assets.
  - Long-run elasticities calculated as in equation (5): sums of exchange rate coefficients divided by (1 − coefficient on lagged dependent variable).

### Empirical results — panel regressions (highlights from Table 9)
- Dependent-variable dynamics:
  - Y (t-1) coefficients (with standard errors in parentheses):
    - All Countries IB Credit: 0.84*** (0.05)
    - All Countries IB Debit: 0.77*** (0.04)
    - AEs IB Credit: 0.83*** (0.04)
    - EMs IB Debit: 0.88*** (0.04)
    - LICs IB Credit: 0.78*** (0.04)
    - LICs IB Debit: 0.74*** (0.05)
- REER contemporaneous and lagged effects (log(REER(t)) and log(REER(t-1))):
  - log(REER(t)):
    - All Countries IB Credit: -0.89*** (0.11)
    - All Countries IB Debit: -0.87*** (0.1)
    - AEs IB Credit: -0.30 (0.24)
    - EMs IB Debit: -0.47*** (0.13)
    - Primary Income Credit: -0.87*** (0.16)
    - Secondary Income Debit: -0.32* (0.17)
  - log(REER(t-1)):
    - All Countries IB Credit: 0.78*** (0.15)
    - All Countries IB Debit: 0.70*** (0.15)
    - AEs IB Credit: 0.31 (0.21)
    - EMs IB Debit: 0.44*** (0.13)
- Other covariates:
  - log(Real GDP(t)): generally insignificant in this table; select value: Primary Income Credit: -0.24** (0.11)
  - log(Fin Asset or Liab. (t-1)): positive and significant in several specifications (e.g., All Countries IB Credit: 0.06** (0.03); All Countries IB Debit: 0.07*** (0.03))
- Sample and fit:
  - Observations: e.g., All Countries IB Credit 2,568; IB Debit 2,572.
  - Number of Countries: e.g., 143 for All Countries regressions.
  - Adjusted R-sq: high values (e.g., 0.96 for All Countries IB Credit).

### Exchange-rate elasticities (Table 10) — key estimates
- Panel A: All Countries (Income flows)
  - 휇휇
퐶퐶퐶퐶퐶퐶퐶퐶_푖푖퐶퐶 = -0.63 (0.74)
  - 휇휇
퐺퐺퐶퐶 퐷퐷_푖푖퐶퐶 = -0.74*** (0.28)
- Panel B: By Country Income Group
  - Advanced: 0.04 (0.45) ; -0.29 (0.83)
  - Emerging: -1.28*** (0.36) ; -0.96*** (0.28)
  - Low-income: -0.28 (0.46) ; -0.65** (0.29)
- Panel C: By Income Type
  - Primary: -1.62*** (0.46) ; -0.64*** (0.25)
  - Secondary: -0.82*** (0.33) ; -0.69* (0.36)
- Panel D: Trade values (comparators)
  - Cubeddu et al (2019): -0.11 +0.57
  - IMF EBA-Lite template: -0.44 +0.29
- Interpretation:
  - Income credits and debits decrease in response to ER appreciation (negative elasticities).
  - Elasticities are larger in magnitude for EMs and LICs than for AEs.
  - Primary income is more responsive in credits than in debits.

### Mechanical vs regression elasticities for investment income (Table 11)
- Focus: 50 countries with FC shares (Bénétrix et al., 2019) used to proxy FC share of investment credits/debits.
- From panel regressions and mechanical calculation (sample means for the 50 countries):
  - Regression estimates (Investment IB):
    - 흁흁
푪푪푪푪 푴푴푪푪푪푪푪푪 = -0.77
    - 흁흁
푫푫푴푴푫푫 푪푪푪푪 = -0.32
  - Mechanical estimates (Investment IB):
    - 흁흁
푴푴푪푪푴푴 ℎ 퐶퐶퐶퐶퐶퐶퐶퐶 = -0.82
    - 흁흁
푴푴푪푪푴푴 ℎ 퐺퐺퐶퐶퐷퐷 = -0.45
- By income-per-capita:
  - Advanced (Investment IB): Regression -0.54; Mechanical -0.61; Regression/Mechanical other entries: -0.68; -0.36
  - Emerging (Investment IB): Regression -1.10; Mechanical -0.40; Regression/Mechanical other entries: -0.98; -0.54
- Conclusion: mechanical elasticity (based on FC shares) is a large and perhaps dominant part of total elasticity of investment income, though differences by income group exist.

### Distribution of income-balance semi-elasticities and ER-gap implications
- Construction:
  - Use elasticities by income group (Panel B of Table 10) and country-specific average S over 2010-19 to compute 휼휼
푰푰푰푰 and 휼휼
푻푻푰푰 per equation (3).
- Cross-sectional distribution (Figure VI-1 and Table 12 summary statistics):
  - 휂휂
퐼퐼퐼퐼:
    - Mean: -0.002
    - Std. Dev.: 0.078
    - p10: -0.095
    - p25: -0.025
    - Median: 0.009
    - p75: 0.030
    - p90: 0.062
  - 휂휂
푇푇퐼퐼:
    - Mean: -0.319
    - Std. Dev.: 0.163
    - p10: -0.532
    - p25: -0.384
    - Median: -0.291
    - p75: -0.201
    - p90: -0.154
  - 휂휂
퐶퐶퐶퐶 = 휂휂
푇푇퐼퐼 + 휂휂
퐼퐼퐼퐼:
    - Mean: -0.345
    - Std. Dev.: 0.203
    - p10: -0.587
    - p25: -0.420
    - Median: -0.304
    - p75: -0.201
    - p90: -0.147
  - ER Gap Multipliers:
    - 1/휂휂
푇푇퐼퐼 (baseline ER Gap Multiplier): Mean -3.982; Std. Dev. 2.247; p10 -6.419; p25 -4.881; Median -3.430; p75 -2.600; p90 -1.879
    - 1/휂휂
퐶퐶퐶퐶 (adjusted ER Gap Multiplier): Mean -4.032; Std. Dev. 2.673; p10 -6.784; p25 -4.972; Median -3.292; p75 -2.379; p90 -1.705
  - Ratio measures:
    - 휂휂
푇푇퐼퐼 / 휂휂
퐶퐶퐶퐶: Mean 0.995; Std. Dev. 0.235; p10 0.693; p25 0.849; Median 0.996; p75 1.114; p90 1.227
    - 휂휂
퐼퐼퐼퐼 / 휂휂
푇푇퐼퐼: Mean 0.062; Std. Dev. 0.257; p10 -0.185; p25 -0.103; Median 0.004; p75 0.177; p90 0.444
- Interpretation:
  - Mode at 0.0; distribution broadly symmetric.
  - 휂휂
퐼퐼퐼퐼 values close to zero for most countries imply the IB channel is generally not an important stabilizing channel for the current account.
  - However, tails of distribution include countries with non-trivially large positive or negative values, with potential for meaningful impact on ER gap estimates.
  - Multiplicative bias from imposing 휂휂
퐼퐼퐼퐼 = 0 is small on average (휂휂
푇푇퐼퐼 / 휂휂
퐶퐶퐶퐶 ≈ 1), but for about a quarter of countries p75 implies multiplicative bias is 11 percent or higher, and p25 implies -15 percent or lower.

### Country examples and REER multiplier effects (Figure VI-3)
- Illustrative country implications (select examples from text):
  - United States:
    - 휂휂
푇푇퐼퐼 = −0.1
    - 휂휂
퐼퐼퐼퐼 = +0.02
    - Resulting CA semi-elasticity 휂휂
퐶퐶퐶퐶 = 0.08, so multiplier ≈ 12.5.
    - Using CA gap of -1.3 percent of GDP (2020 ESR example), implied overvaluation 16 percent instead of 13 percent (trade-only).
  - United Kingdom:
    - CA gap -2.9 percent of GDP (2020 ESR example) implies overvaluation 17 percent instead of 14 percent when IB included.
  - Netherlands:
    - 휂휂
퐼퐼퐼퐼 = +0.13, 휂휂
푇푇퐼퐼 = −0.49; net effect on REER multiplier modest, but large CA gap of 4.9 percent of GDP changes undervaluation from 10 percent to 13 percent.
  - Germany:
    - Gap 4.3 percent of GDP: undervaluation rises from about 15 percent to 17 percent when IB included.
- Caveat: parameter estimates (μ and hence η) are subject to uncertainty; many 휂휂
퐼퐼퐼퐼 and 휂휂
푇푇퐼퐼 estimates are not statistically significantly different from zero. REER multipliers are inverse of often small 휂휂
퐶퐶퐶퐶 so confidence bands can be wide.

### Crisis behavior and external-financing implications
- Event-study findings:
  - The IB is lower before a crisis than in tranquil times.
  - During crises the IB generally deteriorates and remains worse in crisis aftermath than in tranquil times — unlike the trade balance (TB).
  - IB deterioration is stronger for financial crises and more robust among AEs and EMs than LICs.
  - Equity (including FDI) income starts improving on the eve of crises.
- Causal estimates:
  - Crises worsen the IB relative to a no-crisis counterfactual among AEs and EMs.
  - Some evidence crises improve the IB for LICs (potentially due to secondary income importance and small non-FDI investment income share).
- Policy implication drawn:
  - Crisis contingency planning may underestimate external financing gaps and foreign exchange reserve losses if IB deterioration is not accounted for.
  - For financial crises, underestimation of financing needs can be pronounced given IB deterioration patterns.

### Conclusions and policy-relevant implications
- Summary of roles:
  - Over the cycle (mostly tranquil times), negative correlation between IB and TB dampens CAB volatility — a stabilizing effect, stronger in EMs and LICs.
  - During crises, IB typically deteriorates and can act as a destabilizer (worsening external financing needs).
- External-assessment implications:
  - For most countries, IB semi-elasticities close to zero imply omitting IB from external-sector assessments does not materially bias CAB semi-elasticities.
  - For some countries (notably those in distribution tails or key contributors to global imbalances), including the IB channel materially affects REER gap estimates and should be explored.
- Additional considerations:
  - IB dynamics can amplify long-run debtor/creditor asymmetries: a debtor with negative IB facing depreciation can worsen IB and accelerate external liability accumulation; a creditor facing appreciation experiences stabilizing valuation losses on income surpluses.
  - Further research recommended:
    - Decompose primary income into debt and non-debt income.
    - Distinguish income derived from domestic vs foreign yields/dividends.
    - Investigate country characteristics or policies (e.g., financial account openness, exchange rate regime) that bolster IB shock-dampening role.

*Italic: Source — IMF Working Paper excerpt (wpiea2022106-print-pdf).*

### Annex I.

### Annex I.

### A. Stylized Facts
- Appendix Table 1: Summary statistics by income group, percent of GDP (selected indicators, means and percentiles as reported)
  - Income Balance - Net: Advanced Economies Mean -3.13; 25% Prct. -3.07; 75% Prct. 0.54. Emerging Economies Mean 2.53; 25% Prct. -3.81; 75% Prct. 6.78. Low-income Economies Mean 8.54; 25% Prct. 0.16; 75% Prct. 9.51.
  - Debt Investment Income - Net: Advanced Economies Mean 1.89; 25% Prct. -0.97; 75% Prct. 0.37. Emerging Economies Mean -0.74; 25% Prct. -1.1; 75% Prct. -0.14. Low-income Economies Mean 0.22; 25% Prct. -0.39; 75% Prct. 0.
  - Portfolio Equity Income - Net: Advanced Economies Mean -3.55; 25% Prct. -0.27; 75% Prct. 0.32. Emerging Economies Mean -0.01; 25% Prct. -0.01; 75% Prct. 0.03. Low-income Economies Mean 0.26; 25% Prct. -0.01; 75% Prct. 0.
  - FDI Equity Income - Net: Advanced Economies Mean -0.01; 25% Prct. -1.12; 75% Prct. 1.82. Emerging Economies Mean -3.01; 25% Prct. -4.68; 75% Prct. -1.24. Low-income Economies Mean -1.72; 25% Prct. -2.77; 75% Prct. -0.29.
  - Primary Income - Net: Advanced Economies Mean -2.28; 25% Prct. -2.7; 75% Prct. 1.21. Emerging Economies Mean -2.45; 25% Prct. -5.29; 75% Prct. -1.14. Low-income Economies Mean -4.36; 25% Prct. -3.3; 75% Prct. 0.02.
  - Secondary Income - Net: Advanced Economies Mean -0.86; 25% Prct. -1.47; 75% Prct. -0.39. Emerging Economies Mean 4.98; 25% Prct. 0.04; 75% Prct. 7.53. Low-income Economies Mean 12.89; 25% Prct. 3.35; 75% Prct. 9.51.
  - Income - Credit: Advanced Economies Mean 25.23; 25% Prct. 6.41; 75% Prct. 13.39. Emerging Economies Mean 11.78; 25% Prct. 3.95; 75% Prct. 14.57. Low-income Economies Mean 31.04; 25% Prct. 5.49; 75% Prct. 16.03.
  - Income - Debit: Advanced Economies Mean 28.37; 25% Prct. 6.94; 75% Prct. 16.57. Emerging Economies Mean 9.22; 25% Prct. 4.46; 75% Prct. 11.61. Low-income Economies Mean 19.66; 25% Prct. 2.93; 75% Prct. 7.14.
  - Trade Balance: Advanced Economies Mean 6.46; 25% Prct. 0.21; 75% Prct. 6.49. Emerging Economies Mean -5.85; 25% Prct. -11.13; 75% Prct. 2.33. Low-income Economies Mean -20.23; 25% Prct. -20.75; 75% Prct. -3.51.
  - Trade Exports: Advanced Economies Mean 61.59; 25% Prct. 31.91; 75% Prct. 76.88. Emerging Economies Mean 43.99; 25% Prct. 29.01; 75% Prct. 53.54. Low-income Economies Mean 73.69; 25% Prct. 15.03; 75% Prct. 35.21.
  - Trade Imports: Advanced Economies Mean 54.79; 25% Prct. 31.99; 75% Prct. 54.01. Emerging Economies Mean 49.85; 25% Prct. 30.47; 75% Prct. 59.75. Low-income Economies Mean 151.92; 25% Prct. 27.93; 75% Prct. 51.58.
  - Current Account Balance: Advanced Economies Mean 3.33; 25% Prct. -0.95; 75% Prct. 4.79. Emerging Economies Mean -3.32; 25% Prct. -6.18; 75% Prct. 0.67. Low-income Economies Mean -11.69; 25% Prct. -10.43; 75% Prct. -2.12.
  - Current Account - Credit: Advanced Economies Mean 86.82; 25% Prct. 41.27; 75% Prct. 95.17. Emerging Economies Mean 55.59; 25% Prct. 35.73; 75% Prct. 63.53. Low-income Economies Mean 104.73; 25% Prct. 23.55; 75% Prct. 50.35.
  - Current Account - Debit: Advanced Economies Mean 83.16; 25% Prct. 39.69; 75% Prct. 82.71. Emerging Economies Mean 59.07; 25% Prct. 36.33; 75% Prct. 70.62. Low-income Economies Mean 171.58; 25% Prct. 32.36; 75% Prct. 58.27.
- Appendix Figure 1 (descriptive): In the sample, 25 new crises happen in an average year. Crises last 2 years on average.

### B. Definition of Crises
- Four crisis categories followed from the Vulnerability Exercise: external, fiscal, financial, and real.
- External Crisis
  - A) Sudden stops in capital flows: defined when net private capital inflows as a percentage of GDP are at least two percentage points lower than in the previous year and two years before and when the country gets approved to tap large IMF financial support.
  - B) Exchange market pressure (EMP) events: used when sudden stops cannot be measured; EMP events defined when a weighted average of the annual percentage depreciation in the nominal exchange rate lies in the lower 15th-percentile of the whole panel or when the country gets approved for large IMF support. The EMP database covers 192 countries during the 1990-2017 period.
- Fiscal Crises
  - A country is in a fiscal crisis in any year if any of four criteria are met: (1) occurrence of a sovereign default or debt restructuring; (2) substantial official IMF financing; (3) high inflation or accumulation of domestic arrears (implicit default); and (4) loss of market access or spikes in sovereign yields. The crisis dataset covers 188 economies from 1980 to 2017.
- Financial Crises
  - Criteria require evidence of significant financial distress (e.g., sizable bank runs, bank losses, or bank liquidations) and significant policy interventions (e.g., emergency liquidity support, public recapitalization, or nationalization). The crisis dataset covers 117 countries from 1980 to 2017.
- Real Crises
  - Defined using four GDP series and four thresholds: (i) annual growth rate, (ii) cumulative growth over past three years, (iii) growth relative to most recent five-year average, (iv) average GDP level relative to previous three-year average. A value flagged as crisis if it falls below the 10th percentile in one of groups: (i) all countries, (ii) same income group, (iii) income group per WEO 1980 classification, (iv) same tercile of year-on-year growth volatility. Sixteen crisis criteria result; a country-year is a real sector crisis when nine or more indicators signal a crisis.

### C. Assumptions Underpinning FEct Methodology
- Assumption 1 (linear functional form): For any country i and period t,
  - Y_iC = δ_iC G_iC + X_iC' β + α_i + ξ_C + ε_iC  (equation (6) notation as in source).
  - Implications:
    - a) No carry-over effect: Y_iC(G_iC, G_iC-1, ...) = Y_iC(G_iC).
    - b) No lagged dependent variable: past outcomes do not directly affect current variable.
- Assumption 2 (strict exogeneity): For any two countries i and j and any two periods t and s,
  - ε_jA ⟂ G_iC, X_iC', α_i, ξ_C  (equation (7) notation as in source).
  - Implies the parallel trends assumption and that the crisis is not anticipated or predicted before it happens.

### D. Estimation Procedure of FEct
- Three-step estimating strategy:
  1. Estimate a TWFE model using only tranquil times (crisis years and years before/after treated as missing):
     - Y_iC(0) = μ + α_i + ξ_C + ε_iC | G_iC = 0  (equation (8)).
     - Provides estimates μ̂, α̂, ξ̂_C.
  2. Predict counterfactuals for observations retained as missing using coefficients from step 1:
     - Ŷ_iC(0) = μ̂ + α̂ + ξ̂_C | G_iC = 1  (equation (9)).
  3. Compute per-crisis change and average:
     - δ̂_iC = Y_iC − Ŷ_iC(0) | G_iC = 1; average crisis effect = simple empirical average of δ̂_iC.
- Notes:
  - No assumptions imposed on δ_iC being constant across time and country.
  - When no covariates are included, FEct is equivalent to a synthetic control method (each treated observation matched to a weighted sum of control observations).

### E. Causal Inference under FEct
- Potential outcomes notation:
  - Y_iC(0) = X_iC' β + α_i + ξ_C + ε_iC; Y_iC(1) = Y_iC(0) + δ_iC.
  - ATT of interest: CATT = E[δ_iC | G_iC = 1]  (equation (10)).
- Addressing violations of parallel trends:
  - Pre-crisis deviations observed in some IB components can violate parallel trends.
  - Use factor-augmented version of (8) to relax Assumption 1 and restore validity of Assumption 2.
  - Factor-augmented model:
    - Y_iC(0) = μ + X_iC' β + λ_i f_C + α_i + ξ_C + ε_iC  (equation (11)).
    - f_C is vector of time-specific factors; λ_i is country-specific factor loading vector.
  - Number of factors chosen as minimum that satisfies placebo test for divergent pre-trends; reported as zero, one, or two in all cases.
  - Factors are estimated via maximum likelihood in step 1.

### F. Crisis Response Additional Results
- Appendix Figure 2: Income balance dynamic crisis response in different types of crises (figure referenced in source).
- Appendix Figure 3: Dynamic crisis response of PIB, SIB, IIB, and total equity income in any type of crisis (figure referenced in source).

### G. Current Account REER Elasticity Additional Results
- Appendix Table 2: Summary statistics of different constructions of the income balance REER semi-elasticity (reported exactly)
  - Aggregate Sample: Mean -0.025; Std. Dev. 0.112; p10 -0.090; p25 -0.034; p50 0.003; p75 0.024; p90 0.045.
  - Income Group Samples: Mean -0.002; Std. Dev. 0.078; p10 -0.095; p25 -0.025; p50 0.009; p75 0.030; p90 0.062.
  - Primary and Secondary Income: Mean -0.097; Std. Dev. 0.171; p10 -0.274; p25 -0.102; p50 -0.050; p75 -0.007; p90 0.010.
  - Mechanical (Investment): Mean -0.018; Std. Dev. 0.054; p10 -0.052; p25 -0.019; p50 -0.002; p75 0.005; p90 0.015.
- Appendix Table 3: IB and TB semi-elasticity for selected economies (reported exactly as in source)
  - Argentina: η_TI (From EBA 2018) -0.11; η_IB 0.021
  - Australia: -0.15; 0.021
  - Belgium: -0.58; 0.058
  - Brazil: -0.09; 0.020
  - Canada: -0.22; 0.021
  - China: -0.12; -0.005
  - France: -0.22; 0.027
  - Germany: -0.28; 0.027
  - Hong Kong SAR: -; 0.174
  - India: -0.16; -0.031
  - Indonesia: -0.14; 0.020
  - Italy: -0.2; 0.018
  - Japan: -0.12; 0.008
  - Korea: -0.27; 0.009
  - Malaysia: -0.42; 0.029
  - Mexico: -0.27; -0.008
  - Netherlands: -0.53; 0.128
  - Poland: -0.35; 0.023
  - Russia: -0.16; 0.022
  - South Africa: -0.2; 0.026
  - Spain: -0.22; 0.025
  - Sweden: -0.28; 0.037
  - Switzerland: -0.39; 0.089
  - Thailand: -0.41; 0.010
  - Turkey: -0.2; 0.006
  - United Kingdom: -0.2; 0.041

*Annex I, The Current Account Income Balance: Adjustment Channel or Vulnerability Amplifier? IMF Working Paper No. WP/2022/106.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022106-print-pdf.pdf_