## 4.1  Trade balances and NFA returns

## Source details

**Canonical URL:** [4.1  Trade balances and NFA returns](https://www.imf.org/-/media/files/publications/wp/2018/wp1879.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1879.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1879.pdf.json)

---

### Averages — role of NFA returns vs trade balances
- Over 1990-2015, NFA returns have played an important role in driving NFA dynamics, with orders of magnitudes comparable or superior to those of the trade balance.
- NFA returns have become more important in recent years, reflecting both greater valuation changes and income flows, in relation to GDP.
- Return differentials displayed remarkable stability when comparing 1990-2007 vs 2008-15, suggesting post-GFC drop in yields was largely symmetric across largest countries.
- Greater importance of NFA returns (defined in percent of GDP) along with stable return differentials indicates growing role of stock positions in driving NFA returns.
- Cross-section evidence: trade imbalances are a key driver of changes in NFA/GDP ratios and are partially offset by real NFA returns (negative correlation with trade balances).

- Table 1 (annual average of absolute value, in percent of GDP):
  - Trade balance: 3.1 (1990-2015); 2.8 (1990-2007); 3.6* (2008-15)
  - Real NFA return: 4.6 (1990-2015); 4.1 (1990-2007); 5.7*** (2008-15)
    - Valuation change: 4.4 (1990-2015); 3.8 (1990-2007); 5.6*** (2008-15)
    - Investment income: 1.4 (1990-2015); 1.2 (1990-2007); 1.7** (2008-15)
    - Inflation: 0.4 (1990-2015); 0.4 (1990-2007); 0.5* (2008-15)
  - Return differential: 5.7 (1990-2015); 5.7 (1990-2007); 5.7 (2008-15)
    - Yield diff.: 1.5 (1990-2015); 1.4 (1990-2007); 1.7 (2008-15)
    - Exchange rate diff.: 2.0 (1990-2015); 1.9 (1990-2007); 2.3 (2008-15)
    - Asset price diff.: 5.7 (1990-2015); 5.6 (1990-2007); 5.7 (2008-15)

### EMEs vs AEs — differences and magnitudes
- EMEs tend to face higher yields on liabilities, but whether this leads to sustained overall return differentials is an empirical question.
- Evidence points to sizable differences between return differentials for AEs and EMEs, consistent with risk premia.

- Table 2 (annual average, percent of GDP and percent):
  - Trade balance:
    - AEs: -0.7 (1990-2015); -0.6 (1990-2007); -1.0 (2008-15)
    - EMEs: 2.2** (1990-2015); 2.1* (1990-2007); 2.4** (2008-15)
  - Real NFA return:
    - AEs: 0.4 (1990-2015); 0.4 (1990-2007); 0.5 (2008-15)
    - EMEs: -1.9*** (1990-2015); -2.4** (1990-2007); -1.4** (2008-15)
    - Valuation change (EMEs): 0.0 (1990-2015); 0.2 (1990-2007); -0.4 (2008-15)
    - Investment income (EMEs): -1.7*** (1990-2015); -1.8*** (1990-2007); -1.6*** (2008-15)
    - Inflation (EMEs): -0.2 (1990-2015); -0.3 (1990-2007); -0.1 (2008-15)
  - Return differential:
    - AEs: 0.9 (1990-2015); 1.1 (1990-2007); 0.4 (2008-15)
    - EMEs: -4.2*** (1990-2015); -3.8** (1990-2007); -4.7*** (2008-15)
    - Yield diff. (EMEs): -2.7*** (1990-2015); -2.5*** (1990-2007); -3.0*** (2008-15)
    - Exchange rate diff. (EMEs): 1.0 (1990-2015); 1.8** (1990-2007); 0.0 (2008-15)
    - Asset price diff. (EMEs): -2.5** (1990-2015); -3.1** (1990-2007); -1.7** (2008-15)

- Average NFA returns in EMEs are about 2.3 percent of GDP lower than in AEs over the whole sample; difference narrowed in recent years partly due to strengthening NFA positions in EMEs.

### Reserve-currency issuers — elevated returns and offsets
- Reserve-currency countries (US, UK, Euro Area/Germany pre-2001, Japan, Switzerland) show significantly higher real NFA returns on average.
- Over 1990-2015, reserve-currency countries have real NFA returns 2.4 percentage points of GDP above other economies and a higher overall return differential of 3.6 percentage points; differences in average trade balance (~3.7 percentage points of GDP) more than offset NFA return differentials.
- Difference in NFA returns between reserve and non-reserve countries fell in post-crisis period (from about 3 to about 1 percent of GDP), while return differentials were more stable (difference moved from 3.6 to 3.4 percent).
- US stands out with a significantly higher return differential averaging 3.3 percent over the last 25 years.

- Table 3 (annual average, percent of GDP and percent):
  - Trade balance:
    - Reserve issuers: -1.5 (1990-2015); -1.4 (1990-2007); -1.7 (2008-15)
    - Non-reserve issuers: 2.2*** (1990-2015); 2.0*** (1990-2007); 2.5*** (2008-15)
  - Real NFA return:
    - Reserve issuers: 0.8 (1990-2015); 1.0 (1990-2007); 0.4 (2008-15)
    - Non-reserve issuers: -1.6*** (1990-2015); -2.0*** (1990-2007); -0.8 (2008-15)
    - Valuation change (reserve): 0.1 (1990-2015); 0.4 (1990-2007); -0.8 (2008-15)
    - Investment income (reserve): 0.7 (1990-2015); 0.6 (1990-2007); 1.1 (2008-15)
  - Return differential:
    - Reserve issuers: 1.1 (1990-2015); 1.5 (1990-2007); 0.2 (2008-15)
    - Non-reserve issuers: -2.5** (1990-2015); -2.1** (1990-2007); -3.2** (2008-15)
    - Yield diff. (reserve): 0.9 (1990-2015); 0.9 (1990-2007); 1.1 (2008-15)
    - Exchange rate diff. (reserve): -0.3 (1990-2015); -0.1 (1990-2007); -0.6 (2008-15)
    - Asset price diff. (reserve): 0.6 (1990-2015); 0.9 (1990-2007); -0.2 (2008-15)

- Table 4 (reserve-currency countries, 1990-2015; annual averages; NFA, TB and NFA return in percent of GDP; other variables in percent):
  - United States:
    - NFA: -17.8; TB: -3.5; r_NFA: 1.6; r: 3.3**; Yield: 1.4***; XR: -0.3; AP: 2.2*
  - Euro Area:
    - NFA: -13.2; TB: 0.8; r_NFA: -0.9; r: -0.3; Yield: 0.2; XR: -0.2; AP: -0.4
    - (Sample period for the Euro Area is 2001-15)
  - United Kingdom:
    - NFA: -8.0; TB: -2.5; r_NFA: 2.1; r: 0.6; Yield: 0.3*; XR: 0.1; AP: 0.3
  - Japan:
    - NFA: 36.6; TB: 0.6; r_NFA: 0.9; r: -1.6; Yield: 1.7***; XR: -0.3; AP: -3.4**
  - Switzerland:
    - NFA: 103.4; TB: 5.7; r_NFA: -1.8; r: -1.3**; Yield: 0.8**; XR: -1.0; AP: -1.1

- Country-specific notes:
  - US: Return differential (3.3 percent) more than offsets negative NFA position, yielding overall positive NFA return (~one third of average trade deficit); differential largely from yield differential and asset price differential.
  - UK: Return differential ~0.6 percent lower than US but still sizable NFA return (~2 percent of GDP) due to large gross positions (in excess of ten times GDP).
  - Japan: Negative return differential (-1.6 percent) driven by large negative asset price valuation changes, but overall NFA return positive (~1.2 percent) due to positive NFA position and large gross debt asset positions with large yield differential.
  - Switzerland: Negative return differential (~-1.6 percent) from negative exchange rate and asset price differentials, but small positive overall NFA return.

### External positions — creditors vs debtors and trade surplus vs deficit
- Table 5 (NFA returns and return differentials, external position, 1990-2015; GDP-weighted averages):
  - Trade balance:
    - TB surplus: 2.3
    - TB deficit: -2.8***
    - Creditors: 2.8
    - Debtors: -1.4***
  - Real NFA return:
    - TB surplus: -1.0
    - TB deficit: 0.9**
    - Creditors: -0.5
    - Debtors: 0.1
    - Valuation change (TB surplus): -0.5; (TB deficit): 0.5**; (Creditors): -0.8; (Debtors): 0.3**
    - Income balance (TB surplus): -0.3; (TB deficit): 0.1; (Creditors): 0.9; (Debtors): -0.6**
    - Inflation (TB surplus): 0.1; (TB deficit): -0.4***; (Creditors): 0.5; (Debtors): -0.4***
  - Return differential:
    - TB surplus: -2.1
    - TB deficit: 1.9***
    - Creditors: -2.5
    - Debtors: 0.8**
    - Yield diff. (TB surplus): -1.0; (TB deficit): 0.8**; (Creditors): -0.6; (Debtors): 0.0
    - Exchange rate diff. (TB surplus): 0.1; (TB deficit): -0.1; (Creditors): -0.4; (Debtors): 0.2
    - Asset price diff. (TB surplus): -1.4; (TB deficit): 1.2**; (Creditors): -1.7; (Debtors): 0.6*

- Interpretation:
  - Countries with trade surpluses tended to experience lower return differentials, and vice versa.
  - Creditors run negative return differentials and debtors positive ones, suggesting stabilizing role of return differentials for external positions (primarily explained by asset price differentials, with additional contributions from yield and exchange rate differentials).

### NFA-stabilizing mechanisms — econometric evidence and key coefficients
- Panel-regression specification:
  - y_{c,t} = β nfa_{c,t−1} + δ′ X_{c,t} + κ_c + u_{c,t}
  - Negative β (and greater than -1) indicates stable dynamics with finite steady state.
  - Tests: null β = 0 (no adjustment) vs alternative β ≠ 0; negative β indicates stabilizing channel.

- Average effects (Table 6; variables in percent of GDP; panel regression with country fixed effects; standard errors clustered at country level):
  - Coefficients (standard errors in parentheses):
    - ∆nfa: L.nfa = -0.070*** (0.019); R2 = 0.074; N = 1288
    - tb: L.nfa = 0.020 (0.014); R2 = 0.647; N = 1288
    - r_NFA: L.nfa = -0.064*** (0.017); R2 = 0.126; N = 1288
    - g_NFA: L.nfa = -0.015 (0.014); R2 = 0.100; N = 1288
    - val: L.nfa = -0.065*** (0.014); R2 = 0.073; N = 1288
    - ib: L.nfa = 0.018** (0.009); R2 = 0.820; N = 1288
    - π_NFA: L.nfa = -0.016*** (0.001); R2 = 0.897; N = 1288

- Key findings from regressions:
  - After controlling for country fixed effects, NFA positions display stable dynamics on average: L.nfa coefficient -0.070*** implies a 10 percent higher NFA leads on expectation to a -0.7 percent change in the NFA position in the following year.
  - For most countries, implied steady-state NFA levels are close to observed 2015 values (exceptions include Hong Kong SAR, Norway, Germany, Netherlands; Singapore, Switzerland, Belgium would see NFA narrow relative to 2015).
  - NFA returns, not trade balances, are the main NFA-stabilizing mechanism:
    - No evidence that trade balance responds to level of net external wealth (tb column).
    - Evidence that NFA returns are negatively related to existing external wealth (r_NFA column): a 10 percent higher NFA predicts a -0.6 percent of GDP lower NFA return.
  - Trade imbalances show high persistence; other components are less persistent.
  - Overall: persistent trade imbalances have led to diverging NFA positions, while NFA returns have provided an important, though insufficient, stabilizing role.

### Robustness checks and alternative specifications (Appendix A.2 summary)
- Robustness variations reported:
  - Excluding the US;
  - Excluding financial centers (Ireland, Hong Kong SAR, Singapore and Netherlands);
  - Remove GDP weights and winsorize observations with absolute changes in NFA/GDP larger than 20 percent;
  - Consider 5-year lags;
  - Attribute BOP errors and omissions to the valuation residual;
  - Use denomination in local currency instead of USD;
  - Estimate with a system of seemingly unrelated regressions (SUR);
  - Assume panel-corrected errors under different correlation structures.
- Reported alternative-estimation highlights:
  - Excluding US (Table 16): L.nfa examples include -0.071*** (0.022); R2 examples 0.069; N = 1263.
  - Excluding financial centers (Table 17): L.nfa examples include -0.070*** (0.023); R2 examples 0.079; N = 1188.
  - Unweighted, |∆nfa| ≤ 20 (Table 18): L.nfa examples include -0.058*** (0.013); R2 examples 0.085; N = 1216.
  - 5-Year lag (Table 19): L5.nfa examples include -0.079*** (0.017); R2 examples 0.401; N = 1080.
  - BOP E&Os → valuation residual (Table 20): L.nfa -0.072*** (0.020); -0.090*** (0.015); N = 1288.
  - Local-currency denomination (Table 21): L.nfa examples -0.057** (0.024); -0.060*** (0.018); N ≈ 1277–1288.
  - SUR (Table 22): L.nfa -0.064*** (0.013); N = 1288.
  - Panel-corrected SEs (Table 23): L.nfa examples -0.062*** (0.022); -0.059*** (0.022); N = 1288.

### Implied steady-state NFA positions and dynamics
- Implied steady-state NFA is equal to minus the country fixed effect divided by β_NFA (from column 1 in Table 6); these implied steady-state values do not reflect IMF country team projections.
- Table 7 (autocorrelation / dynamics), selected coefficients (all variables in percent of GDP):
  - L.nfa: -0.007** (0.003)
  - tbr: -0.064*** (0.019)
  - NFA_g: -0.016 (0.013)
  - NFA_valibπ: -0.065*** (0.016)
  - Lag dependent var. entries: 0.830*** (0.021); 0.783***; 0.455*** (respective standard errors reported)
  - R2 examples: 0.881, 0.129, 0.117, 0.075, 0.925, 0.926
  - N observations examples: 1286, 1236, 1237, 1236, 1287, 1237
- Table 8 (stabilizing return differentials): Selected coefficients
  - L.nfa on rYield diff.: -0.026 (0.027)
  - L.nfa on XR diff: 0.075** (0.034)
  - L.nfa on AP diff: -0.100*** (0.025)
  - L.nfa on AP liab.: -0.099*** (0.023)
  - Baseline interpretation: a 10 percent of GDP higher NFA is associated with a 1 percent lower return differential; estimate may be an upper bound and is halved in unweighted regressions.

### EMEs vs AEs: speed of adjustment
- Speed of NFA adjustment is significantly higher for (individual) EMEs than for AEs by approximately a factor of 5.
  - Implied half-lives:
    - AEs: about 13 years
    - EMEs: about 2.5 years
- For both groups, external adjustment comes from NFA returns.
- Faster adjustment for EMEs holds for both debtor and creditor EMEs (interaction coefficient insignificant).

### Reserve-currency economies: slower adjustment and stabilizing components
- Reserve-currency (RC) issuers appear to adjust at a slower pace than other economies.
  - Implied half-lives:
    - RC issuers: about 16 years
    - Other economies: about 6 years
  - Difference only marginally statistically significant.
- Table 10 (selected coefficients):
  - Baseline L.nfa: -0.070*** (0.019); -0.064*** (0.017); -0.088*** (0.018) in various columns.
  - L.nfa_RC examples: 0.068* (0.035); -0.014 (0.023); 0.045 (0.030)
  - R2 examples: 0.074, 0.647, 0.126; N observations: 1288.

### NFA returns as insurance against income shocks — domestic and global
- Regression (5-year averages): y_c,t = α + β GDPgr_c,t + δ′X_c,t + κ_c + u_c,t, dependent variable = 5-year average NFA return (percent of domestic GDP) or components.

- Insurance against domestic GDP shocks (Tables 11–12, 5-year intervals):
  - Table 11 (NFA return components):
    - g_5-year coefficient on r_NFA: -0.633*** (0.136)
    - g_5-year coefficient on ibvalπ: -0.088 (0.062)
    - Alternate g_5-year on r_NFA: -0.505*** (0.153)
    - g_5-year on other component: -0.177** (0.081)
    - R2 examples: 0.508, 0.870, 0.315, 0.839; N examples: 1078, 1084
    - Interpretation: A 1 percent drop in the 5-year average GDP growth is associated with an average positive NFA return of 0.6 percent of GDP over the same period; effect comes largely from valuation changes, with marginal effects through inflation and income.
  - Table 12 (return differentials):
    - g_5-year on rYield diff.: -0.726*** (0.182)
    - g_5-year on XR diff: -0.091 (0.060)
    - g_5-year on AP diff: -0.200 (0.134)
    - g_5-year on another component: -0.460* (0.233)
    - R2 examples: 0.586, 0.877, 0.262, 0.538; N examples: 1077, 1042, 1085
    - Interpretation: Asset price differential is the main component providing insurance, with exchange rate and yield differentials also contributing.
  - Country-level evidence: For the majority of countries, NFA returns provide insurance against domestic shocks; for others, effects not statistically different from zero. Applies to AEs and EMEs.

- Global insurance providers and reserve-currency roles (Tables 13–14, 5-year intervals):
  - Table 13: NFA return over World GDP (5-year intervals), selected coefficients:
    - g_ROW: -0.012*** (0.001) across columns
    - g_ROW_RC examples: 0.066*** (0.018); 0.053*** (0.001); 0.231** (0.090)
    - g_ROW_US: 0.281*** (0.000)
    - g_ROW_EA: -0.017 (0.019)
    - g_ROW_UK: -0.178* (0.090)
    - g_ROW_JAP: -0.005 (0.018)
    - g_ROW_CHE: -0.052*** (0.018)
    - R2 examples: 0.517, 0.569; N = 870
    - Interpretation: Reserve-currency countries are providers of insurance against income shocks to the rest of the world. As a group, reserve-currency economies transfer, on average, 0.26 percent of world GDP in wealth to non-reserve countries when GDP growth in the latter group slows down by 1 percent. The US contributes with 0.32 percent of world GDP; Switzerland plays a more limited role.
  - Table 14: NFA return over Domestic GDP (5-year intervals), selected coefficients:
    - g_ROW: -0.869*** (0.129) across columns
    - g_ROW_RC examples: 1.529*** (0.133); 1.505*** (0.131); 1.669*** (0.350)
    - g_ROW_US: 0.507*** (0.019)
    - g_ROW_EA: -0.596*** (0.034)
    - g_ROW_JAP: 0.356*** (0.029)
    - g_ROW_CHE: -0.164*** (0.030)
    - R2 = 0.446; N = 870
    - Interpretation: On average, reserve-currency economies transfer about 0.7 percent of their own GDP in response to a 1 percent income shock in the rest of the world. Across individual reserve countries, contribution ranges between 0.1 and 0.9 (US top insurer, closely followed by Japan and the UK; Switzerland below average; EA at the bottom).

### Main conclusions (empirical findings)
- Persistent trade imbalances have been the main driving force of diverging NFA positions over 1990-2015, especially for systemic economies, with no sign of stock position–driven adjustment.
- NFA returns have played an important but insufficient stabilizing role, primarily through asset prices of countries’ liabilities.
- Cross-country heterogeneity:
  - EMEs adjust NFA positions faster than AEs (factor of ~5).
  - Reserve-currency issuers adjust at a slower pace than other economies.
- NFA returns serve as an insurance mechanism against domestic and global income shocks with macro-economically meaningful magnitudes; this insurance role is broad across countries.
- Evidence extends prior findings on the US as a global insurer to other reserve-currency countries (Euro area, UK, Japan, Switzerland), which also provide insurance against global income shocks in a scale broadly commensurate to the size of their economies.

*Source: wp1879 - 4.1 Trade balances and NFA returns (including Appendix A.2 robustness checks)*

### 4.1  Trade balances and NFA returns

### 4.1  Trade balances and NFA returns

### Averages
- Over 1990-2015, NFA returns have played an important role in driving NFA dynamics, with orders of magnitudes comparable or superior to those of the trade balance.
- NFA returns have become more important in recent years, reflecting both greater valuation changes and income flows, in relation to GDP.
- Return differentials displayed remarkable stability when comparing 1990-2007 vs 2008-15, suggesting post-GFC drop in yields was largely symmetric across largest countries.
- Greater importance of NFA returns (defined in percent of GDP) along with stable return differentials indicates growing role of stock positions in driving NFA returns.
- Cross-section evidence: trade imbalances are a key driver of changes in NFA/GDP ratios and are partially offset by real NFA returns (negative correlation with trade balances).

- Table 1 (annual average of absolute value, in percent of GDP):
  - Trade balance: 3.1 (1990-2015); 2.8 (1990-2007); 3.6* (2008-15)
  - Real NFA return: 4.6 (1990-2015); 4.1 (1990-2007); 5.7*** (2008-15)
    - Valuation change: 4.4 (1990-2015); 3.8 (1990-2007); 5.6*** (2008-15)
    - Investment income: 1.4 (1990-2015); 1.2 (1990-2007); 1.7** (2008-15)
    - Inflation: 0.4 (1990-2015); 0.4 (1990-2007); 0.5* (2008-15)
  - Return differential: 5.7 (1990-2015); 5.7 (1990-2007); 5.7 (2008-15)
    - Yield diff.: 1.5 (1990-2015); 1.4 (1990-2007); 1.7 (2008-15)
    - Exchange rate diff.: 2.0 (1990-2015); 1.9 (1990-2007); 2.3 (2008-15)
    - Asset price diff.: 5.7 (1990-2015); 5.6 (1990-2007); 5.7 (2008-15)

### EMEs vs AEs
- EMEs tend to face higher yields on liabilities, but whether this leads to sustained overall return differentials is an empirical question.
- Evidence points to sizable differences between return differentials for AEs and EMEs, consistent with risk premia.

- Table 2 (annual average, percent of GDP and percent):
  - Trade balance:
    - AEs: -0.7 (1990-2015); -0.6 (1990-2007); -1.0 (2008-15)
    - EMEs: 2.2** (1990-2015); 2.1* (1990-2007); 2.4** (2008-15)
  - Real NFA return:
    - AEs: 0.4 (1990-2015); 0.4 (1990-2007); 0.5 (2008-15)
    - EMEs: -1.9*** (1990-2015); -2.4** (1990-2007); -1.4** (2008-15)
    - Valuation change (EMEs): 0.0 (1990-2015); 0.2 (1990-2007); -0.4 (2008-15)
    - Investment income (EMEs): -1.7*** (1990-2015); -1.8*** (1990-2007); -1.6*** (2008-15)
    - Inflation (EMEs): -0.2 (1990-2015); -0.3 (1990-2007); -0.1 (2008-15)
  - Return differential:
    - AEs: 0.9 (1990-2015); 1.1 (1990-2007); 0.4 (2008-15)
    - EMEs: -4.2*** (1990-2015); -3.8** (1990-2007); -4.7*** (2008-15)
    - Yield diff. (EMEs): -2.7*** (1990-2015); -2.5*** (1990-2007); -3.0*** (2008-15)
    - Exchange rate diff. (EMEs): 1.0 (1990-2015); 1.8** (1990-2007); 0.0 (2008-15)
    - Asset price diff. (EMEs): -2.5** (1990-2015); -3.1** (1990-2007); -1.7** (2008-15)

- Average NFA returns in EMEs are about 2.3 percent of GDP lower than in AEs over the whole sample; difference narrowed in recent years partly due to strengthening NFA positions in EMEs.

### Reserve-currency issuers
- Reserve-currency countries (US, UK, Euro Area/Germany pre-2001, Japan, Switzerland) show significantly higher real NFA returns on average.
- Over 1990-2015, reserve-currency countries have real NFA returns 2.4 percentage points of GDP above other economies and a higher overall return differential of 3.6 percentage points; differences in average trade balance (~3.7 percentage points of GDP) more than offset NFA return differentials.
- Difference in NFA returns between reserve and non-reserve countries fell in post-crisis period (from about 3 to about 1 percent of GDP), while return differentials were more stable (difference moved from 3.6 to 3.4 percent).
- US stands out with a significantly higher return differential averaging 3.3 percent over the last 25 years.

- Table 3 (annual average, percent of GDP and percent):
  - Trade balance:
    - Reserve issuers: -1.5 (1990-2015); -1.4 (1990-2007); -1.7 (2008-15)
    - Non-reserve issuers: 2.2*** (1990-2015); 2.0*** (1990-2007); 2.5*** (2008-15)
  - Real NFA return:
    - Reserve issuers: 0.8 (1990-2015); 1.0 (1990-2007); 0.4 (2008-15)
    - Non-reserve issuers: -1.6*** (1990-2015); -2.0*** (1990-2007); -0.8 (2008-15)
    - Valuation change (reserve): 0.1 (1990-2015); 0.4 (1990-2007); -0.8 (2008-15)
    - Investment income (reserve): 0.7 (1990-2015); 0.6 (1990-2007); 1.1 (2008-15)
  - Return differential:
    - Reserve issuers: 1.1 (1990-2015); 1.5 (1990-2007); 0.2 (2008-15)
    - Non-reserve issuers: -2.5** (1990-2015); -2.1** (1990-2007); -3.2** (2008-15)
    - Yield diff. (reserve): 0.9 (1990-2015); 0.9 (1990-2007); 1.1 (2008-15)
    - Exchange rate diff. (reserve): -0.3 (1990-2015); -0.1 (1990-2007); -0.6 (2008-15)
    - Asset price diff. (reserve): 0.6 (1990-2015); 0.9 (1990-2007); -0.2 (2008-15)

- Table 4 (reserve-currency countries, 1990-2015; annual averages; NFA, TB and NFA return in percent of GDP; other variables in percent):
  - United States:
    - NFA: -17.8; TB: -3.5; r_NFA: 1.6; r: 3.3**; Yield: 1.4***; XR: -0.3; AP: 2.2*
  - Euro Area:
    - NFA: -13.2; TB: 0.8; r_NFA: -0.9; r: -0.3; Yield: 0.2; XR: -0.2; AP: -0.4
    - (Sample period for the Euro Area is 2001-15)
  - United Kingdom:
    - NFA: -8.0; TB: -2.5; r_NFA: 2.1; r: 0.6; Yield: 0.3*; XR: 0.1; AP: 0.3
  - Japan:
    - NFA: 36.6; TB: 0.6; r_NFA: 0.9; r: -1.6; Yield: 1.7***; XR: -0.3; AP: -3.4**
  - Switzerland:
    - NFA: 103.4; TB: 5.7; r_NFA: -1.8; r: -1.3**; Yield: 0.8**; XR: -1.0; AP: -1.1

- Country-specific notes:
  - US: Return differential (3.3 percent) more than offsets negative NFA position, yielding overall positive NFA return (~one third of average trade deficit); differential largely from yield differential and asset price differential.
  - UK: Return differential ~0.6 percent lower than US but still sizable NFA return (~2 percent of GDP) due to large gross positions (in excess of ten times GDP).
  - Japan: Negative return differential (-1.6 percent) driven by large negative asset price valuation changes, but overall NFA return positive (~1.2 percent) due to positive NFA position and large gross debt asset positions with large yield differential.
  - Switzerland: Negative return differential (~-1.6 percent) from negative exchange rate and asset price differentials, but small positive overall NFA return.

### External positions
- Table 5 (NFA returns and return differentials, external position, 1990-2015; GDP-weighted averages):
  - Trade balance:
    - TB surplus: 2.3
    - TB deficit: -2.8***
    - Creditors: 2.8
    - Debtors: -1.4***
  - Real NFA return:
    - TB surplus: -1.0
    - TB deficit: 0.9**
    - Creditors: -0.5
    - Debtors: 0.1
    - Valuation change (TB surplus): -0.5; (TB deficit): 0.5**; (Creditors): -0.8; (Debtors): 0.3**
    - Income balance (TB surplus): -0.3; (TB deficit): 0.1; (Creditors): 0.9; (Debtors): -0.6**
    - Inflation (TB surplus): 0.1; (TB deficit): -0.4***; (Creditors): 0.5; (Debtors): -0.4***
  - Return differential:
    - TB surplus: -2.1
    - TB deficit: 1.9***
    - Creditors: -2.5
    - Debtors: 0.8**
    - Yield diff. (TB surplus): -1.0; (TB deficit): 0.8**; (Creditors): -0.6; (Debtors): 0.0
    - Exchange rate diff. (TB surplus): 0.1; (TB deficit): -0.1; (Creditors): -0.4; (Debtors): 0.2
    - Asset price diff. (TB surplus): -1.4; (TB deficit): 1.2**; (Creditors): -1.7; (Debtors): 0.6*

- Interpretation:
  - Countries with trade surpluses tended to experience lower return differentials, and vice versa.
  - Creditors run negative return differentials and debtors positive ones, suggesting stabilizing role of return differentials for external positions (primarily explained by asset price differentials, with additional contributions from yield and exchange rate differentials).

### NFA-Stabilizing Mechanisms (econometric evidence)
- Panel-regression specification:
  - y_{c,t} = β nfa_{c,t−1} + δ′ X_{c,t} + κ_c + u_{c,t}
  - y_{c,t} is change in nfa_{c,t} or one of its drivers; X_{c,t} vector of controls; κ_c country fixed effect.
  - Negative β (and greater than -1) indicates stable dynamics with finite steady state.
  - Tests: null β = 0 (no adjustment) vs alternative β ≠ 0; negative β indicates stabilizing channel.

- Average effects (Table 6; variables in percent of GDP; panel regression with country fixed effects; standard errors clustered at country level):
  - Coefficients (standard errors in parentheses):
    - ∆nfa: L.nfa = -0.070*** (0.019); R2 = 0.074; N = 1288
    - tb: L.nfa = 0.020 (0.014); R2 = 0.647; N = 1288
    - r_NFA: L.nfa = -0.064*** (0.017); R2 = 0.126; N = 1288
    - g_NFA: L.nfa = -0.015 (0.014); R2 = 0.100; N = 1288
    - val: L.nfa = -0.065*** (0.014); R2 = 0.073; N = 1288
    - ib: L.nfa = 0.018** (0.009); R2 = 0.820; N = 1288
    - π_NFA: L.nfa = -0.016*** (0.001); R2 = 0.897; N = 1288

- Key findings from regressions:
  - After controlling for country fixed effects, NFA positions display stable dynamics on average: L.nfa coefficient -0.070*** implies a 10 percent higher NFA leads on expectation to a -0.7 percent change in the NFA position in the following year.
  - For most countries, implied steady-state NFA levels are close to observed 2015 values (exceptions include Hong Kong SAR, Norway, Germany, Netherlands; Singapore, Switzerland, Belgium would see NFA narrow relative to 2015).
  - NFA returns, not trade balances, are the main NFA-stabilizing mechanism:
    - No evidence that trade balance responds to level of net external wealth (tb column).
    - Evidence that NFA returns are negatively related to existing external wealth (r_NFA column): a 10 percent higher NFA predicts a -0.6 percent of GDP lower NFA return.
  - Trade imbalances show high persistence; other components are less persistent.
  - Overall: persistent trade imbalances have led to diverging NFA positions, while NFA returns have provided an important, though insufficient, stabilizing role.

*Source: wp1879 - 4.1  Trade balances and NFA returns*

### Appendix A.2 shows that these results are robust to a number of variations, such as:

### wp1879 - Appendix A.2 shows that these results are robust to a number of variations, such as:

### Robustness checks and alternative specifications
- Excluding the US, or running an unweighted regression, to limit the importance of large economies in the results.
- Focusing on 5-year averages to reduce year-to-year noise in the data (e.g., from errors and omissions).
- Treating BOP errors and omissions as part of valuation changes, as mismeasurement can be due to the current account or to the financial account.
- Computing valuation changes on the basis of variables denominated in local currency (instead of US dollars).
- Estimating the regressions with a system of seemingly unrelated regressions (imposing equation 2 as a constraint) and assuming panel-corrected errors.
- Excluding economies with large gross stock positions (often referred to as ’financial centers’), since large stocks and flows in these economies may reflect tax arbitrage and distort measured NFA returns.

### Implied steady-state NFA positions
- Notes: Percentage of GDP. The steady-state NFA is equal to minus the country fixed effect divided by β_NFA, from column 1 in Table 6. These implied stead-state values do not reflect, and may be different from, IMF country team projections.

### NFA dynamics and stabilizing components (key estimated coefficients and fit)
- Table 7 (autocorrelation / dynamics): Selected coefficients (all variables in percent of GDP)
  - L.nfa: -0.007** (standard error (0.003))
  - tbr: -0.064*** (0.019)
  - NFA_g: -0.016 (0.013)
  - NFA_valibπ: -0.065*** (0.016)
  - lag dep. var.: 0.830*** (0.021) in column 1; other reported lag dep. var. entries include 0.783*** and 0.455*** with respective standard errors.
  - R2 values reported across columns: 0.881 0.129 0.117 0.075 0.925 0.926
  - N observations: 1286 1236 1237 1236 1287 1237
- Table 8 (stabilizing return differentials): Selected coefficients
  - L.nfa on rYield diff.: -0.026 (0.027)
  - L.nfa on XR diff: 0.075** (0.034)
  - L.nfa on AP diff: -0.100*** (0.025)
  - L.nfa on AP liab.: -0.099*** (0.023)
  - Additional reported coefficients: 0.006 (0.004); 0.006 (0.007); 0.066** (0.027)
  - R2 values across columns: 0.081 0.045 0.193 0.78 0.092 0.156 0.05
  - N observations: 1247 in all columns
- Key interpretation:
  - Stabilizing effect of NFA returns operates through NFA valuation changes, particularly asset prices (liabilities).
  - Baseline estimate: a 10 percent of GDP higher NFA is associated with a 1 percent lower return differential. This estimate may be an upper bound; it is reduced by half in an unweighted regression, suggesting part of the result is driven by large economies.

### EMEs vs AEs: speed of adjustment
- Speed of NFA adjustment is significantly higher for (individual) EMEs than for AEs by approximately a factor of 5.
  - Implied half-lives:
    - AEs: about 13 years
    - EMEs: about 2.5 years
- For both groups, external adjustment comes from NFA returns.
- The faster adjustment for EMEs holds for both debtor and creditor EMEs (interaction coefficient insignificant).

### Reserve-currency economies: slower adjustment and stabilizing components
- Reserve-currency (RC) issuers appear to adjust at a slower pace than other economies:
  - Implied half-lives:
    - RC issuers: about 16 years
    - Other economies: about 6 years
  - Difference is only marginally statistically significant.
- Table 10 (selected coefficients):
  - Baseline L.nfa: -0.070*** (0.019) in baseline; -0.064*** (0.017) in RC specification; -0.088*** (0.018) in another column.
  - L.nfa_RC: 0.068* (0.035); -0.014 (0.023); 0.045 (0.030) in respective columns.
  - R2 values: 0.074 0.647 0.126 0.082 0.654 0.129
  - N observations: 1288 in reported columns

### NFA returns as insurance against income shocks
- Regression framework (5-year averages): y_c,t = α + β GDPgr_c,t + δ′X_c,t + κ_c + u_c,t, where dependent variable is 5-year average NFA return (percent of domestic GDP) or components.
- Insurance against domestic GDP shocks (Table 11 and Table 12):
  - Table 11 (NFA return components, 5-year intervals):
    - g_5-year coefficient on r_NFA: -0.633*** (0.136)
    - g_5-year coefficient on ibvalπ: -0.088 (0.062)
    - g_5-year coefficient on r_NFA (alternate): -0.505*** (0.153)
    - g_5-year coefficient on other component: -0.177** (0.081)
    - R2 values: 0.508 0.870 0.315 0.839
    - N: 1078 1084 1078 1084
  - Interpretation: A 1 percent drop in the 5-year average GDP growth is associated with an average positive NFA return of 0.6 percent of GDP over the same period. Effect comes largely from valuation changes, with marginal effects through inflation and income.
  - Table 12 (return differentials, 5-year intervals):
    - g_5-year on rYield diff.: -0.726*** (0.182)
    - g_5-year on XR diff: -0.091 (0.060)
    - g_5-year on AP diff: -0.200 (0.134)
    - g_5-year on another component: -0.460* (0.233)
    - R2 values: 0.586 0.877 0.262 0.538
    - N: 1077 1042 1085 1042
  - Asset price differential is the main component providing insurance, with exchange rate and yield differentials also contributing.
  - Country-level analysis (Figure 5): For the majority of countries, NFA returns provide insurance against domestic shocks; for others, effects not statistically different from zero. Applies to AEs and EMEs.

### Global insurance providers and reserve-currency roles
- Regression modified to use g_ROW (5-year average GDP growth in rest of world) with dependent variable = 5-year average NFA return expressed in percent of world or domestic GDP.
- Table 13: NFA return over World GDP (5-year intervals), selected coefficients (standard errors in parentheses)
  - g_ROW: -0.012*** (0.001) across columns
  - g_ROW_RC: 0.066*** (0.018); 0.053*** (0.001); 0.066*** (0.019); 0.231** (0.090); 0.066*** (0.018); 0.066*** (0.018)
  - g_ROW_US: 0.281*** (0.000)
  - g_ROW_EA: -0.017 (0.019)
  - g_ROW_UK: -0.178* (0.090)
  - g_ROW_JAP: -0.005 (0.018)
  - g_ROW_CHE: -0.052*** (0.018)
  - R2 values: 0.517 0.569 0.517 0.549 0.517 0.517
  - N: 870 in all columns
- Interpretation:
  - Reserve-currency countries are providers of insurance against income shocks to the rest of the world.
  - As a group, reserve-currency economies are found to transfer, on average, 0.26 percent of world GDP in wealth to non-reserve countries when GDP growth in the latter group slows down by 1 percent.
  - The US stands out, contributing with 0.32 percent of world GDP; Switzerland plays a more limited role (results partly reflect relative sizes of insurer economies).
- Table 14: NFA return over Domestic GDP (5-year intervals), selected coefficients
  - g_ROW: -0.869*** (0.129) across columns
  - g_ROW_RC: 1.529*** (0.133); 1.505*** (0.131); 1.542*** (0.134); 1.669*** (0.350); 1.528*** (0.133); 1.529*** (0.133)
  - g_ROW_US: 0.507*** (0.019)
  - g_ROW_EA: -0.596*** (0.034)
  - g_ROW_UK: -0.152 (0.325)
  - g_ROW_JAP: 0.356*** (0.029)
  - g_ROW_CHE: -0.164*** (0.030)
  - R2: 0.446 across columns
  - N: 870 in all columns
- Interpretation:
  - On average, reserve-currency economies transfer about 0.7 percent of their own GDP in response to a 1 percent income shock in the rest of the world.
  - Across individual reserve countries, this contribution ranges between 0.1 and 0.9 (US top insurer, closely followed by Japan and the UK; Switzerland below average; EA at the bottom). Range is much narrower than in Table 13 before rescaling, indicating provision of insurance broadly commensurate with economy size.

### Main conclusions (empirical findings)
- Persistent trade imbalances have been the main driving force of diverging NFA positions over 1990-2015, especially for systemic economies, with no sign of stock position–driven adjustment.
- NFA returns have played an important but insufficient stabilizing role, primarily through asset prices of countries’ liabilities.
- Cross-country heterogeneity:
  - EMEs adjust NFA positions faster than AEs (factor of ~5).
  - Reserve-currency issuers adjust at a slower pace than other economies.
- NFA returns serve as an insurance mechanism against domestic and global income shocks with macro-economically meaningful magnitudes; this insurance role is broad across countries.
- Evidence extends prior findings on the US as a global insurer to other reserve-currency countries (Euro area, UK, Japan, Switzerland), which also provide insurance against global income shocks in a scale broadly commensurate to the size of their economies.

*Source: wp1879 - Appendix A.2 shows that these results are robust to a number of variations, such as:*

### References

### References

### Bibliographic Sources
- Alberola, E., Estrada, Á., and Viani, F. (2017). Global Imbalances from a Stock Perspective. Banco de Espana Working Paper No. 1742.
- Bénétrix, A. S., Lane, P. R., and Shambaugh, J. C. (2015). International Currency Exposures, Valuation Effects and the Global Financial Crisis. Journal of International Economics, 96:S98–S109.
- Curcuru, S. E., Dvorak, T., and Warnock, F. E. (2008). Cross-Border Returns Differentials. The Quarterly Journal of Economics, 123(4):1495–1530.
- Curcuru, S. E., Thomas, C. P., and Warnock, F. E. (2013). On Returns Differentials. Journal of International Money and Finance, 36:1–25.
- Darvas, Z. and Hüttl, P. (2017). Returns on Foreign Assets and Liabilities: Exorbitant Privileges and Stabilising Adjustments. Bruegel Working Paper Issue 07.
- Devereux, M. B. and Sutherland, A. (2010). Valuation Effects and the Dynamics of Net External Assets. Journal of International Economics, 80:129–143.
- Forbes, K., Hjortsoe, I., and Nenova, T. (2017). Current Account Deficits During Heightened Risk: Menacing or Mitigating? The Economic Journal, 127:571–623.
- Ghironi, F., Lee, J., and Rebucci, A. (2015). The Valuation Channel of External Adjustment. Journal of International Money and Finance, 57:86–114.
- Gourinchas, P.-O. and Rey, H. (2007). International Financial Adjustment. Journal of Political Economy, 115(4):665–703.
- Gourinchas, P.-O. and Rey, H. (2014). External Adjustment, Global Imbalances, Valuation Effect. Handbook of International Economics, 4:585–645.
- Gourinchas, P.-O., Rey, H., Govillot, N., et al. (2010). Exorbitant Privilege and Exorbitant Duty.
- International Monetary Fund (2016). 2016 External Sector Report.
- International Monetary Fund (2017a). 2017 External Sector Report.
- International Monetary Fund (2017b). Article IV Staff Report for Ireland.
- Lane, P. R. et al. (2017). The Treatment of Global Firms in National Accounts. Economic Letters Series, 2017(1).
- Lane, P. R. and Milesi-Ferretti, G. M. (2009). Where Did All the Borrowing Go? A Forensic Analysis of the US External Position. Journal of the Japanese and International Economies, 23(2):177–199.
- Lane, P. R. and Milesi-Ferretti, G. M. (2017). The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970–2004. IMF Working Paper 17/115.
- Nguyen, H. (2011). Valuation Effects with Transitory and Trend Productivity Shocks. Journal of International Economics, 85:245–255.
- Obstfeld, M. and Rogoff, K. S. (2005). Global Current Account Imbalances and Exchange Rate Adjustments. Brookings Papers on Economic Activity, 2005:67–123.
- Pavlova, A. and Rigobon, R. (2010). An Asset-pricing View of External Adjustment. Journal of International Economics, 80:144–156.
- Tille, C. (2008). Composition of International Assets and the Long-run Current Account. Economic Notes, 37:283–313.
- Tille, C. (2017). The Changing International Linkages of Switzerland: An Overview. Graduate Institute of International and Development Studies International Economics Department Working Paper Series, No. HEIDWP14-2017.

*Source: wp1879 - References (IMF working paper references and appendix material).*

### Appendix: Robustness of Empirical Results

### A.1 Stylized Facts
- Table 15: NFA Returns and Return Differentials, Unweighted
  - Periods: 1990-2015, 1990-2007, 2008-15
  - Trade balance: 4.1, 4.0, 4.3
  - Real NFA return: 6.6, 6.3, 7.4
    - Valuation change: 6.1, 5.7, 6.8*
    - Investment income: 2.6, 2.5, 2.9**
    - Inflation: 0.8, 0.8, 0.7
  - Return differential: 7.5, 8.2, 5.9**
    - Yield diff.: 2.1, 2.0, 2.1
    - Exchange rate diff.: 2.3, 2.0, 2.3
    - Asset price diff.: 7.4, 8.4, 5.5***
  - Notes: Unweighted averages. Trade balance and NFA returns (and its components) are reported as percent of GDP. Return differentials (and their components) are reported as percentages. The last row is calculated as a residual. Asterisks indicate whether the differences across subsamples are statistically significant at the 1, 5 and 10 percent confidence levels respectively, with standard errors clustered at the country level.

### A.2 NFA-Stabilizing Mechanisms — Robustness Checks Summary
- Main regression results are maintained under the following alternative specifications:
  - i) exclude the US;
  - ii) exclude financial centers;
  - iii) remove GDP weights and winsorize observations with absolute changes in NFA/GDP larger than 20 percent;
  - iv) consider 5-year lags;
  - v) attribute BOP errors and omissions to the valuation residual;
  - vi) use denomination in local currency instead of USD;
  - vii) estimate with a system of seemingly unrelated regressions; and
  - viii) assume panel-corrected errors under different correlation structures.
- Tables reporting alternative estimations (with exact reported coefficients, standard errors, R-squared and N):
  - Table 16: Excluding US
    - L.nfa coefficients and standard errors reported (e.g., -0.071*** (0.022), 0.012 (0.013), -0.060*** (0.019), -0.015 (0.016), -0.068*** (0.016), 0.025*** (0.008), -0.017*** (0.001), -0.082*** (0.020))
    - R2 values: 0.069, 0.539, 0.119, 0.099, 0.079, 0.833, 0.901, 0.166
    - N: 1263 repeated across specifications
  - Table 17: Excluding Financial Centers (Ireland, Hong Kong SAR, Singapore and Netherlands)
    - L.nfa coefficients and standard errors reported (e.g., -0.070*** (0.023), 0.018 (0.019), -0.074*** (0.020), 0.000 (0.013), -0.082*** (0.016), 0.024** (0.012), -0.016*** (0.001), -0.109*** (0.019))
    - R2 values: 0.079, 0.603, 0.116, 0.075, 0.081, 0.800, 0.899, 0.207
    - N: 1188 repeated across specifications
  - Table 18: Unweighted, excluding observations with |∆nfa| > 20 percent of GDP
    - L.nfa coefficients and standard errors reported (e.g., -0.058*** (0.013), 0.02 (0.012), -0.049*** (0.012), -0.022 (0.014), -0.048*** (0.014), 0.017** (0.008), -0.019*** (0.001), -0.049** (0.020))
    - R2 values: 0.085, 0.527, 0.187, 0.137, 0.128, 0.818, 0.906, 0.363
    - N: 1216 repeated across specifications (one column shows N = 10)
  - Table 19: 5-Year Lag
    - L5.nfa coefficients and standard errors reported (e.g., -0.079*** (0.017), 0.016 (0.015), -0.059*** (0.017), -0.027** (0.013), -0.057*** (0.012), 0.013 (0.008), -0.014*** (0.001), -0.072*** (0.021))
    - R2 values: 0.401, 0.775, 0.518, 0.483, 0.348, 0.876, 0.932, 0.647
    - N: 1080 repeated across specifications (one column shows N = 73)
  - Table 20: BOP E&Os attributed to valuation residual
    - L.nfa coefficients: -0.072*** (0.020), -0.090*** (0.015)
    - R2: 0.143, 0.108
    - N: 1288 repeated
  - Table 21: Local-Currency Denomination (instead of USD)
    - L.nfa coefficients and standard errors reported (e.g., -0.057** (0.024), -0.060*** (0.018), -0.015*** (0.004), -0.091** (0.044))
    - R2 values: 0.117, 0.083, 0.40, 0.089
    - N: 1277, 1288, 1277, 1281 as reported
  - Table 22: SUR (Seemingly Unrelated Regressions)
    - L.nfa coefficients and standard errors reported (e.g., -0.064*** (0.013), 0.015*** (0.004), -0.065*** (0.012), -0.014** (0.006))
    - R2 values: 0.070, 0.65, 0.130, 0.10
    - N: 1288 repeated
    - Note: The coefficient on ∆nfa is imposed equal to the sum of coefficients on the trade balance, the NFA return and the growth component.
  - Table 23: Panel-Corrected Standard Errors under different correlation structures
    - Columns I–III report L.nfa coefficients and standard errors under assumptions: AR(1) errors uncorrelated across countries with common autocorrelation; heterogeneous autocorrelation across countries; contemporaneous correlation with common autocorrelation.
    - Example coefficients and standard errors reported (selected): -0.062*** (0.022), 0.006 (0.007), -0.069*** (0.022); -0.059*** (0.022), 0.003 (0.006), -0.064*** (0.022); -0.062** (0.029), 0.006 (0.009), -0.069** (0.029)
    - R2 values: 0.075, 0.319, 0.123, 0.077, 0.331, 0.134, 0.075, 0.319, 0.123 (as reported)
    - N: 1288 repeated across specifications

*Source: wp1879 - References (Appendix A tables and notes).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1879.pdf_