## Section 5: Pre- and Post‑Crisis Adjustment of Aggregate Foreign Currency Exposures ## Source details **Canonical URL:** [Section 5: Pre- and Post‑Crisis Adjustment of Aggregate Foreign Currency Exposures](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019299-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2019/wpiea2019299-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019299-print-pdf.pdf.json) --- ### Dataset construction and coverage - Builds on Lane and Shambaugh (2010a) and Bénétrix et al. (2015) but uses actual data reported to the IMF by national statistical authorities and more granular information, substantially reducing reliance on estimated currency weights. - Combines actual data sources with improved estimation methods to fill gaps and extend coverage to 1990-2017. - Focus sample: 50 countries that together account for over 90 percent of world’s GDP. - Data start adjustments: - Russia data start in 1993; Czech Republic data reported since 1993 following the split of Czechoslovakia. - Actual data sources: - IMF survey to country authorities requesting currency breakdown from 1990 across five SDR currencies (US dollar, euro, Japanese yen, Pound sterling and renminbi), domestic currency, and “other currencies”. - Coordinated Portfolio Investment Survey (CPIS) Table 2. - EU countries’ reports to the ECB. - BIS Locational Banking Statistics (LBS). - Reporting coverage: - For recent years ~52 percent of countries reported some survey data; coverage declines going backwards. ### Improvements in methodology and item treatment - Distinguishes FDI equity and FDI debt currency composition; FDI debt can represent up to 40 percent of total FDI in some countries. - Equity in FDI shares sourced from IMF International Financial Statistics (IFS); where missing, synthetic series constructed: - 1990–2008: use data from Bénétrix et al. (2015). - 2009–2017: use CDIS outward and inward stocks of direct investment for 108 reporting countries. - Portfolio equity estimation: - Uses CPIS geographical data and gravity-type regression models to impute missing country pairs for pre-2001. - Regression specification: log(1 + EQ_ijt) = ψ_j + θ_t + β Z_ijt + γ X_it + ε_ijt with reported R2 ~ 0.75 for geography imputations. - Portfolio debt synthetic coverage follows Lane and Shambaugh (2010a): geography from CPIS combined with currency composition of host countries’ international bond issuance (BIS). - Other investment proxied with IMF survey and BIS LBS (banking assets largest component). - Reserves: 1990–2012 from Bénétrix et al. (2015); 2010–2017 constructed from central bank/MoF publications and IMF COFER; series merged via interpolation. - Liabilities treatment: - Portfolio equity and FDI equity liabilities assumed in host currency (implies exposure in domestic currency). - FDI debt uses survey when available, extended using portfolio debt liability breakdown. - For 19 emerging market countries, adjust for domestic-currency issuance held by foreigners using Arslanalp and Tsuda (2014). ### Sample countries (50) - Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, China, Colombia, Czech Republic, Denmark, Egypt, Finland, France, Germany, Greece, Guatemala, Hong Kong SAR, Hungary, India, Indonesia, Ireland, Israel, Italy, Japan, Korea, Malaysia, Mexico, Morocco, Netherlands, New Zealand, Norway, Pakistan, Peru, Philippines, Poland, Portugal, Russia, Singapore, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Thailand, Tunisia, Turkey, United Kingdom, United States, Uruguay. ### Comparison of actual versus estimated currency weights — key contrasts - General: Estimated data capture main patterns but notable discrepancies exist. - Asset side: - US dollar weights are underestimated in synthetic data, particularly for portfolio equity and FDI equity. - Special Purpose Entities (SPEs) may drive further discrepancies. - Liability side: - Domestic currency weights of portfolio debt liabilities are underestimated, especially for emerging countries, because BIS does not report currency composition of domestically issued debt securities nor the share held by non-residents. - Other investment liabilities show substantial dispersion; US dollar weights slightly underestimated in early–mid 2000s. - Given these differences, prioritizing incorporation of actual data is important. ### Data products and derived indicators - Public dataset contains asset, debt asset, liability, and debt liability currency weights. - Combines shares of each IIP component from the External Wealth of Nations dataset with calculated currency weights. - Debt assets include portfolio debt, FDI debt, other investment and reserve assets. - Debt liabilities include portfolio debt, FDI debt and other investment. - Dataset enables computation of indicators: - Valuation effects combining financial exchange rates and international financial integration. - Currency composition metrics for external assets and liabilities, including US dollar and euro measures of international financial integration. - Core focus: aggregate foreign currency exposure measure FX_AGG. ### Valuation effects and financial exchange rate indices — definitions and construction - Valuation effects measure (Lane and Shambaugh, 2010a): - VAL_XR_{i,t+1} = %∆I_F_{i,t+1} IFI_{i,t} - IFI_{i,t} = (A_{i,t} + L_{i,t})/GDP_{i,t} - I_F_{i,t+1} = I_F_{i,t} (1 + ∑ ω^F_{ij,t} %∆E_{ij,t}) - ω^F_{ij,t} = ω^A_{ij,t} s^A_{i,t} − ω^L_{ij,t} s^L_{i,t}; s^A_{i,t} = A_{i,t}/(A_{i,t}+L_{i,t}); s^L_{i,t} = L_{i,t}/(A_{i,t}+L_{i,t}) - Interpretation: - Depreciation of domestic currency may be amplified or mitigated by valuation changes depending on foreign-currency long/short positions and external balance sheet size. ### Empirical comparisons: financial vs trade-weighted exchange rates and valuation effects - Trade-weighted exchange rates are limited for assessing wealth effects on external balance sheets. - Financial indices and trade-weighted indices can move in opposite directions; financial indices are crucial for assessing sustainability of external imbalances. - Confirmatory findings for 50-country sample (1990–2017): - Heterogeneity and weak correlations with trade-weighted indices. - United States: financial exchange rate index shows much less volatility than trade-weighted index. - Colombia: high comovement but financial index lower volatility since 2000s reflecting domestic currency liabilities. - Germany: financial index lower volatility than trade-weighted index. - Table 1 highlights: - Asset and liability financial indices exhibit high correlation. - Correlations between financial indices and trade-weighted index are weak. - Financial exchange rate volatility significantly lower than trade-weighted index overall; financial index for emerging and developing economies is more volatile than for advanced economies. - Validation: VAL_XR tracks valuation changes of external net financial assets and liabilities (example: Germany series tracks Deutsche Bundesbank series). ### Aggregate foreign currency exposure — definition and interpretation (FX_AGG) - FX_AGG_i,t = ω^F_Ai,t s_Ai,t − ω^F_Li,t s_Li,t - Ranges between -1 and 1. - Positive: country “long on foreign currency”; Negative: country “short on foreign currency”. - Uniform valuation effect: - VAL_XR,U_i,t+1 = FX_AGG_i,t IFI_i,t %∆E_U_i,t+1 - Data limitations: - Dataset does not capture currency denomination of cross-border financial derivatives nor extent of hedging. - Net financial derivatives positions tend to be less than 1 percent of GDP. - Local (within-resident) foreign currency positions data are sparse. ### Decomposition of FX_AGG and main components - FX_AGG = (A−L)/(A+L) + FX_AGG^o - (A−L)/(A+L): net foreign asset position scaled by external balance sheet. - FX_AGG^o: net liability position in domestic currency (L_DC − A_DC) as proportion of external balance sheet; "foreign currency mix". - Further: - (A−L)/(A+L) = (A_NR − L)/(A+L) + FXR/(A+L) - FX_AGG^o = PEQL_DC + FDIL_DC /(A+L) + DEBTL_DC /(A+L) − A_DC_NR /(A+L) - PEQL_DC: portfolio equity liabilities in domestic currency. - FDIL_DC: FDI equity liabilities in domestic currency. - DEBTL_DC: debt liabilities in domestic currency. - A_DC_NR: non-reserve assets in domestic currency. - Group patterns: - Advanced economies: FX_AGG^o mainly driven by debt liabilities in domestic currency, followed by equity liabilities. - Emerging economies: FX_AGG^o mainly driven by equity component of liabilities; domestic-currency debt component small but positive. - Stylized interplay: - Advanced economies: negative link between (A−L)/(A+L) and FX_AGG^o, suggesting debtor countries hedge negative net positions with long foreign currency mix. - Emerging economies: positive correlation for many countries; net debtor position can be exacerbated by a short foreign currency mix. ### Summary statistics (2017 snapshot, full sample and groups) - All countries (2017): - FXAGG: Mean 0.14; Median 0.18; p25 0.06; p75 0.32. - (A−L)/(A+L): Mean -0.13; Median -0.06; p25 -0.26; p75 0.05. - FXAGGo: Mean 0.27; Median 0.28; p25 0.20; p75 0.36. - (PEQL+FDIL)/(A+L): Mean 0.24; Median 0.24; p25 0.16; p75 0.31. - DEBTL_DC/(A+L): Mean 0.12; Median 0.07; p25 0.04; p75 0.14. - A_DC_NR/(A+L): Mean 0.08; Median 0.02; p25 0.00; p75 0.10. - Advanced (2017): - FXAGG: Mean 0.17; Median 0.11. - (A−L)/(A+L): Mean -0.03; Median 0.00. - FXAGGo: Mean 0.20; Median 0.20. - DEBTL_DC/(A+L): Mean 0.20; Median 0.15. - A_DC_NR/(A+L): Mean 0.18; Median 0.14. - Emerging and Developing (2017): - FXAGG: Mean 0.12; Median 0.18. - (A−L)/(A+L): Mean -0.20; Median -0.13. - FXAGGo: Mean 0.33; Median 0.34. - (PEQL+FDIL)/(A+L): Mean 0.28; Median 0.28. - DEBTL_DC/(A+L): Mean 0.05; Median 0.05. - A_DC_NR/(A+L): Mean 0.01; Median 0.00. ### Key stylized facts and implications - Cross-country shift toward long foreign-currency positions between 1990 and 2017. - 60 percent of countries had net negative positions in foreign currencies in 1990; proportion declined to 20 percent in 2017. - Emerging markets dominated net negative positions historically; in 1990 emerging economies accounted for 22 out of 29 short-position countries; in 2017 they represent 9 of the 10 short countries. - In 2017 the only advanced economy with negative positions is Greece. - Drivers of the shift: - Improvement in net international investment positions (A−L)/(A+L), largely from non-reserve components. - Increase in equity liabilities denominated in domestic currency (FX_AGG^o), especially in emerging economies. - Policy-relevant implications: - Countries with negative net positions in foreign currency can suffer significant negative wealth effects from depreciation. - Shifts toward issuance of foreign liabilities in domestic currency can materially alter wealth effects from currency movements. - For many emerging economies, recessions with depreciating currencies can yield capital gains on net external position, providing a partial hedge. ### Long-term dynamics and distributional shifts (1990–2017) - Cumulative distribution shifts: - 1990 vs 2017: rightward shift in FX_AGG; 35 percent of countries were net creditors in 2017 vs 20 percent in 1990. - Size of net debtor positions (scaled by A+L) shrank substantially. - Component dynamics (1990→2017): - (A_NR − L)/(A+L) improvement drove net IIP improvement. - Foreign currency mix FX_AGG^o moved towards longer positions. - Equity liabilities in domestic currency increased substantially, major contributor to FX_AGG^o shift; especially marked in emerging economies. - Distribution of debt liabilities in domestic currency similar in 1990 and 2017 for around 80 percent of countries. ### Pre- and post-crisis adjustments (descriptive evidence, 1997–2017) - Distributional prevalence: - 1997: 48 percent of countries had negative FX_AGG. - 2007: 14 percent negative. - 2012: 24 percent negative. - 2017: 20 percent negative. - Scale of adjustments: - 1997–2007 adjustment ranges from -0.2 to +0.6. - 2007–2017 adjustment ranges from -0.2 to +0.2. - Pre-crisis (1997–2007): - Negative relation between initial FX_AGG and subsequent change: countries short in 1997 improved by 2007 (exception: Sri Lanka). - Emerging markets improved markedly from larger initial short positions. - Post-crisis (2007–2017): - Relation between initial exposure and change changed: advanced economies mostly moved to larger long positions; trend line flat for advanced group. - Emerging economies: relation slightly positive; some with positive FX_AGG in 2007 moved further long (e.g., Philippines); some with negative FX_AGG in 2007 moved further short (Turkey, Sri Lanka, Pakistan, Tunisia). - Egypt and Morocco moved from long in 2007 to short in 2017. - Overall: pre-crisis period marked strong adjustments toward longer FX_AGG; post-crisis period marked persistence and some divergence across emerging countries. ### Regression analysis: determinants and associations (pooled regressions for 1997, 2007, 2017) - Model and covariates: - Dependent variables: FX_AGG and subcomponents ((A−L)/(A+L), FX_AGG^o, (A_NR − L)/(A+L), FXR/(A+L), (PEQL_DC + FDIL_DC)/(A+L), DEBTL_DC/(A+L), A_DC_NR/(A+L)). - Core covariates: Trade, vol(GDP), vol(π), I (financial exchange rates index), corr(GDP,I), KA open (Chinn and Ito), Peg (Shambaugh fixed exchange rate dummy), EMU membership, Size (log population), GDPpc, year dummies 2007 and 2017. - Model performance: - Explains more than 50 percent of cross-country variation in aggregate foreign-currency exposures; R^2 across subcomponents ranges from 0.40 to 0.90. - Main empirical findings (All countries): - Trade openness positively associated with long FX_AGG (0.07*** (0.02) in Table 4 column (1)); effect driven by net external positions (A−L)/(A+L). - vol(GDP) positively associated with FX_AGG (0.33** (0.15) in column (1)). - Exchange rate volatility significantly associated with short aggregate positions in foreign currency; negative link with domestic-currency debt liabilities and non-reserve assets. - corr(GDP,I) positively associated with long FX_AGG (stabilizing channel). - KA open not associated with aggregate FX exposure overall but negatively associated with FX_AGG^o. - EMU membership associated with short aggregate foreign-currency positions (EMU coefficient -0.20*** (0.03) in column (1)); linked to lower reserve assets and larger external assets denominated in domestic currency. - Size and GDPpc positively associated with long FX_AGG via net positions; GDPpc negatively related with reserves. - Year 2017 dummy: 0.09** (0.04) in column (1) indicating rightward shift toward longer FX_AGG. - Country group differences (Advanced vs Emerging): - Trade openness effect driven by advanced economies (Trade 0.10* (0.05) in advanced sample). - Exchange rate volatility effects concentrated in emerging markets. - corr(GDP,I) strong and significant in emerging countries (0.12*** (0.04) in Table 6 column (1)); not significant in advanced group. - GDP per capita positively associated with FX_AGG in both groups. - Year dummies: advanced economies show limited conditional movement; emerging economies show movement toward long FX_AGG^o in 2017 and positive association of 2007/2017 with equity liabilities in domestic currency. - Robustness checks: results robust to country fixed effects, alternative IFI measures, financial deepening proxies, financial center dummy, alternative volatility measures. ### Key takeaways (concise) - Cross-country distribution shifted toward long foreign-currency exposures between 1990 and 2017 (from 60 percent net negative in 1990 to 20 percent in 2017). - Main drivers: improvement in net international investment positions (excluding reserves) and increases in equity liabilities denominated in domestic currency, especially in emerging economies. - Trade openness and output volatility associated with longer FX_AGG (trade effect driven by advanced economies). - Exchange rate volatility and EMU membership associated with shorter aggregate foreign-currency positions (exchange rate volatility effects driven by emerging economies). - For many emerging economies, depreciations in recessions can yield capital gains on net external position, acting as partial hedge. ### Appendix A — data sources and estimation methods (high-level) - Actual data sources: IMF survey (52 countries invited; response rate 85 percent), CPIS Table 2, ECB reporting, BIS LBS. - A.2 Combining actual and estimated data: - Use actual data when available; splice backward with estimated currency weights to create 1990-2017 series (Russia and Czech Republic start in 1993). - Financial derivatives excluded due to unavailable currency breakdowns. - Item-specific methods summarized: - Portfolio equity: CPIS + gravity regression imputations; R2 ~ 0.75. - FDI: split equity/debt using IFS; CDIS used for 2009–2017; gravity model for gaps. - Portfolio debt: geography + host-country issuance currency from BIS IDS; netting out domestic holders using US Treasury, ECB, BoJ data. - Other investment: IMF survey + BIS LBS reported and mirror data approaches; hierarchy: survey → BIS reported → BIS mirror. - Reserves: Bénétrix et al. (2015) 1990-2012 merged with constructed 2010-2017 series using COFER and central bank publications; merged via linear interpolation. - Accuracy assessments: - Correlations between actual and estimated series high for several items (e.g., other investment assets US dollar 0.94, euro 0.91). - Noted systematic biases: synthetic data tend to underestimate US dollar weights for many asset items and underestimate domestic-currency share of portfolio debt liabilities for emerging markets. _italics: Source: wpiea2019299-print-pdf — Section 5 (and related sections 3.2, 4.2, 5.1, Appendix A) of the supplied IMF chapter excerpt_ ### Section 5 discusses the pre- and post- crisis adjustment of aggregate foreign currency exposures ### Section 5: Pre- and Post‑Crisis Adjustment of Aggregate Foreign Currency Exposures ### Dataset construction and coverage - Builds on Lane and Shambaugh (2010a) and Bénétrix et al. (2015) but uses actual data reported to the IMF by national statistical authorities and more granular information, substantially reducing reliance on estimated currency weights. - Combines actual data sources with improved estimation methods to fill gaps and extend coverage to 1990-2017. - Focus sample: 50 countries (part of the IMF External Balance Assessment and/or External Sector Report) that together account for over 90 percent of world’s GDP. - Data start adjustments: Russia data start in 1993; Czech Republic data reported since 1993 following the split of Czechoslovakia. - Actual data sources: - IMF survey to country authorities requesting currency breakdown from 1990 across five SDR currencies (US dollar, euro, Japanese yen, Pound sterling and renminbi), domestic currency, and “other currencies”. - Coordinated Portfolio Investment Survey (CPIS) Table 2. - EU countries’ reports to the ECB. - BIS Locational Banking Statistics (LBS). - For recent years ~52 percent of countries reported some survey data; coverage declines going backwards. ### Improvements in methodology and item treatment - Distinguishes FDI equity and FDI debt currency composition (actual data show different currency mixes), improving on prior studies that treated FDI as equity only. - Equity in FDI shares sourced from IMF International Financial Statistics (IFS); where missing, synthetic series constructed: - 1990–2008: use data from Bénétrix et al. (2015). - 2009–2017: use CDIS outward and inward stocks of direct investment for 108 reporting countries. - Portfolio equity estimation uses CPIS geographical data (assumes equity issued by a country is denominated in that country’s currency) and gravity-type models to impute missing country pairs for pre-2001. - Portfolio debt synthetic coverage follows Lane and Shambaugh (2010a): geography from CPIS combined with currency of host countries’ bonds issued in international markets. - Other investment (mainly bank-related) proxied with IMF survey and BIS LBS (banking assets largest component). - Reserves: 1990–2012 from Bénétrix et al. (2015); 2010–2017 constructed from central bank/MoF publications and IMF COFER; series merged via interpolation. - Liabilities: portfolio equity and FDI equity liabilities assumed in host currency (implies exposure in domestic currency). FDI debt uses survey when available, extended using portfolio debt liability breakdown. - For portfolio debt liabilities, BIS international debt issuance statistics used; for 19 emerging market countries, adjust for domestic-currency issuance held by foreigners using Arslanalp and Tsuda (2014). ### Sample countries - Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, China, Colombia, Czech Republic, Denmark, Egypt, Finland, France, Germany, Greece, Guatemala, Hong Kong SAR, Hungary, India, Indonesia, Ireland, Israel, Italy, Japan, Korea, Malaysia, Mexico, Morocco, Netherlands, New Zealand, Norway, Pakistan, Peru, Philippines, Poland, Portugal, Russia, Singapore, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Thailand, Tunisia, Turkey, United Kingdom, United States, Uruguay. ### Comparison of actual versus estimated currency weights — key contrasts - Overall, estimated data capture main patterns of actual data, but notable discrepancies exist: - Asset side: - US dollar weights are underestimated in synthetic data, particularly for portfolio equity and FDI equity (synthetics rely on geography; some cross-border positions denominated in US dollars are not captured by geography). - Special Purpose Entities (SPEs) may drive further discrepancies. - Liability side: - Domestic currency weights of portfolio debt liabilities are underestimated, especially for emerging countries, because BIS does not report currency composition of domestically issued debt securities nor the share held by non-residents. - Other investment liabilities show substantial dispersion; US dollar weights slightly underestimated in early–mid 2000s (other investment proxied mainly by banking data). - Given these differences, prioritizing incorporation of actual data is important. ### Data products and derived indicators - Publicly available dataset contains asset, debt asset, liability, and debt liability currency weights. - Combines shares of each IIP component from the External Wealth of Nations dataset (Lane and Milesi‑Ferretti, 2007) with calculated currency weights. - Debt assets include portfolio debt, FDI debt, other investment and reserve assets. - Debt liabilities include portfolio debt, FDI debt and other investment. - Dataset enables computation of indicators used in analysis: - A measure of valuation effects in response to exchange rate movements combining financial exchange rates and international financial integration. - Currency composition metrics for external assets and liabilities, including US dollar and euro measures of international financial integration. - Core focus: a measure of aggregate foreign currency exposure. ### Valuation effects and financial exchange rate indices — definitions and construction - Valuation effects measure (as proposed by Lane and Shambaugh, 2010a): - V AL_XR_{i,t+1} = %∆I_F_{i,t+1} IFI_{i,t} - Where V AL_XR_{i,t+1} indicates the currency-induced valuation change related to currency movements for country i; %∆I_F_{i,t+1} is the percentage change in the net financial exchange rate index during period t+1; IFI_{i,t} denotes the size of the external balance sheet measured by the international financial integration index. - IFI_{i,t} = (A_{i,t} + L_{i,t})/GDP_{i,t}, where A_{i,t} and L_{i,t} are end-of-period gross stocks of external assets and liabilities expressed in current US dollars. - Financial exchange rate index dynamics: - I_F_{i,t+1} = I_F_{i,t} (1 + ∑ ω^F_{ij,t} %∆E_{ij,t}) - %∆E_{ij,t} is the percentage change in the bilateral end-of-period nominal exchange rate between currency of country i and foreign currency j between t and t+1. - ω^F_{ij,t} = ω^A_{ij,t} s^A_{i,t} − ω^L_{ij,t} s^L_{i,t}, where ω^A_{ij,t} (ω^L_{ij,t}) is the proportion of assets (liabilities) denominated in currency j; s^A_{i,t} = A_{i,t}/(A_{i,t}+L_{i,t}); s^L_{i,t} = L_{i,t}/(A_{i,t}+L_{i,t}). By construction s^A_{i,t} + s^L_{i,t} = 1. - Interpretation: - Depreciation of domestic currency associated with net financial outflow (via interest parity) may be amplified or mitigated by valuation changes in gross external positions depending on whether the country is long or short in foreign currencies and the size of the external balance sheet. - Financial exchange rate indices are constructed using net foreign currency exposures and provide a direct measure of sensitivity of external balance sheets to currency movements. ### Empirical comparisons: financial vs trade-weighted exchange rates and valuation effects - Trade-weighted exchange rates are limited for assessing wealth effects on external balance sheets because: - They proxy financial account via current account rather than focusing on the financial account itself. - Geography of trade does not necessarily match investment counterparties or currencies of denomination. - They do not provide a one-to-one link between change in value of external positions and exchange rate movements. - Financial and trade-weighted indices can move in opposite directions; financial indices are crucial for assessing sustainability of external imbalances. - Confirmatory findings: - Computing financial exchange rates for the 50-country sample (1990–2017) confirms heterogeneity and weak correlations with trade-weighted indices documented in Lane and Shambaugh (2010a). - Examples: - United States: financial exchange rate index shows much less volatility than trade-weighted index (due to low foreign-currency component of US external liabilities and trade index weighting of partners not aligned with investment destinations). - Colombia: both indices show high comovement but financial index exhibits substantially lower volatility since the 2000s reflecting growing importance of domestic currency in liabilities. - Germany: financial index exhibits lower volatility than trade-weighted index, typical for advanced economies with significant domestic currency liabilities and investment partners differing from trading partners. - Table 1 (described) findings: - Asset and liability financial indices exhibit high correlation (smaller for emerging and developing economies). - Correlations between financial indices and trade-weighted index are weak. - Volatility: financial exchange rate volatility significantly lower than trade-weighted index overall; financial index for emerging and developing economies is more volatile than for advanced economies. - Validation: valuation effects measure V AL_XR tracks valuation changes of external net financial assets and liabilities (example of Germany compared with Deutsche Bundesbank series shows close tracking, including valuation loss after the global financial crisis and subsequent recovery). ### Implications and analytical use - The calculated financial exchange rates and valuation effects quantify effects of exchange rate movements on external balance sheets. - For countries with negative net positions in foreign currency, depreciation of domestic currency can produce significant negative wealth effects (prototype: developing country). - Shifts toward greater issuance of foreign liabilities in domestic currency (increasing equity and domestic-currency debt issuance) in many developing countries can materially alter wealth effects from currency movements. - The formula for valuation effects (Equation 1) can be applied across the sample to assess country-specific exposure and wealth impact. *Source: wpiea2019299-print-pdf — Section 5 discusses the pre- and post- crisis adjustment of aggregate foreign currency exposures* ### 3.2 International Financial Integration ### 3.2 International Financial Integration ### Trends in the IFI measure and currency composition - The IFI measure (sum of external assets and external liabilities scaled by the weighted average of each country’s GDP) increased two-folds from early 1990s levels to 360 percent of GDP by 2017 and exhibits a decline in the aftermath of the global financial crisis. - As of 2017, around seventy percent of the cross-border holdings are denominated in US dollars or euros. - In the full sample: - Euro-denominated cross-border holdings expanded rapidly between 1999 and 2007, but declined considerably after the global financial crisis. - US dollar positions dropped in 2008, but quickly recovered and showed an upward trend, gaining a clear predominance over the euro since 2014. - Excluding the US and the Euro Area (dotted lines), the shift away from the euro and into the dollar is still present; US dollar cross-border positions were three times larger in 2017. - For advanced economies: - The IFI measure in US dollars for the sample of advanced economies (solid blue line) is very similar to advanced economies excluding the US and the Euro Area (dotted blue line). - There is a substantial decline in positions in euros, especially when excluding the US and Euro Area. - For emerging economies: - Emerging markets account for a small but growing share of cross-border holdings. - The majority of cross-border holdings of assets and liabilities are denominated in US dollars; the euro plays a muted role. - Factors driving the dollar’s dominance include: - Euro Area banks persistently deleveraged from cross-border positions since the crisis. - Uncertainty from the Euro Area sovereign debt crisis led investors to shift away from euro positions. - US dollar appreciation and high liquidity of dollar assets during the crisis, and a lack of supply of safe euro-denominated assets relative to the US dollar. ### Foreign currency exposure: definition and use - Aggregate foreign currency exposure indicator (FX_AGG_i,t): - FX_AGG_i,t = ω^F_Ai,t s_Ai,t − ω^F_Li,t s_Li,t - s_Ai,t and s_Li,t are defined earlier and capture whether the country exhibits a creditor or debtor external position. - ω^F_Ai,t and ω^F_Li,t are the proportions of assets and liabilities denominated in foreign currency. - Indicator range and interpretation: - Ranges between -1 (all liabilities in foreign currency and all assets in domestic currency) and 1 (all assets in foreign currency and all liabilities in domestic currency). - A country is “long on foreign currency” if FX_AGG is positive and “short on foreign currency” if it is negative. - FX_AGG captures sensitivity of a country’s external position to a uniform appreciation or depreciation of its currency vis-à-vis all other currencies. - Uniform valuation effect formula: - VAL_XR,U_i,t+1 = FX_AGG_i,t IFI_i,t %∆E_U_i,t+1 - %∆E_U denotes a uniform shift in the value of the domestic currency against all foreign currencies. - Data limitations noted: - Dataset does not capture currency denomination of cross-border financial derivatives nor extent of derivatives used to hedge cross-border currency mismatches. - Net financial derivatives positions tend to be a minor component of the external balance sheet (less than 1 percent of GDP). - Local (within-resident) foreign currency positions data are sparse; partial hedging between cross-border and local exposures observed in banking sectors of some advanced economies. ### Components and decomposition of FX_AGG - Aggregate decomposition: - FX_AGG = (A−L)/(A+L) + FX_AGG^o - First component: net foreign asset position (A−L) scaled by external balance sheet (A+L). - FX_AGG^o: net liability position in domestic currency (L_DC − A_DC) as proportion of external balance sheet; referred to as the foreign currency mix. - FX_AGG^o can be interpreted as aggregate foreign currency exposure evaluated at a zero net foreign asset position. - Illustrative example: - If ω^F_Ai,t = ω^F_Li,t = 0.6 and (A−L)/(A+L) = −0.6, then FX_AGG = −0.36. - Further decomposition of net position: - (A−L)/(A+L) = (A_NR − L)/(A+L) + FXR/(A+L) - A_NR are foreign assets excluding reserves; FXR denotes foreign exchange reserves. - Net position dominated by non-reserve component. - Advanced economies mean value of (A_NR − L)/(A+L) is -0.04; for emerging countries it is -0.33. - Foreign exchange reserves represent a larger share of external assets in emerging countries. - Decomposition of FX_AGG^o: - FX_AGG^o = PEQL_DC + FDIL_DC /(A+L) + DEBTL_DC /(A+L) − A_DC_NR /(A+L) - Definitions: - PEQL_DC: portfolio equity liabilities denominated in domestic currency. - FDIL_DC: equity component of direct investment liabilities denominated in domestic currency. - DEBTL_DC: debt liabilities denominated in domestic currency. - A_DC_NR: non-reserve assets denominated in domestic currency. - Drivers by country group: - Advanced economies: FX_AGG^o mainly driven by debt liabilities in domestic currency, followed by equity liabilities (FDI equity and portfolio equity). - Emerging economies: FX_AGG^o mainly driven by the equity component of liabilities; domestic-currency debt component is small but positive. - Noted trend: emerging market governments increasingly borrowed in their own currency over the last decade, consistent with Alfaro and Kanczuk (2019). External debt liabilities still include significant foreign currency components (FDI debt and other investment). ### Summary statistics and correlations (1990–2017, snapshot in 2017) - In 2017 summary statistics (full sample of 50 countries, advanced and emerging groups): - Mean and median in all country groupings show a long foreign currency position, FX_AGG > 0, ranging from 0.11 to 0.18. - Some emerging economies show a short position. - Considering components: - Many countries exhibit a debtor position with (A−L)/(A+L) ranging from -0.13 to -0.2 across mean and median countries. - All statistics show a long position for the foreign currency mix, FX_AGG^o > 0. - Long foreign currency exposures in 2017 were mostly the result of the currency mix of assets and liabilities rather than net creditor positions. - Interquartile range shows more cross-country variation in (A−L)/(A+L) than in FX_AGG^o. - Correlation patterns (Table 3 summary): - (A−L)/(A+L) and FX_AGG^o are positively correlated with FX_AGG. - FX_AGG is more responsive to changes in (A−L)/(A+L) than in FX_AGG^o. - Mean and median correlations between (A−L)/(A+L) and FX_AGG^o are very small (negative and positive signs respectively), but interquartile range is wide (large negative and positive correlations). - Strong positive correlation between (A−L)/(A+L) and (A_NR − L)/(A+L). - Correlation between (A−L)/(A+L) and FXR/(A+L) is positive at mean and median but negative correlations appear in the interquartile range. - For FX_AGG^o: - Equity liability components ((PEQL + FDIL)/(A+L)) are highly correlated with FX_AGG^o (mean 0.76; median 0.97). - DEBTL_DC/(A+L) has a positive but not uniformly strong relation with FX_AGG^o; interquartile range from −0.25 (25th percentile) to 0.74 (75th percentile). - A_DC_NR/(A+L) correlation with FX_AGG^o is negative by construction; mean and median correlations are negative but small in absolute value. ### Key stylized facts and implications - Mean and median countries show long foreign currency positions in 2017, driven by a positive foreign currency mix, FX_AGG^o. - In advanced economies this long FX_AGG^o is mainly due to debt liabilities in domestic currency; in emerging economies the key driver is equity liabilities. - Stylized fact on correlation: - Advanced economies show a negative link between (A−L)/(A+L) and FX_AGG^o, suggesting debtor countries hedge negative net positions with a long foreign currency mix; exchange rate movements can improve the net external position. - Emerging economies show a positive correlation between these items for at least half of the countries in the group; for them a net debtor position is exacerbated by a short foreign currency mix, so a depreciation of the exchange rate would act as a destabilizing force, increasing a net debtor position further. - Interpretation: - Emerging countries with negative net external positions may be deemed inherently more risky, reducing their ability to issue liabilities in domestic currency. - Foreign exchange reserves are more strongly associated with (A−L)/(A+L) and FX_AGG^o in emerging countries. *Source: 3.2 International Financial Integration (from the supplied IMF chapter excerpt)* ### 4.2 Long-term dynamics ### 4.2 Long-term dynamics ### Overview - Figure 6: cumulative cross-country distribution of FX_AGG positions at the start and end of sample for advanced (red) and emerging economies (blue). - Shift towards long positions in foreign currency since 1990, consistent with Lane and Shambaugh (2010a) and Bénétrix et al. (2015). - 60 percent of countries in sample exhibited net negative positions in foreign currencies in 1990; this proportion declined to 20 percent in 2017. - Emerging market economies dominate net negative positions in foreign currencies historically (“original sin”); in 1990 emerging economies accounted for 22 out of the 29 countries with short foreign currency positions, and in 2017 they represent 9 of the 10 countries short in foreign currency. In 2017 the only advanced economy with negative positions is Greece. ### Decomposition of FX_AGG - Figure 7: cumulative distribution of components of FX_AGG (Equation 6). - The rightward shift in FX_AGG between 1990 and 2017 was mainly driven by improvement in the net international investment positions, (A−L)/(A+L). - 35 percent of the countries in the sample were net creditors in 2017, compared to 20 percent in 1990. - Size of net debtor positions (scaled by total external assets and liabilities) shrank substantially. - Foreign currency mix, FX_AGG^o, moved towards longer positions since 1990, contributing to the rightward shift. - Figure 8: decomposition of (A−L)/(A+L) (Equation 7) into (A_NR − L)/(A+L) and FXR/(A+L). - Improvement in net foreign assets (scaled by A+L) since 1990 was overwhelmingly driven by the non-reserve components of the IIP. - Emerging economies experienced an expansion in foreign reserves since the mid-1990s, but reserves played a limited role in the overall improvement of the net IIP. - Figure 9: decomposition of FX_AGG^o (Equation 8) into equity liabilities in domestic currency (portfolio and FDI), domestic currency debt liabilities, and external assets denominated in domestic currency (all scaled by A+L). - Asset component in domestic currency unchanged for most of sample since 1990, but increased substantially for Euro Area countries due to euro adoption (redenomination of cross-border assets from foreign to domestic currency). - Overall rightward shift mainly driven by larger equity liabilities denominated in domestic currency; importance even more marked for emerging economies. - Distribution of debt liabilities in domestic currency looks very similar in 1990 and 2017 for around 80 percent of countries. - Key interpretation: - Improvement in foreign currency exposure in emerging economies dominated by increase in equity liabilities in domestic currency and to a lesser extent increase in portfolio debt liabilities in domestic currency. - Overall shift since 1990 mainly driven by improving net international investment positions (excluding reserves) and to a lesser extent larger net liability positions in domestic currency (driven by larger equity liabilities). ### Regression Analysis (Section 4.3) - Objective: assess associations between foreign currency exposures and macro variables (trade openness, macro risk, exchange rate regimes, country size and development). - Data and model: - Pooled regressions using three years of data: 1997, 2007 and 2017 (Table 4). - Core covariates: Trade (exports + imports scaled by GDP), vol(GDP) (coefficient of variation over ten-year non-overlapping window), vol(π) (inflation volatility), I (financial exchange rates index as in equation (2)), corr(GDP,I) (correlation coefficient between GDP and financial exchange rate index). - Additional controls: KA open (Chinn and Ito (2006) de jure financial openness), Peg (Shambaugh (2004) fixed exchange rate dummy), EMU membership dummy (zero for all countries in 1997, one for member countries in 2007 and 2017), Size (log population), GDPpc, year dummies for 2007 and 2017. - Dependent variables: FX_AGG and subcomponents: (A−L)/(A+L), FX_AGG^o, (A_NR − L)/(A+L), FXR/(A+L), (PEQL_DC + FDIL_DC)/(A+L), DEBTL_DC/(A+L), A_DC_NR/(A+L). - Model performance: - Model explains more than 50 percent of cross-country variation in aggregate foreign-currency exposures. - R^2 across subcomponents ranges from 0.40 to 0.90. - Main empirical findings: - Trade openness is positively associated with long positions in foreign currency (driven by effect on net external positions, (A−L)/(A+L), not FX_AGG^o). Trade is unrelated to FX_AGG^o via reserves (column (5)). - Trade openness: - Negatively associated with debt liabilities in domestic currency. - Positively associated with non-reserve assets in foreign currency. - No uncovered relation with equity liabilities (column (6)). - GDP volatility: - Positively associated with long position in foreign currency in aggregate (column (1)); driven by longer foreign currency positions in non-reserve assets (column (8)). - Inflation volatility: no significant effect on aggregate exposure and main subcomponents; small positive link with FX reserves. - Exchange rate volatility: - Significantly associated with short aggregate positions in foreign currency. - Negative link between exchange rate volatility and both domestic currency denominated debt liabilities and non-reserve assets (columns (7) and (8)). - Correlation between GDP and financial exchange rate (corr(GDP,I)): - Positively associated with long aggregate position in foreign currency — interpreted as stabilizing: recessions associated with depreciation would yield capital gains on net external position. - Effect driven by net position (A−L)/(A+L), mostly via net position excluding reserves (column (4)). - KA open (de jure financial openness): - Not associated with aggregate FX exposure due to counterbalancing effects. - Negatively and statistically related with FX_AGG^o only. - Negatively associated with reserve assets and equity liabilities (columns (5) and (6)). - Positively associated with debt liabilities and non-reserve assets in domestic currency (columns (7) and (8)). - Peg (fixed exchange rate dummy): - Not statistically significant for FX_AGG and main subcomponents overall. - Positive link with reserves and negative link with equity liabilities in domestic currency. - EMU membership: - Strong role for aggregate currency exposures: associated with short aggregate foreign-currency positions (column (1)), driven by negative links with (A−L)/(A+L) and FX_AGG^o. - Effect driven by lower reserve assets for Euro Area countries (column (5)). - Statistically significant links with all subcomponents (columns (6)-(8)); larger external assets denominated in domestic currency (the euro) lead to lower net liabilities in domestic currency. - Size and development: - Size and GDP per capita positively associated with long aggregate foreign currency exposures via net positions (A−L)/(A+L), not FX_AGG^o. - GDP per capita negatively related with stock of foreign exchange reserves. - Larger and richer countries hold more debt liabilities and non-reserve assets denominated in domestic currency. - Year dummies: - Confirm rightward shift toward longer foreign currency position unconditionally and conditional on covariates. - Conditional movement in aggregate currency exposures explained by a foreign currency mix effect (column (3)), not by net foreign asset position for full sample. - 2007 associated with reduction in non-reserve net position; 2017 associated with increase in foreign exchange reserves. - Year dummies positively associated with equity liabilities in domestic currency and negatively with external assets denominated in domestic currency. - Robustness checks: - Alternative specifications including country fixed effects, IFI indicator for financial integration, proxies for financial deepening (stock market capitalization, credit and deposits), financial center dummy, alternative volatility measures — results remain robust. ### Country Group Comparison: Advanced vs. Emerging Economies (Section 4.3.1) - Motivation: issuers of dominant currencies are in advanced group; potential fundamental differences justify separate analysis (Tables 5 and 6). - Trade openness effect: - Positive effect between trade openness and long foreign currency exposures dominated by advanced economies; not significant for emerging markets (column (1)). - GDP volatility: - Country-split does not reveal clear dominance; point estimates insignificant in both groups but signs differ (negative in advanced, positive in emerging). - Advanced economies: GDP volatility yields negative and significant coefficient for (A−L)/(A+L) and positive for FX_AGG^o — associated with smaller net creditor positions (column (4)) and no significant relation with reserves (column (5)). - Emerging economies: GDP volatility associated with more positive external positions (column (2)) and positively significant effect on non-reserve domestic assets denominated in domestic currency (column (8)). - Inflation volatility: - Advanced group: associated with long foreign currency mix, higher reserves, and larger equity liabilities in domestic currency (columns (3), (5), (6)). - Emerging group: positive association between inflation volatility and reserves (column (5)). - Exchange rate volatility: - Significant only for emerging markets in the split; effects in full sample likely driven by emerging group. - Relation comes from link between exchange rate volatility and reserve assets (column (5)). - Stabilizing role (corr(GDP,I)): - Strong effect in emerging countries driven by effect on net position (column (2)); not significant in advanced countries. - Size and GDP per capita: - Present in both groups. - Advanced: GDP per capita associated with long foreign currency position via net positions (driven by (A_NR − L)/(A+L) in column (4)); effect on FX_AGG^o explained by negative association between GDP per capita and debt liabilities in domestic currency (column (7)). - Emerging: positive relation with FX_AGG driven by net position and unrelated to reserves. - Year dummies: - Advanced economies: no statistically significant relation for aggregate and net positions; 2017 dummy negatively linked to non-reserve net positions (column (4)). - Emerging economies: movement toward long foreign currency mix FX_AGG^o > 0 present for both groups in 2017 and only for emerging economies in 2007; 2007 and 2017 dummies positively associated with equity liabilities in domestic currency (column (6)). - Robustness: results robust to alternative specifications and controls. ### Key takeaways - Cross-country distribution shifted significantly toward long foreign-currency positions between 1990 and 2017 (from 60 percent net negative in 1990 to 20 percent in 2017). - Main drivers of the shift: - Improvement in net international investment positions (excluding reserves), (A−L)/(A+L). - Increase in equity liabilities denominated in domestic currency (FX_AGG^o), especially pronounced in emerging economies. - Trade openness (driven by advanced economies) and output volatility are associated with longer positions in foreign currency. - Exchange rate volatility (driven by emerging economies) and EMU membership are associated with shorter aggregate positions in foreign currency. - Among emerging economies, recessions with depreciating currencies tend to yield capital gains on the net external position, providing a hedge against domestic output fluctuations. *Source: 4.2 Long-term dynamics (chapter section) from the provided IMF content unit.* ### 5.1 Descriptive Evidence ### 5.1 Descriptive Evidence ### Distributional shift in FX_AGG (1990–2017) - Rightward shift in the distribution of FX_AGG documented relative to earlier section. - Cross-country prevalence: - 48 percent of countries had negative FX_AGG in 1997. - 14 percent showed negative values in 2007. - 24 percent displayed negative values in 2012. - 20 percent displayed negative values in 2017. - Between 2012 and 2017 the curve moved slightly to the right. - Scale of FX_AGG adjustments: - For 2007–2017 the scale of the FX_AGG adjustment ranges between -0.2 and +0.2. - For 1997–2007 it ranges from -0.2 to +0.6. ### Pre-crisis (1997–2007) adjustments — scatter-plot evidence - Relation between initial foreign-currency exposures and subsequent change in the run-up to the crisis is negative for: - the full sample, - advanced economies, - emerging economies. - All countries with short positions in foreign currency in 1997 exhibited improvements in FX_AGG between 1997 and 2007, with the exception of Sri Lanka. - Emerging market economies generally started with substantially larger short positions and improved markedly. - Countries with long positions in 1997: - Emerging economies moved towards longer positions. - Advanced economies movements were heterogeneous, but all advanced economies with long positions in 1997 still exhibited long positions in 2007. ### Post-crisis (2007–2017) adjustments — scatter-plot evidence - Relation between initial exposure in 2007 and subsequent adjustment to 2017 changed markedly: - Advanced economies: most move towards larger long positions; only Greece ended up slightly negative in 2017. Flat trend line indicates no significant relation between initial level and subsequent change. - Emerging economies: relation between initial exposure and change in following ten years turns slightly positive. - Many emerging economies with positive FX_AGG in 2007 moved further towards longer foreign-currency exposures (notably the Philippines). - Countries with negative FX_AGG in 2007 moved further into shorter positions — includes Turkey, Sri Lanka, Pakistan, and Tunisia. - Egypt and Morocco recorded short foreign-currency positions in 2017 despite starting with long exposures in 2007. - Summary statement: The pre-crisis period was marked by strong adjustments toward longer FX_AGG positions, while the post-crisis period is characterized by persistence in currency exposures. ### Key country-level movements highlighted - Sri Lanka: exception to the general improvement among short-position countries in 1997 (did not improve by 2007). - Philippines: notably large move toward longer foreign-currency exposures after 2007. - Turkey, Sri Lanka, Pakistan, Tunisia: worsened foreign-currency exposures since the crisis (moved further short). - Egypt, Morocco: moved from long in 2007 to short in 2017. - Greece: among advanced economies ended slightly negative in 2017. --- *Italicized source: Content excerpt from "5.1 Descriptive Evidence" (wpiea2019299-print-pdf).* ### References ### References ### Cited works (selection from the source) - Alfaro, L. and F. Kanczuk(2019): “Debt redemption and reserve accumulation,” IMF Economic Review, 67, 261–287. - Arslanalp, S. and T. Tsuda(2014): “Tracking global demand for emerging market sovereign debt,” IMF Working Papers 14/39, International Monetary Fund. - Blanchard, O. and J. Acalin(2016): “What does measured FDI actually measure?” Policy brief 16-17, Peterson institute for International Economics. - Bénétrix, A. S., P. R. Lane, and J. C. Shambaugh(2015): “International currency exposures, valuation effects and the global financial crisis,” Journal of International Economics, 96, 98–109. - Bénétrix, A. S. and M. Schmitz(2019): “Euro-US dollar exposures in cross-border banking,,” The international role of the euro 2019, European Central Bank. - Chinn, M. D. and H. Ito(2006): “What matters for financial development? Capital controls, institutions, and interactions,” Journal of Development Economics, 81, 163 – 192. - Coeurdacier, N. and P. Martin(2009): “The geography of asset trade and the euro: Insiders and outsiders,” Journal of the Japanese and International Economies, 23, 90–113. - Damgaard, J., T. Elkjaer, and N. Johannesen(2019): “The rise of phantom investments,” Finance and Development 12/13, International Monetary Fund. - Di Giovanni, J. and J. C. Shambaugh(2008): “The impact of foreign interest rates on the economy: The role of the exchange rate regime,” Journal of International Economics, 74, 341–361. - Eichengreen, B., R. Hausmann, and U. Panizza(2003): “The Pain of Original Sin,” in Debt denomination and financial instability in emerging-market economies, ed. by B. Eichengreen, R. Hausmann, and U. Panizza, University of Chicago Press Economics Books. - Emter, L., M. Schmitz, and M. Tirpák(2019): “Cross-border banking in the EU since the crisis: What is driving the great retrenchment?” Review of World Economics (Weltwirtschaftliches Archiv), 155, 287–326. - Georgiadis, G. and A. Mehl(2016): “Financial globalisation and monetary policy effectiveness,” Journal of International Economics, 103, 200–212. - Goldberg, L. S. and C. Tille(2008): “Vehicle currency use in international trade,” Journal of International Economics, 76, 177–192. - Gopinath, G. and J. C. Stein(2018): “Banking, trade, and the making of a dominant currency,” NBER Working Papers 24485, National Bureau of Economic Research, Inc. - Haussman, R. and U. Panizza(2011): “Redemption or abstinence? Original sin, currency mismatches and counter cyclical policies in the new millennium,” Journal of Globalization and Development, 2, 1–35. - Ilzetzki, E., C. Reinhart, and K. Rogoff(2019): “Why is the euro punching below its weight?” mimeo. - International Monetary Fund (IMF)(2000): “The dynamics of external adjustment,” External sector report, Washington, DC. - Jeanne, O.(2007): “International reserves in emerging market countries: too much of a good thing?” Brookings Papers on Economic Activity, 38, 1–80. - Lane, P. R.(2015): “A financial perspective on the UK currency account deficit,” Tech. Rep. 234, National Institute Economic Review. - Lane, P. R. and G. M. Milesi-Ferretti(2001): “The external wealth of nations: measures of foreign assets and liabilities for industrial and developing countries,” Journal of International Economics, 55, 263–294. - Lane, P. R. and G. M. Milesi-Ferretti(2003): “International financial integration,” IMF Staff Papers, 50, 82–113. - Lane, P. R. and G. M. Milesi-Ferretti(2007): “The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970-2004,” Journal of International Economics, 73, 223–250. - Lane, P. R. and G. M. Milesi-Ferretti(2008): “International investment patterns,” The Review of Economics and Statistics, 90, 538–549. - Lane, P. R. and G. M. Milesi-Ferretti(2009): “Where did all the borrowing go? A forensic analysis of the U.S. external position,” Journal of the Japanese and International Economies, 23, 177–199. - Lane, P. R. and G. M. Milesi-Ferretti(2014): “Global imbalances and external adjustment after the crisis,” IMF Working Papers 14/151, International Monetary Fund. - Lane, P. R. and G. M. Milesi-Ferretti(2018): “The External Wealth of Nations Revisited: International Financial Integration in the Aftermath of the Global Financial Crisis,” IMF Economic Review, 66, 189–222. - Lane, P. R. and J. C. Shambaugh(2010a): “Financial exchange rates and international currency exposures,” American Economic Review, 100, 518–540. - Lane, P. R. and J. C. Shambaugh(2010b): “The long or short of it: determinants of foreign currency exposure in external balance sheets,” Journal of International Economics, 80, 33 – 44, special Issue: JIE Special Issue on International Macro-Finance. - Maggiori, M., B. Neiman, and J. Schreger(2019a): “International currencies and capital allocation,” Journal of Political Economy, forthcoming. - Maggiori, M., B. Neiman, and J. Schreger(2019b): “The rise of the dollar and fall of the euro as international currencies,” AEA Papers and Proceedings, 109, 521–26. - McCauley, R. N., A. S. Bénétrix, P. M. McGuire, and G. von Peter(2019): “Financial deglobalisation in banking?” Journal of International Money and Finance, 94, 116–131. - McQuade, P. and M. Schmitz(2017): “The great moderation in international capital flows: A global phenomenon?” Journal of International Money and Finance, 73, 188–212. - Milesi-Ferretti, G. and C. Tille(2011): “The great retrenchment: international capital flows during the global financial crisis,” Economic Policy, 26, 285–342. - Obstfeld, M. and K. Rogoff(2001): “The six major puzzles in international macroeconomics: is there a common cause?” in NBER Macroeconomics Annual 2000, Volume 15, National Bureau of Economic Research, Inc, NBER Chapters, 339–412. - Obstfeld, M., J. C. Shambaugh, and A. M. Taylor(2010): “Financial stability, the trilemma, and international reserves,” American Economic Journal: Macroeconomics, 2, 57–94. - Shambaugh, J. C.(2004): “The effect of fixed exchange rates on monetary policy,” The Quarterly Journal of Economics, 119, 301–352. - Tille, C.(2003): “The impact of exchange rate movements on U.S. foreign debt,” Current Issues in Economics and Finance, 9. ### Figures — notes and key descriptions - Figure 1: Financial and Trade weighted exchange rates — comparison for USA, Colombia, Germany. Financial Exchange Rate Index calculated using weights of US dollar, euro, Pound sterling, Japanese yen and domestic currency; base year 1990. - Figure 2: Valuation Effects: Comparison of VALXR for Germany — compares authors' VALXR (2019) with VALXR (Bundesbank). - Figure 3: International Financial Integration: All Countries — shows IFI (USD) and IFI (EUR), with series excluding USA & Euro Area. - Figure 4: International Financial Integration: Advanced Economies — shows IFI (USD) and IFI (EUR), with series excluding USA & Euro Area. - Figure 5: International Financial Integration: Emerging and Developing Economies — shows IFI (USD) and IFI (EUR). - Figure 6: FXAGG long-term dynamics: 1990 vs 2017 — cumulative distribution for net aggregate foreign-currency exposures FXAGG_i,t = ω^F_Ai,t s_Ai,t − ω^F_Li,t s_Li,t; FXAGG ranges between -1 and 1; sample includes 50 countries; for Russia and Czech Republic use FXAGG for 1993. - Figures 7–11: Decompositions and dynamics of FXAGG across periods (1990, 1997, 2007, 2012, 2017) and pre-/post-crisis changes; definitions and decompositions as in main text (e.g., FXAGG = (A−L)/(A+L) + FXAGGo). ### Tables — key statistics and empirical results - Table 1: Correlations of Exchange Rate Indices, 1990-2017 — within-country correlations (means and medians): - All Countries: Mean (A,L) 0.854; Mean (A,T) 0.453; Mean (L,T) 0.398; Mean (I,T) 0.107. Median (A,L) 0.927; Median (A,T) 0.458; Median (L,T) 0.365; Median (I,T) 0.201. - Advanced: Mean (A,L) 0.953; Mean (A,T) 0.479; Mean (L,T) 0.446; Mean (I,T) 0.162. Median (A,L) 0.955; Median (A,T) 0.462; Median (L,T) 0.414; Median (I,T) 0.279. - Emerging and Developing: Mean (A,L) 0.750; Mean (A,T) 0.426; Mean (L,T) 0.346; Mean (I,T) 0.046. Median (A,L) 0.778; Median (A,T) 0.448; Median (L,T) 0.354; Median (I,T) 0.109. - Table 2: Foreign Currency Exposure (FXAGG) and Subcomponents — 2017 cross-country summary (Mean, Median, p25, p75): - All countries: - FXAGG: Mean 0.14; Median 0.18; p25 0.06; p75 0.32. - (A−L)/(A+L): Mean -0.13; Median -0.06; p25 -0.26; p75 0.05. - FXAGGo: Mean 0.27; Median 0.28; p25 0.20; p75 0.36. - (A_NR − L)/(A+L): Mean -0.21; Median -0.13; p25 -0.42; p75 0.00. - FXR/(A+L): Mean 0.08; Median 0.08; p25 0.01; p75 0.13. - (PEQL+FDIL)/(A+L): Mean 0.24; Median 0.24; p25 0.16; p75 0.31. - DEBTL_DC/(A+L): Mean 0.12; Median 0.07; p25 0.04; p75 0.14. - A_DC_NR/(A+L): Mean 0.08; Median 0.02; p25 0.00; p75 0.10. - Advanced: - FXAGG: Mean 0.17; Median 0.11; p25 0.07; p75 0.23. - (A−L)/(A+L): Mean -0.03; Median 0.00; p25 -0.14; p75 0.05. - FXAGGo: Mean 0.20; Median 0.20; p25 0.11; p75 0.27. - (A_NR − L)/(A+L): Mean -0.04; Median 0.00; p25 -0.14; p75 0.04. - FXR/(A+L): Mean 0.02; Median 0.01; p25 0.00; p75 0.02. - (PEQL+FDIL)/(A+L): Mean 0.17; Median 0.18; p25 0.14; p75 0.20. - DEBTL_DC/(A+L): Mean 0.20; Median 0.15; p25 0.08; p75 0.29. - A_DC_NR/(A+L): Mean 0.18; Median 0.14; p25 0.03; p75 0.30. - Emerging and Developing: - FXAGG: Mean 0.12; Median 0.18; p25 -0.01; p75 0.34. - (A−L)/(A+L): Mean -0.20; Median -0.13; p25 -0.40; p75 0.02. - FXAGGo: Mean 0.33; Median 0.34; p25 0.28; p75 0.38. - (A_NR − L)/(A+L): Mean -0.33; Median -0.32; p25 -0.52; p75 -0.12. - FXR/(A+L): Mean 0.13; Median 0.12; p25 0.09; p75 0.16. - (PEQL+FDIL)/(A+L): Mean 0.28; Median 0.28; p25 0.24; p75 0.34. - DEBTL_DC/(A+L): Mean 0.05; Median 0.05; p25 0.02; p75 0.08. - A_DC_NR/(A+L): Mean 0.01; Median 0.00; p25 0.00; p75 0.01. - Table 3: FXAGG decomposition, correlations 1990-2017 — cross-country correlation statistics (mean, median, p25, p75) for FXAGG and subcomponents (selected highlights): - All countries: (A−L)/(A+L) mean 0.67; median 0.83; FXAGGo mean 0.56; median 0.77. - Advanced: (A−L)/(A+L) mean 0.47; median 0.68; FXAGGo mean 0.51; median 0.70. - Emerging and Developing: (A−L)/(A+L) mean 0.81; median 0.90; FXAGGo mean 0.60; median 0.80. - Notes: table reports mean, median, 25th and 75th percentile of the cross-country correlation coefficient between FXAGG and its sub-components; correlations computed for 1990-2017. - Tables 4–6: FXAGG determinants — regression results (robust standard errors in parentheses). Selected coefficient estimates (All countries, Advanced, Emerging): - Table 4 (All countries): Observations 150. R^2 values across specifications: 0.54, 0.57, 0.40, 0.70, 0.61, 0.37, 0.69, 0.90. Selected coefficients (standard errors in parentheses): - Trade: 0.07*** (0.02) in column (1). - vol(GDP): 0.33** (0.15) in column (1). - EMU: -0.20*** (0.03) in column (1). - Size: 0.05*** (0.01) in column (1). - GDPpc: 0.10*** (0.02) in column (1). - 2017: 0.09** (0.04) in column (1). - Table 5 (Advanced): Observations 63. R^2 values across specifications: 0.56, 0.54, 0.36, 0.60, 0.51, 0.14, 0.80, 0.88. Selected coefficients: - Trade: 0.10* (0.05) in column (1). - EMU: -0.15*** (0.04) in column (1). - GDPpc: 0.13** (0.06) in column (1). - Table 6 (Emerging countries): Observations 87. R^2 values across specifications: 0.60, 0.57, 0.41, 0.61, 0.37, 0.37, 0.16, 0.19. Selected coefficients: - vol(I): -0.12*** (0.04) in column (1). - corr(GDP,I): 0.12*** (0.04) in column (1). - GDPpc: 0.14*** (0.03) in column (1). - Table 7: Ten-year changes pre- and post-crisis vs. initial values — FXAGG and subcomponents (coefficients for 1997-2007 and 2007-2017, different samples and specifications). Selected estimates: - Full sample, 1997-2007, Bivariate: FXAGG coefficient -0.53***. - Full sample, 2007-2017, Bivariate: FXAGG coefficient 0.14. - Advanced, 1997-2007, Full controls: FXAGG coefficient -0.76**. - Emerging, 1997-2007, Full controls: FXAGG coefficient -0.41***. - Notes: Robust standard errors in parentheses; ***, **, and * denote p <0.01, p <0.05, and p <0.1. *Italic: Content extracted from "wpiea2019299-print-pdf - References" (figures and tables notes included).* ### Appendix A Data ### Appendix A Data ### A.1 Actual Data — sources and coverage - Actual currency-composition data incorporated from: - IMF survey to authorities (52 countries invited; response rate 85 percent). - Coordinated Portfolio Investment Survey (CPIS), including Table 2 (as of December 2017 a total of 82 countries reported core data; as of December 2017 a total of 59 countries reported to CPIS Table 2). - Euro Area / EU reporting to the ECB (quarterly national data on currency denomination of portfolio debt asset and liability positions available for most EU countries since 2008). - BIS Locational Banking Statistics (LBS) — cross-border banking positions with currency breakdowns including US dollar, euro, Japanese yen, Pound sterling, renminbi, domestic currency, and other currencies. - IMF survey specifics: - Survey requested data from 1990 on IIP components broken down into five SDR currencies (US dollar, euro, Japanese yen, Pound sterling, renminbi), domestic currency, and “other currencies.” - Sample includes 52 countries (list provided in source) that together account for over 90 percent of World’s GDP. - Treatment: requested data labeled public unless authorities objected; any template with confidential data was not used. - Coverage dynamics: around 52 percent of countries reported some data for recent years; only 1 country submitted partial data covering 1990; around 7 countries reported some information at the start of the 2000s; 27 countries in 2017. - CPIS usage: - Used CPIS Table 2 currency-of-denomination data where survey data were not available or were confidential; CPIS provides Table 2 currency detail as an encouraged item. - ECB usage: - ECB Guideline ECB/2011/23 requires transmission of quarterly national data for portfolio debt denomination (euro, US dollar, other currencies); series available for most EU countries since 2008. - BIS LBS usage: - Quarterly cross-border banking positions from late 1970s for up to 36 reporters in later sample years. - LBS provides currency breakdown for the five SDR currencies and domestic currency. - LBS reporting countries in sample (34): Australia, Austria, Belgium, Brazil, Canada, Chile, China, Denmark, Finland, France, Germany, Greece, Hong Kong, India, Indonesia, Ireland, Italy, Japan, Korea, Malaysia, Mexico, Netherlands, Norway, Philippines, Portugal, Russia, Singapore, South Africa, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States. Data for the United States complemented with TIC System data. ### A.2 Combining estimated and actual data — general approach - Strategy: use actual data whenever available; extend coverage backwards and fill gaps using estimation methods (splicing actual data backwards using estimated currency weights for each IIP component) to create series covering 1990-2017 (data for Russia start in 1993; Czech Republic reported since 1993). - Builds on Lane and Shambaugh (2010) and Bénétrix et al. (2015) with improvements: - Access to superior data sources (including actual data). - Refined estimation methods. - New FDI dimension splitting FDI into equity and debt components (FDI debt can represent up to 40 percent of total FDI in some countries). - Financial derivatives excluded due to largely unavailable currency breakdowns. ### A.2.1 Foreign assets — item-specific estimation methods and data use - Asset categories: portfolio equity, foreign direct investment (equity and debt), portfolio debt, other investment (largely bank-related), and reserves. - Portfolio equity: - Use actual data from IMF Survey and CPIS Table 2 where available. - For missing countries, rely on geography of cross-border positions from CPIS (2001 onward) and a gravity-based regression for missing country pairs. - Regression model: log(1 + EQ_ijt) = ψ_j + θ_t + β Z_ijt + γ X_it + ε_ijt, where EQ_ijt are portfolio equity assets from source country i to host j in period t; ψ_j host fixed effects; θ_t year fixed effects; Z_ijt bilateral variables (trade, distance, colonial dummies, regional dummies, relative GDP measures); X_it source-country characteristics (latitude, landlocked status, population, capital controls, GDP per capita, gross cross-border financial stocks). - Regression explanatory power: R2 values in the region of 0.75. - For country group with no currency-mix information (China, Guatemala, Morocco, Peru, Sri Lanka, Tunisia), gravity weights used for 1990-2017. - For years before 1998, follow Lane and Shambaugh (2010) using CPIS trial survey 1997 weights where available; otherwise synthetic gravity-model weights for 1990-1997; CPIS 2001 weights applied for 1998-2001. - Foreign Direct Investment (FDI): - Split into equity and debt components; equity share from IMF’s International Financial Statistics (IFS). - Equity component: - Use IMF Survey actual currency weights where available. - Assume equity component denominated in host country currency. - 1990-2008: update Bénétrix et al. (2015) using UNCTAD bilateral stock/flow data where applicable. - 2009-2017: use IMF Coordinated Direct Investment Survey (CDIS) outward/inward stocks (coverage for 2017: 108 and [number missing in source], reporting countries). - For Argentina, China, Egypt, Indonesia, Singapore, Tunisia, Uruguay (limited/no CDIS coverage), apply gravity approach; regression results reported in Table A.5. - Chain-link CDIS-based bilateral positions with UNCTAD-based positions pre-2009 to form estimated series. - Debt component: - Use IMF Survey actual data where available; debt and equity components show very different currency distributions. - Because FDI debt can represent up to 40 percent of total FDI in some countries, estimate separately; when lacking, extend coverage using portfolio debt assets as proxy. - Portfolio debt: - Combine geography of portfolio debt asset positions with currency of denomination of host countries’ international bond issuance (BIS International Debt Statistics). - Apply host-country issuance currency distribution to source-country bilateral positions, netting out holdings by US residents of US-dollar debt, Euro Area holdings of euro, and Japanese residents’ holdings of yen — netting done using US Treasury Report on US Portfolio Holdings of Foreign Securities, ECB, and Bank of Japan data. - Example: host country issuance split 50% US dollar / 50% euro; if 80% of US dollar debt held by US and 50% of euro issuance held by Euro Area, then residual attribution to another source country position = 29 percent US dollar and 71 percent euro. - Synthetic data used to expand time-series coverage where actual data absent. - Other investment (bank-related): - Survey data available for 13 countries for some years. - Complement with BIS LBS reporter country aggregate and bilateral cross-border bank positions (currency composition for five SDR currencies and domestic currency). - LBS coverage: 34 reporting countries in sample 1990-2017 (list above). - Mirror-data approach: derive currency composition of foreign assets of non-reporting countries by using aggregate currency composition of foreign liabilities of reporting banking systems. - Justification: aggregate cross-border liabilities of reporter countries account for around 90% of reporter countries’ external assets; mirror LBS weights align with reported LBS weights (US dollar weights correlation about 90% between reported and mirror series). - Assume domestic currency share in assets of a non-reporter corresponds to “other” currencies share in aggregate reporters’ liabilities vis-à-vis that non-reporter. - Pre-1999 euro series in LBS are sums of legacy currencies; derive legacy-currency shares using reported “domestic currency” and BIS positions in individual legacy currencies; for non-euro-area countries assume legacy-currency distribution follows euro-area aggregate pattern. - Final hierarchy for other investment weights: priority given to survey data, then BIS LBS reported data, then BIS LBS derived (mirror) data. - Reserves: - Baseline data 1990-2012 from Bénétrix et al. (2015), which in turn builds on Eichengreen and Mathieson (2000) model predicting reserve currency shares using trade shares, debt shares, and exchange rate regime. - Extended dataset 2010-2017 constructed by obtaining or estimating reserve weights from non-confidential Central Bank / Ministry publications for a set of countries (Canada, Chile, Colombia, Peru, Poland, Sweden, Switzerland, United Kingdom, USA, Uruguay) and COFER-derived inferences for others. - COFER constraints: - COFER disseminates aggregated end-of-period data for total, advanced economies, and emerging/developing economies; individual country COFER details are confidential except optional SDR breakdown publicly available since 2016 for nine countries (Australia, Belgium, Brazil, Finland, Germany, Ireland, Netherlands (only 2017), Norway, Portugal); used these where available and filled gaps for 2010-2015 using a two-year moving average subject to SDR basket totals. - For countries with only aggregate SDR / non-SDR weights (Argentina, Austria, France, Italy, Japan, Korea, Czech Republic, China, Egypt, Greece, Hungary, India (only for 2010), Indonesia, Israel, Malaysia, Mexico, Morocco, New Zealand, Philippines, Russia, Singapore, South Africa, Thailand, Turkey), infer currency composition by applying group (advanced/emerging) currency weights re-weighted by country-specific SDR/non-SDR composition. - For Euro Area countries, use combined weight of euro and US dollar of “advanced economies” group as proxy for US dollar weight (euro weight set to zero). - For Czech Republic, Russia, Norway refine estimates using ECB “The International Role of the Euro” publication share-of-euro in total reserves for 2010-2017. - Merge Bénétrix et al. (2015) series (1990-2012) with constructed 2010-2017 series by linearly interpolating starting from 5 years preceding availability of second dataset (typically from 2005; for second-dataset starting in 2008 interpolation starts in 2003). ### A.2.2 Foreign liabilities — item-specific estimation methods and data use - Liability categories: portfolio equity, foreign direct investment (equity and debt), portfolio debt, other investment. - Portfolio equity liabilities: - Assumed denominated in currency of host country; thus total equity liabilities assumed in domestic currency and generate no foreign-currency exposure. - Foreign Direct Investment liabilities: - Split into equity and debt; equity share from IMF IFS. - Equity liabilities assumed denominated in host-country currency. - Debt liabilities treated using currency composition of portfolio debt liabilities when actual data unavailable. - Portfolio debt liabilities: - Use survey and ECB actual data where available. - Baseline synthetic sources: BIS International Debt Statistics and Arslanalp and Tsuda (2014) for a sample of 19 emerging markets. - BIS IDS covers debt securities issued by non-residents with comprehensive currency breakdown, but tends to underestimate domestic-currency share for emerging markets because domestically issued debt currency composition and foreign-held proportions are not fully captured. - For 19 emerging countries (Argentina, Brazil, Chile, China, Colombia, Egypt, Hungary, India, Indonesia, Malaysia, Mexico, Peru, Philippines, Poland, Russia, South Africa, Thailand, Turkey, Uruguay) use Arslanalp and Tsuda (2014) share of central government securities denominated in domestic currency held by foreigners as proxy for domestic-currency share of portfolio debt liabilities; data available 2004-2017 and spliced backwards using BIS international issuance data. - Assembled actual data spliced backwards with synthetic series; Table A.4 summarizes sources and years. - Other investment liabilities: - Built analogously to other investment assets: survey information complemented with country-level BIS LBS data. - Survey data on other investment liabilities available for 14 countries for specific years. - LBS coverage of liabilities aligned with assets coverage. ### A.3 Uses of the dataset (brief) - Dataset provides currency composition of IIP items for 1990-2017 (with country-specific start-years noted: Russia from 1993; Czech Republic from 1993). - Enables analysis that distinguishes FDI equity from FDI debt and assesses currency exposures across IIP items using a mix of actual and estimated weights. ### A.4 Comparison of actual and estimated data (brief) - When survey and BIS LBS reported data overlap for other investment, weights show very high correlation (around 90 percent). - Mirror LBS weights closely align with reported LBS weights (US dollar weights correlation about 90 percent). - Regression approaches for geography-based imputations show strong explanatory power (R2 values in the region of 0.75); regression estimates and data-source details reported in Table A.5 (regression outputs available upon request). *Source: Appendix A, wpiea2019299-print-pdf — Data* ### Section 2.1.4). Hence, we are able to rely on reporter countries’ data, complemented with ### wpiea2019299-print-pdf - Section 2.1.4) ### Method for non-reporting countries and euro legacy currencies - Currency composition of foreign liabilities of non-reporting countries is derived using aggregate currency composition of the foreign banking assets of the reporting countries (“mirror data”). - Assumption: domestic currency share in liabilities of a non-reporter corresponds to the share of “other” currencies in aggregate reporters’ assets vis-à-vis this non-reporter. - For euro legacy currencies: - Use series reported for ‘domestic currency’ and BIS information on positions in other individual legacy currencies to derive shares of individual legacy currencies in foreign assets of euro area Member States before 1999. - For non-euro area countries, assume distribution into legacy currencies follows the aggregate distribution derived from assets of euro area countries. - Final other investment liability weights are obtained by complementing survey data using the same hierarchy of sources as for assets. ### Dataset composition and uses (A.3) - Dataset contains: - asset, debt asset, liability, and debt liability weights. - decomposition of each variable into domestic and foreign currency components. - breakdown into the five SDR currencies: US dollar, euro, Japanese yen, Pound sterling and renminbi, and an aggregate “other” which includes all currencies other than the five SDRs. - Definitions: - Debt assets include portfolio debt, FDI debt, other investment and reserve assets. - Debt liabilities include portfolio debt, FDI debt and other investment. - Reporting: - The sum of domestic currency and foreign currency weights equals 1. - The sum of US dollar, euro, Japanese yen, Pound sterling, renminbi and “other” weights equals 1. - The dataset reports share of each variable broken down by currency and in proportion of GDP, and includes main indicators used in the paper. ### Comparison Between Actual and Estimated Currency Weights (A.4) — overview - Estimated currency weights are compared to actual data only for years with available actual data; estimated data are used for comparison purposes but are not included in the final dataset when actual data exist. - Figures A.2-A.9 display average cross-country differences between actual and estimated data by year for different IIP items. - Currency categories used in comparisons: US dollar, euro, domestic currency, and all other foreign currencies (bundling Japanese yen, Pound sterling, renminbi and other currencies). - Note: only years with more than 2 observations (or number of countries) are plotted. ### Asset-side findings (enumerated) 1. When number of observations is small, differences between actual and estimated data tend to be larger and confidence interval bands widen. 2. For many asset items, notably portfolio equity and FDI equity, US dollar weights are underestimated by estimates: - Synthetic data rely on geography of cross-border positions; dominance of US dollar implies some cross-border positions outside the US are issued in US dollars and not captured by geography. 3. Portfolio equity assets: - Differences between euro weights in actual and estimated data are very small. - US dollar weights tend to be underestimated. - Estimated approach assumes equity is denominated in host country currency, implying no domestic-currency exposure from equity assets. - Actual data reveal domestic currency exposure is negligible with very few non-zero observations. - Discrepancies are concentrated where gravity-type models fill data gaps and are more marked in emerging economies. - Correlations between actual and estimated portfolio equity assets: - US dollars: 0.90 for advanced economies, 0.5 for emerging economies. - Euros: 0.82 for advanced economies, 0.5 for emerging economies. 4. FDI equity assets: - US dollar weights are underestimated; differences for euro weights are small. - Estimation assumes equity denominated in host country currency (no domestic-currency exposure); actual domestic-currency exposure is tiny. - Discrepancies arise for countries where gravity model is used, financial centers, and countries with SPEs (special purpose entities). 5. FDI debt assets: - Differences tend to be larger than for other items. - Euro currency weights are overestimated while US dollar weights are underestimated, especially in the latest part of the sample. - Item is largely concentrated on dominant currencies, posing estimation challenges. 6. Portfolio debt assets: - Euro weights are overestimated; US dollar weights are underestimated. - Differences in dollar weights are larger at the beginning of the sample. - Gap between actual and estimated data is, for the most part, consistent across countries. 7. Other investment assets: - US dollar weights are underestimated. - Confidence bands are relatively wide in the earlier sample period. - Euro weights are overestimated in the first years of the sample (few observations). - Estimation largely based on banking; other investment is broader, and lack of data on non-banking components may explain differences. - Correlations between actual and estimated data: US dollar 0.94, euro 0.91. ### Liability-side findings (enumerated) 1. FDI debt liabilities: - Estimates are, on average, close to actual data, especially after 2004. - In earlier years US dollar weights are overestimated while domestic currency weights are underestimated. 2. Portfolio debt liabilities: - US dollar weights are overestimated since 2009, with the difference growing larger since 2010. - Euro weights are also overestimated but differences are relatively small. - Domestic currency weights are underestimated, particularly in emerging countries. - Main source for estimation is international issuance statistics from the BIS LBS; lack of information on domestically issued debt securities held by non-residents (and their currency composition) may drive discrepancies. 3. Other investment liabilities: - In the early 2000s, euro weights are overestimated while US dollar weights are underestimated. - Estimation largely based on banking; other investment is broader and lack of data on remaining components could explain differences. - Correlations between actual and estimated data: US dollar 0.87, euro 0.88. *Source: wpiea2019299-print-pdf - Section 2.1.4).* ### 4. We do not show the comparison for FDI equity liabilities and portfolio equity liabilities ### 4. We do not show the comparison for FDI equity liabilities and portfolio equity liabilities ### Key assumption and high-level finding - FDI equity liabilities and portfolio equity liabilities are assumed to be in domestic currency. - A comparison with actual data reveals that this assumption "is the case." - The analysis provides perspective on how close actual and estimated data are: although there is some heterogeneity, "the main trends in the data seem to be captured by the estimates." - The exercise highlights the importance of incorporating as much actual data as possible to build the dataset. ### Data reporting and survey templates - Figure A.1: Shows the share of countries reporting IIP currency composition statistics at any level of disaggregation for each year (authors’ own elaboration based on survey responses). - Table A.1: Survey template for Total Assets (years listed 1990–2017). Currency columns include: - US dollar - Euro - Japanese yen - Pound sterling - Renminbi - National currency (if different from listed above) 1/ - Other currencies 2/ - Instrument breakdowns: Direct Investment (Equity and investment fund shares, Debt instruments), Portfolio Investment (Equity and investment fund shares, Debt instruments), Financial Derivatives, Other Investment. - Table A.2: Survey template for Total Liabilities (years listed 1990–2017). Same currency and instrument structure as Table A.1. - Notes for templates: - 1/ denotes national currency exposures when different from US dollar, euro, Japanese yen, Pound sterling and renminbi. - 2/ excludes US dollar, euro, Japanese yen, Pound sterling, renminbi and own currency. ### Coverage of actual vs synthetic data (Tables A.3 and A.4) - Table A.3: Assets — Actual and Synthetic Data - Lists per-country coverage periods for: - FDI Equity Assets (Actual Data / Synthetic data) - FDI Debt Assets (Actual Data / Synthetic data) - Portfolio Equity Assets (Actual Data / Synthetic data) - Portfolio Debt Assets (Actual Data / Synthetic data) - Other Investment Assets (Actual Data / Synthetic data) - Examples preserving exact entries (selected): - Argentina: FDI Equity Assets Actual Data 1990-2017 | Synthetic data 1990-2017; FDI Debt Assets Actual Data 2016-2017**, Synthetic data 1990-2015; Portfolio Equity Assets Actual Data 2016-2017**, Synthetic data 1990-2015; Other Investment Assets Actual Data 1990-2017, Synthetic data 1990-2017***. - Belgium: FDI Equity Assets Actual Data 2013-2017, Synthetic data 1990-2012; FDI Debt Assets Actual Data 2013-2017, Synthetic data 1990-2012; Portfolio Equity Assets Actual Data 2013-2017, Synthetic data 1990-2012; Portfolio Debt Assets Actual Data 2008-2017, Synthetic data 1990-2007; Other Investment Assets Actual Data 2013-2017, Synthetic data 1990-2012***. - United States: FDI Equity Assets Actual Data 1990-2017, Synthetic data 1990-2017; FDI Debt Assets Actual Data 1990-2017, Synthetic data 1990-2017; Portfolio Equity Assets Actual Data 1990-2017, Synthetic data 1990-2017; Portfolio Debt Assets Actual Data 2003-2017**, Synthetic data 1990-2002; Other Investment Assets Actual Data 1990-2017, Synthetic data 1990-2017***. - Notes: - "Actual data are from IMF survey unless otherwise indicated." - * denotes actual data from the ECB. - ** identifies actual data from CPIS T2. - *** indicates that the country is a BIS reporter. - Table A.4: Liabilities — Actual and Synthetic Data - Lists per-country coverage periods for: - FDI Debt Liabilities (Actual Data / Synthetic data) - Portfolio Debt Liabilities (Actual Data / Synthetic data) - Other Investment Liabilities (Actual Data / Synthetic data) - Selected exact entries: - Austria: FDI Debt Liabilities Actual Data 1990-2017, Synthetic data 2006-2017*; Portfolio Debt Liabilities Actual Data 1990-2005, Synthetic data 1990-2017; Other Investment Liabilities Actual Data 1990-2017, Synthetic data 1990-2017***. - Japan: FDI Debt Liabilities Actual Data 1990-2017, Synthetic data 2014-2017; Portfolio Debt Liabilities Actual Data 1990-2013, Synthetic data 1990-2017; Other Investment Liabilities Actual Data 1990-2017, Synthetic data 1990-2017***. - Turkey: FDI Debt Liabilities Actual Data 2014-2017, Synthetic data 1990-2013; Portfolio Debt Liabilities Actual Data 2014-2017, Synthetic data 1990-2013; Other Investment Liabilities Actual Data 2014-2017, Synthetic data 1990-2013***. - Notes: - "Actual data are from IMF survey unless otherwise indicated." - * denotes actual data from the ECB. - ** identifies actual data from CPIS T2 and - *** indicates that the country is a BIS reporter. - Important methodological note: "Portfolio equity liabilities and FDI equity liabilities are assumed to be denominated in the currency of the host country and were therefore excluded from the table since all data are synthetic." ### Gravity models used to fill gaps (Table A.5) - Purpose: Panel data models used to fill gaps in the geographical distribution of portfolio equity assets (PEQA), portfolio debt assets (PDA) and foreign direct investment (FDI). - Sample: Covers all available years and countries from CPIS and CDIS data sets. - Model structure: Each model incorporates three dimensions: country source, country host and time. All models include host country fixed effects and year fixed effects. Models also include source and host country dummy variables for EMU and CEEC memberships and additional gravity variables. - Reported coefficient estimates and robust standard errors (in parenthesis) — preserve exact numeric values: - Dependent variables in columns: (1) PEQA, (2) PDA, (3) FDI - Trade_ij: - (1) 0.121*** (0.00) - (2) 0.102*** (0.00) - (3) 0.516*** (0.01) - Distance_ij: - (1) -0.492*** (0.01) - (2) -0.490*** (0.01) - (3) -0.550*** (0.02) - Both EMU: - (1) 0.178*** (0.04) - (2) 1.437*** (0.04) - (3) -0.045 (0.06) - Both CEEC: - (1) 0.753*** (0.05) - (2) 0.872*** (0.05) - (3) 0.221*** (0.08) - Both Same continent: - (1) -0.261*** (0.02) - (2) 0.204*** (0.02) - (3) -0.274*** (0.03) - Both South Asia: - (1) 0.034 (0.10) - (2) 0.273** (0.11) - (3) 0.553*** (0.09) - Both East Asia: - (1) -0.342*** (0.03) - (2) -0.192*** (0.03) - (3) 0.468*** (0.06) - Both Sub-Saharan Africa: - (1) 0.512*** (0.07) - (2) 0.689*** (0.09) - (3) -0.015 (0.04) - Both Latin America: - (1) -0.280*** (0.02) - (2) -0.465*** (0.02) - (3) -0.022 (0.05) - Both Middle East North Africa: - (1) 0.177** (0.07) - (2) 0.171** (0.08) - (3) -0.126 (0.38) - Both high income: - (1) 1.761*** (0.01) - (2) 1.636*** (0.01) - (3) 1.041*** (0.02) - Both middle income: - (1) -0.416*** (0.01) - (2) -0.577*** (0.01) - (3) -0.520*** (0.02) - Both least developed: - (1) 0.794 (0.72) - (2) 0.103* (0.06) - (3) 0.273*** (0.05) - Both low income: - (1) 0.293*** (0.03) - (2) 0.445*** (0.03) - (3) 0.194*** (0.03) - Same language: - (1) 0.488*** (0.02) - (2) 0.224*** (0.02) - (3) 0.522*** (0.04) - Same colonizer: - (1) 0.268*** (0.02) - (2) 0.128*** (0.02) - (3) 0.340*** (0.03) - Constant: - (1) 1.771** (0.79) - (2) -3.563*** (0.28) - (3) 2.280*** (0.69) - Model fit and sample sizes: - R-squared: - (1) 0.747 - (2) 0.764 - (3) 0.727 - Observations: - (1) 9698 - (2) 8988 - (3) 8228 - (4) 44076 [note: table shows a sequence "969889822844076" corresponding to column-aligned Observations; preserved as separate observation counts above] - Country pairs: - 7222 - 722 - 263 - 378 [preserved from table row "Country pairs722272226378"] - Periods: - (1) 1997-2017 - (2) 1997-2017 - (3) 2009-2017 - Additional model inputs and data sources (verbatim): - Additional gravity variables include region (South Asia, East Asia, Sub-Saharan Africa, Latin America, Middle East and North Africa) and income (low, less developed middle and high income), available in Andrew Rose’s webpage (http://faculty.haas.berkeley.edu/arose/RecRes.htm). - Area, landlocked, latitude and longitude included for source and host countries’ are from the Centre d’Etudes Prospectives et d’Informations Internationales (CEPII). - The regressions include a measure of capital account openness sourced from Chinn and Ito (2006) and the overall portfolio investment from the External Wealth of Nations dataset. - GDP per capita is included to capture different degrees of development, and is sourced from the World Bank World Development Indicators database. ### Accuracy assessment: Average differences between actual and estimated data (Figures A.2–A.9) - For each instrument/category the report provides figures of "Average Difference between Actual and Estimated Data" by currency grouping: - Currency groupings shown in the figures: EUR, USD, DC (domestic currency), FC* (broad category used for the Japanese yen, the Pound sterling, the renminbi and other currencies). - Each figure plots: - Cross-country average difference between actual and estimated data (in blue). - The 95 percent confidence bands (in teal). - The number of observations which represents the number of countries (red dot, right axis). - Note: "Only those years with more than 2 observations are plotted." - Figures included (instrument-specific): - Figure A.2: Portfolio Equity Assets — time span plotted 2001–2017 (selected years shown in axis labels). - Figure A.3: FDI Equity Assets — time span plotted 2004–2016. - Figure A.4: FDI Debt Assets — time span plotted 2004–2016. - Figure A.5: Portfolio Debt Assets — time span plotted 2001–2017. - Figure A.6: Other Investment Assets — time span plotted 1999–2017. - Figure A.7: FDI Debt Liabilities — time span plotted 2000–2016. - Figure A.8: Portfolio Debt Liabilities — time span plotted 2000–2016. - Figure A.9: Other Investment Liabilities — time span plotted 2000–2016. - General finding from figures: cross-country average differences are plotted with confidence bands and observation counts to assess how closely estimated series match actual series across currencies and instruments; authors note heterogeneity but overall main trends are captured. *Source: IMF working paper content (figures, tables and notes reproduced verbatim from the supplied content).* --- _Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019299-print-pdf.pdf_