## wpiea2026169-source-pdf ## Source details **Canonical URL:** [wpiea2026169-source-pdf](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026169-source-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026169-source-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026169-source-pdf.pdf.json) --- ### 3.1 Rapid increase in USD exposures driven by non-banks - Trends in USD exposures - USD-denominated assets amounted to 42 trillion in 2024. - USD-denominated assets nearly doubled from 19 trillion in 2013. - Foreign holdings of USD bonds and equities increased sixfold in the last two decades (from $5.5 trillion in 2002) (Du et al. (2024) cited). - As a share of GDP, the average for countries in the sample rose from 132% in 2013 to 195% in 2024. - USD-denominated liabilities are much smaller relative to assets—about forty percent less as of 2024. - Absolute USD liabilities increased over the past decade but less than assets; foreign holdings of USD-denominated liabilities relative to GDP remained roughly unchanged, peaking in 2015 and declining steadily thereafter. - Investor composition: banks vs non-banks - Growth in USD-denominated assets is primarily driven by rising non-banks’ holdings; USD assets held by the banking sector have remained stable. - For USD-denominated liabilities, both banks and non-banks show stable patterns. - Increase in non-banks’ USD assets aligns with Du et al. (2024), which documents greater allocation to USD securities across insurance, pensions and mutual funds. - Drivers of non-bank increases: higher allocations to non-domestic investments and active rebalancing within non-domestic investments toward USD-denominated assets. - Cross-country heterogeneity (advanced economies vs emerging markets) - Emerging market economies (EMs) exhibit more balanced USD assets and USD liabilities; these have remained aligned over the past decade. - In advanced economies (AEs), USD assets noticeably exceed USD liabilities; the gap widened due to a rapid increase in USD assets. - The rise in global USD exposure has been primarily driven by advanced economies. - Non-banks’ dominance in AEs: - In 2013, banks and non-banks each held roughly half of USD assets. - By 2024, non-banks accounted for over 65 percent of USD assets in AEs, leaving banks with about 35 percent. - In EMs, banks and non-banks hold similar amounts of USD assets. - USD liabilities composition: non-banks constitute a larger share of liabilities than banks in both AEs and EMs, especially in EMs; liability composition has remained relatively stable. - Net dollar positions: - In AEs, both banks and non-banks are net holders of dollar assets (more dollar assets than liabilities). - In EMs, only banks are net dollar asset holders; non-banks have more dollar liabilities than assets. - Marginal investors of USD asset growth (regression evidence) - Regression setup (following Fang et al. (2025)): changes in USD assets by investor type scaled by lagged total USD assets regressed on changes in total USD assets (coefficients for banks and non-banks sum to one); country fixed effects included; Driscoll-Kraay standard errors used. - Key estimates (Table (1) summary): - Advanced economies: non-banks drive 83% of the increase in USD assets; banks drive 18%. - Emerging markets: non-banks account for 27% of the increase; banks account for 73%. - Conclusion: non-bank investors are the primary marginal investors in USD assets in advanced economies; banks are the primary marginal investors in EMs. - USD exposures scaled by FX market depth - USD exposures scaled using BIS Triennial Survey daily FX turnover figures converted to monthly; total USD assets divided by corresponding monthly FX turnover volumes. - Scaling by FX turnover measures exposure relative to FX market depth; deeper FX markets can accommodate hedging demands more easily. - The level of USD assets scaled by GDP does not perfectly correlate with USD assets scaled by FX turnover. - Country examples: - Luxembourg: relatively high USD assets to GDP, but comparatively lower USD assets scaled by FX turnover due to the depth of the Euro FX market. - Taiwan: substantial USD assets relative to the size of its FX market (high when scaled by FX turnover) but smaller ratio when scaled by GDP, reflecting a relatively shallow FX market for the Taiwanese dollar. - Note: BIS Triennial Survey numbers for year t are applied to years t−1 and t+1 as approximations. - CIP deviations: definition and measurement - One-period CIP condition (cash vs FX swap markets) and introduction of CIP deviation x_{i/$t} as a wedge when equality does not hold. - In log terms, CIP deviation: - x_{i/$t} = y_{$t} − y_{i t} + ρ_{i/$t}, where ρ_{i/$t} is the forward premium (log difference between forward and spot). - Sign interpretation: - Negative CIP deviation: USD borrowing costs more in the FX swap market (forward rate too low); investors selling dollar forwards incur a cost. - Positive CIP deviation: USD borrowing more expensive in the cash market; investors buying dollar forwards incur a cost. - Focus on 3-month CIP deviations: - Chosen to mitigate credit risk contamination and because short-term forwards and interbank lending markets are more liquid. - Correlation between 3-month CIP deviations and sovereign CDS spreads in the sample is near zero, supporting that the short-term measure does not reflect sovereign default risk. - Summary statistics (Table (2)): CIP deviations are mostly negative in the sample; the median for both AEs and EMs is below zero. - Empirical relationships between USD exposures and CIP deviations - Baseline cross-sectional regression: CIP_it = γ_t + β DollarExp_it + ν_it, where DollarExp_it is net dollar exposure (dollar assets minus liabilities) divided by monthly FX turnover; time fixed effects included; Driscoll-Kraay standard errors used. - Main findings (Table (5)): - Advanced economies: relationship between net dollar exposures and CIP deviations is statistically and significantly negative. - Emerging markets: relationship is statistically and significantly positive. - Interpretation for AEs: higher net dollar exposures increase hedging demand, pushing up dollar borrowing cost in the FX swap market and making CIP deviations more negative. - For EMs: positive coefficient partly aligns with Dao and Gourinchas (2025) who find a weak positive correlation due to measurement challenges; robustness to inclusion of sovereign CDS spreads is confirmed in Appendix. - Banks vs non-banks analysis (regression CIP_it = γ_t + β1 DollarExp_bk,it + β2 DollarExp_nb,it + ν_it): - Advanced economies (Table (6)): coefficients on net dollar exposures for both banks and non-banks are significantly negative. - Emerging markets (Table (6)): coefficient on banks’ net dollar exposure is statistically insignificant; coefficient on non-banks’ net dollar exposure is statistically positive. - Robustness: findings remain robust to inclusion of sovereign credit risk as a control. - Interpretation for EMs: non-banks typically have more dollar liabilities than assets; under the hedging demand channel, hedging of open dollar liability positions by non-banks would supply dollars into the FX swap market and lower dollar borrowing costs (implying a negative relationship). The empirically observed significantly positive coefficient for non-banks in EMs suggests an alternative mechanism—linked to substantial dollar-denominated debt held by non-bank sectors in EMs—is driving CIP deviations in these countries. ### 4.3 Divergent relations between CIP deviations and dollar exposures - Overview and objective - Explain why CIP deviations can be positively related to non-banks’ net dollar exposure in emerging markets by linking CIP deviations to firms’ currency choice for debt issuance and asset holdings. - Approach: illustrative example of dollar debt issuance and a formal model (inspired by Liao (2020)) that endogenizes CIP deviations via firms’ currency choices and constrained-capital traders who arbitrage CIP deviations. - Focus: rationalize cross-sectional variation in CIP deviations across countries driven by the stock of net dollar exposures. - CIP deviations and dollar debt issuance (illustrative example) - Local-currency borrowing cost: r_i^t (equation (7) in source). - USD borrowing cost (hedged via forward/swap): (1/S_{i/$}^t)(1 + r_ $^t)F_{i/$}^t − 1 (equation (8) in source). - Borrowing-cost differential Φ defined (equation (9) in source) and log-linearized: - Φ = r_i^t − r_ $^t − f_t + s_t (equation (10) in source). - Decomposition with risk-free yields (equation (11) in source): - Φ = (r_i^t − y_i^t) − (r_ $^t − y_ $^t) − (y_ $^t − y_i^t + ρ_{i/$}^t) - Interpretation: A more negative CIP deviation increases Φ, making USD borrowing cheaper; firms issuing USD liabilities benefit from negative CIP deviations by effectively providing dollar funding where there is a dollar shortage. - Mechanism implication: More negative CIP deviations incentivize firms to increase dollar debt issuance → higher USD liabilities → smaller net USD exposures (potential positive cross-sectional relation when driven by supply factors). - Model: firm choices, trader arbitrage, and equilibrium CIP deviations - Firm liabilities: representative firm issues debt D, chooses USD debt share μ_d given hedging ratio h_d, maximizes CRRA expected utility (equations (12)-(13) in source). - End-of-period wealth W expression (equation (13) in source) distinguishes components: locally funded debt, hedged USD portion (cost r_ $^d + x), and unhedged USD portion (cost r_ $^d + φ) where φ has mean φ̄ and variance V. - Optimal USD debt share (first-order/Taylor expansion) (equation (14) in source): - μ*_d = [r_i^d − (r_ $^d + h_d x + (1−h_d) φ̄)] / [ρ_d (1−h_d)^2 V] - Comparative statics: larger ρ_d or larger V → smaller μ*_d. Higher x raises USD hedging cost → reduces μ*_d; negative x lowers USD funding cost → increases μ*_d. - Firm assets: dollar asset holder funded by LC debt chooses USD asset share μ_a with analogous setup (equations (15)-(17) in source). - Optimal USD asset share (equation (17) in source): - μ*_a = [r_ $^a + h_a x + (1−h_a) φ̄ − r_i^a] / [ρ_a (1−h_a)^2 V] - Negative x (x < 0) implies a hedging cost that reduces dollar asset returns, lowering μ*_a ceteris paribus. - Trader supply: constrained-capital forward/swap trader allocates capital between exploiting CIP deviations x and alternative investment I, with haircut H = γ s (equations (18)-(20) in source). - Trader objective (equation (19) in source) leads to optimal forward position s (equation (21) in source) and, under assumption W = φ_0 / φ_1 (from Liao (2020)), simplified relations: - s = − x / (γ^2 φ_1) (equation (23) in source) - Equivalently x = − γ^2 φ_1 s (equation (24) in source) - Market clearing for forwards (equation (25) in source): - μ_d D h_d − μ_a A h_a + s = 0 - Substituting optimal μ*_d, μ*_a, and s yields equilibrium condition (equation (26) in source): - D h_d [r_i^d − (r_ $^d + h_d x + (1−h_d) φ̄)] / [ρ_d (1−h_d)^2 V] − A h_a [r_ $^a + h_a x + (1−h_a) φ̄ − r_i^a] / [ρ_a (1−h_a)^2 V] − x / (γ^2 φ_1) = 0 - Interpretation: First two terms = net hedging demand from dollar debt and dollar asset holders (functions of x); last term = limited-capital traders’ supply of forwards. Equilibrium x determined by interaction of demand and supply. - Equilibrium intuition and cross-sectional implications - Demand-driven variation: - More dollar assets (higher A) raise hedging demand for selling dollar forwards → CIP deviation must become more negative to clear market → ∂x*/∂A < 0. - More dollar liabilities (higher D) increase demand for buying dollar forwards → x must be less negative / more positive → ∂x*/∂D > 0. - Net implication: Demand-side driven variation produces a negative cross-sectional correlation between CIP deviations and net dollar exposures. - Supply-driven variation: - If supply of arbitrage (traders) is constrained (higher haircut γ → traders supply fewer swaps), then ∂x*/∂γ < 0 (higher γ pushes x more negative). - More negative x reduces μ*_a (fewer dollar assets) and increases μ*_d (more dollar debt) → country’s net dollar exposure falls. - Net implication: Supply-side driven variation can produce a positive cross-sectional correlation between CIP deviations and net dollar exposures. - Institutional interpretation: - Advanced economies: CIP deviations largely shaped by hedging demand (shared global bank suppliers), producing a negative correlation between net dollar exposures and CIP deviations. - Emerging markets: cross-border arbitrage constrained by capital controls, taxation, legal/regulatory differences, and higher haircuts; supply-side differences dominate, producing a positive correlation. - Empirical findings (selected quantitative results) - Cross-sectional relationships (Table 3; regression of CIP deviations on DollarExp): - All countries (column (1)): coefficient = −0.15 (standard error 0.245) - Advanced economies (column (2)): coefficient = −1.06*** (standard error 0.243) - Emerging markets (column (3)): coefficient = 3.43*** (standard error 0.639) - Observations: 210 (all), 139 (advanced), 71 (emerging). Time Fixed Effects: Yes. - Banks vs non-banks (Table 4; regressions of CIP deviations on bank and non-bank DollarExp): - DollarExp_bk (banks): - All: −0.78* (0.366) - Advanced: −1.36** (0.514) - Emerging: 1.14 (0.942) - DollarExp_nb (non-banks): - All: 0.32 (0.512) - Advanced: −0.76*** (0.078) - Emerging: 6.10*** (0.961) - Observations: 210, 139, 71. Time Fixed Effects: Yes. - Robustness with CDS controls (Tables 5 and 6 in Appendix B): - Table 5 (CIP deviations on DollarExp, control=CDS): - Advanced economies coefficient = −1.06*** (0.243); Emerging = 4.69*** (0.985) - Observations: 210, 139, 57. - Table 6 (banks/non-banks with CDS control): - DollarExp_bk (advanced) = −1.36** (0.514); (emerging) = 1.18 (1.148) - DollarExp_nb (advanced) = −0.76*** (0.078); (emerging) = 7.23*** (1.200) - Observations: 210, 139, 57. - Summary statistics of CIP deviations (Table 2; annual CIP deviations 2013–2024): - All countries (Obs 236): P10 = −90.11, P25 = −59.48, P50 = −33.09, P75 = −8.32, P90 = 3.18 - Advanced economies (Obs 142): P10 = −69.58, P25 = −50.56, P50 = −23.51, P75 = −9.73, P90 = −4.3 - Emerging markets (Obs 94): P10 = −108.42, P25 = −75.49, P50 = −44.73, P75 = −2.09, P90 = 53.21 - Main conclusions and policy-relevant implications - The cross-sectional sign of the correlation between net dollar exposures and CIP deviations depends on whether variations in CIP deviations originate from demand-side or supply-side factors. - Demand-side dominated (typical of advanced economies with common global bank suppliers): negative correlation between net dollar exposures and CIP deviations. - Supply-side dominated (typical of emerging markets with heterogeneous arbitrage capacity, capital controls, and differing haircuts γ): positive correlation between net dollar exposures and CIP deviations. - Practical implication: Determining whether observed CIP deviations across countries are driven primarily by hedging demand or by local supply frictions (e.g., higher haircuts, limited cross-border arbitrage) is crucial for interpreting the link between CIP deviations and countries’ dollar exposures and for designing policies that affect dollar funding and hedging markets. *Source: wpiea2026169-source-pdf* ### 3.1 Rapid increase in USD exposures driven by non-banks ### 3.1 Rapid increase in USD exposures driven by non-banks ### Trends in USD exposures - USD-denominated assets amounted to 42 trillion in 2024. - USD-denominated assets nearly doubled from 19 trillion in 2013. - Foreign holdings of USD bonds and equities increased sixfold in the last two decades (from $5.5 trillion in 2002) (Du et al. (2024) cited). - As a share of GDP, the average for countries in the sample rose from 132% in 2013 to 195% in 2024. - USD-denominated liabilities are much smaller relative to assets—about forty percent less as of 2024. - Absolute USD liabilities increased over the past decade but less than assets; foreign holdings of USD-denominated liabilities relative to GDP remained roughly unchanged, peaking in 2015 and declining steadily thereafter. ### Investor composition: banks vs non-banks - Growth in USD-denominated assets is primarily driven by rising non-banks’ holdings; USD assets held by the banking sector have remained stable. - For USD-denominated liabilities, both banks and non-banks show stable patterns. - Increase in non-banks’ USD assets aligns with Du et al. (2024), which documents greater allocation to USD securities across insurance, pensions and mutual funds. - Drivers of non-bank increases: higher allocations to non-domestic investments and active rebalancing within non-domestic investments toward USD-denominated assets. ### Cross-country heterogeneity (advanced economies vs emerging markets) - Emerging market economies (EMs) exhibit more balanced USD assets and USD liabilities; these have remained aligned over the past decade. - In advanced economies (AEs), USD assets noticeably exceed USD liabilities; the gap widened due to a rapid increase in USD assets. - The rise in global USD exposure has been primarily driven by advanced economies. - Non-banks’ dominance in AEs: - In 2013, banks and non-banks each held roughly half of USD assets. - By 2024, non-banks accounted for over 65 percent of USD assets in AEs, leaving banks with about 35 percent. - In EMs, banks and non-banks hold similar amounts of USD assets. - USD liabilities composition: non-banks constitute a larger share of liabilities than banks in both AEs and EMs, especially in EMs; liability composition has remained relatively stable. - Net dollar positions: - In AEs, both banks and non-banks are net holders of dollar assets (more dollar assets than liabilities). - In EMs, only banks are net dollar asset holders; non-banks have more dollar liabilities than assets. ### Marginal investors of USD asset growth (regression evidence) - Regression setup (following Fang et al. (2025)): changes in USD assets by investor type scaled by lagged total USD assets regressed on changes in total USD assets (coefficients for banks and non-banks sum to one); country fixed effects included; Driscoll-Kraay standard errors used. - Key estimates (Table (1) summary): - Advanced economies: non-banks drive 83% of the increase in USD assets; banks drive 18%. - Emerging markets: non-banks account for 27% of the increase; banks account for 73%. - Conclusion: non-bank investors are the primary marginal investors in USD assets in advanced economies; banks are the primary marginal investors in EMs. ### USD exposures scaled by FX market depth - USD exposures are scaled by FX turnover using BIS Triennial Survey daily FX turnover figures converted to monthly; total USD assets divided by corresponding monthly FX turnover volumes. - Scaling by FX turnover provides a measure of exposure relative to FX market depth; deeper FX markets can accommodate hedging demands more easily. - The level of USD assets scaled by GDP does not perfectly correlate with USD assets scaled by FX turnover. - Country examples highlighting disparity: - Luxembourg: relatively high USD assets to GDP, but comparatively lower USD assets scaled by FX turnover due to the depth of the Euro FX market. - Taiwan: substantial USD assets relative to the size of its FX market (high when scaled by FX turnover) but smaller ratio when scaled by GDP, reflecting a relatively shallow FX market for the Taiwanese dollar. - Note on BIS Triennial Survey timing: since the Triennial Survey is conducted every three years, reported numbers for year t are applied to years t−1 and t+1 as approximations. ### CIP deviations: definition and measurement - One-period CIP condition (cash vs FX swap markets) and introduction of CIP deviation x_{i/$t} as a wedge when equality does not hold. - In log terms, CIP deviation: - x_{i/$t} = y_{$t} − y_{i t} + ρ_{i/$t}, where ρ_{i/$t} is the forward premium (log difference between forward and spot). - Sign interpretation: - Negative CIP deviation: USD borrowing costs more in the FX swap market (forward rate too low); investors selling dollar forwards incur a cost. - Positive CIP deviation: USD borrowing more expensive in the cash market; investors buying dollar forwards incur a cost. - Focus on 3-month CIP deviations: - Chosen to mitigate credit risk contamination and because short-term forwards and interbank lending markets are more liquid. - Correlation between 3-month CIP deviations and sovereign CDS spreads in the sample is near zero, supporting that the short-term measure does not reflect sovereign default risk. - Summary statistics (Table (2)): - CIP deviations are mostly negative in the sample; the median for both AEs and EMs is below zero. ### Empirical relationships between USD exposures and CIP deviations - Baseline cross-sectional regression: CIP_it = γ_t + β DollarExp_it + ν_it, where DollarExp_it is net dollar exposure (dollar assets minus liabilities) divided by monthly FX turnover; time fixed effects included; Driscoll-Kraay standard errors used. - Main findings (Table (5)): - Advanced economies: relationship between net dollar exposures and CIP deviations is statistically and significantly negative. - Emerging markets: relationship is statistically and significantly positive. - Interpretation for AEs: higher net dollar exposures increase hedging demand, pushing up dollar borrowing cost in the FX swap market and making CIP deviations more negative. - For EMs: positive coefficient partly aligns with Dao and Gourinchas (2025) who find a weak positive correlation due to measurement challenges; robustness to inclusion of sovereign CDS spreads is confirmed in Appendix. - Banks vs non-banks analysis (regression CIP_it = γ_t + β1 DollarExp_bk,it + β2 DollarExp_nb,it + ν_it): - Advanced economies (Table (6)): coefficients on net dollar exposures for both banks and non-banks are significantly negative. - Emerging markets (Table (6)): coefficient on banks’ net dollar exposure is statistically insignificant; coefficient on non-banks’ net dollar exposure is statistically positive. - Robustness: findings remain robust to inclusion of sovereign credit risk as a control. - Interpretation for EMs: non-banks typically have more dollar liabilities than assets; under the hedging demand channel, hedging of open dollar liability positions by non-banks would supply dollars into the FX swap market and lower dollar borrowing costs (implying a negative relationship). The empirically observed significantly positive coefficient for non-banks in EMs suggests an alternative mechanism—linked to substantial dollar-denominated debt held by non-bank sectors in EMs—is driving CIP deviations in these countries. *Source: wpiea2026169-source-pdf - 3.1 Rapid increase in USD exposures driven by non-banks* ### 4.3 Divergent relations between CIP deviations and dollar exposures ### 4.3 Divergent relations between CIP deviations and dollar exposures ### Overview - Purpose: Explain why CIP deviations can be positively related to non-banks’ net dollar exposure in emerging markets by linking CIP deviations to firms’ currency choice for debt issuance and asset holdings. - Approach: Provide an illustrative example of dollar debt issuance, then present a formal model (inspired by Liao (2020)) that endogenizes CIP deviations via firms’ asset/liability currency choices and constrained-capital traders who arbitrage CIP deviations. - Focus: Rationalize cross-sectional variation in CIP deviations across countries driven by the stock of net dollar exposures, rather than high-frequency time-series dynamics. ### CIP deviations and dollar debt issuance (illustrative example) - Local-currency borrowing cost: r_i^t (equation (7) in source). - USD borrowing cost (hedged via forward/swap): (1/S_{i/$}^t)(1 + r_ $^t)F_{i/$}^t − 1 (equation (8) in source). - Borrowing-cost differential Φ defined (equation (9) in source) and log-linearized to: - Φ = r_i^t − r_ $^t − f_t + s_t (equation (10) in source). - With risk-free yields included, Φ decomposes into (equation (11) in source): - Φ = (r_i^t − y_i^t) − (r_ $^t − y_ $^t) − (y_ $^t − y_i^t + ρ_{i/$}^t) - Interpretation: A more negative CIP deviation increases Φ, making USD borrowing cheaper; firms issuing USD liabilities benefit from negative CIP deviations by effectively providing dollar funding where there is a dollar shortage. - Mechanism implication: More negative CIP deviations incentivize firms to increase dollar debt issuance → higher USD liabilities → smaller net USD exposures (i.e., a potential positive relationship between CIP deviations and net dollar exposures across countries when driven by supply factors). ### Model: firm choices, trader arbitrage, and equilibrium CIP deviations - Firm side (liabilities): Representative firm issues debt D, chooses USD debt share μ_d given hedging ratio h_d, maximizes CRRA expected utility (equations (12)-(13) in source). - End-of-period wealth W expression (equation (13) in source) distinguishes locally funded debt, hedged USD portion (cost r_ $^d + x), and unhedged USD portion (cost r_ $^d + φ) where φ has mean φ̄ and variance V. - Optimal USD debt share (first-order/Taylor expansion) (equation (14) in source): - μ*_d = [r_i^d − (r_ $^d + h_d x + (1−h_d) φ̄)] / [ρ_d (1−h_d)^2 V] - Comparative statics: larger ρ_d or larger V → smaller μ*_d. Higher x (more positive CIP deviation) raises USD hedging cost → reduces μ*_d; negative x lowers USD funding cost → increases μ*_d. - Firm side (assets): Dollar asset holder funded by LC debt chooses USD asset share μ_a with analogous setup (equations (15)-(17) in source): - Optimal USD asset share (equation (17) in source): - μ*_a = [r_ $^a + h_a x + (1−h_a) φ̄ − r_i^a] / [ρ_a (1−h_a)^2 V] - Negative x (x < 0) implies a hedging cost that reduces dollar asset returns, lowering μ*_a ceteris paribus. - Trader (supply) side: Constrained-capital forward/swap trader allocates capital between exploiting CIP deviations x and alternative investment I, with haircut H = γ s (equations (18)-(20) in source). - Trader objective (equation (19) in source) leads to optimal forward position s (equation (21) in source) and, under assumption W = φ_0 / φ_1 (from Liao (2020)), simplified relations: - s = − x / (γ^2 φ_1) (equation (23) in source) - Equivalently x = − γ^2 φ_1 s (equation (24) in source) - Market clearing for forwards (equation (25) in source): - μ_d D h_d − μ_a A h_a + s = 0 - Substituting optimal μ*_d, μ*_a, and s yields equilibrium condition (equation (26) in source): - D h_d [r_i^d − (r_ $^d + h_d x + (1−h_d) φ̄)] / [ρ_d (1−h_d)^2 V] − A h_a [r_ $^a + h_a x + (1−h_a) φ̄ − r_i^a] / [ρ_a (1−h_a)^2 V] − x / (γ^2 φ_1) = 0 - Interpretation: First two terms = net hedging demand from dollar debt and dollar asset holders (functions of x); last term = limited-capital traders’ supply of forwards. Equilibrium x determined by interaction of demand and supply. ### Equilibrium intuition and cross-sectional implications - Demand-driven variation: - If cross-country differences in x stem from demand: more dollar assets (higher A) raise hedging demand for selling dollar forwards → CIP deviation must become wider (more negative) to clear market, so ∂x*/∂A < 0. - More dollar liabilities (higher D) increase demand for buying dollar forwards → x must be less negative / more positive, so ∂x*/∂D > 0. - Net implication: Demand-side driven variation produces a negative cross-sectional correlation between CIP deviations and net dollar exposures (illustrated in first panel of Figure (12) in source). - Supply-driven variation: - If cross-country differences stem from supply (e.g., higher haircut γ → constrained traders supply fewer swaps), then ∂x*/∂γ < 0 (higher γ pushes x more negative). - More negative x (wider negative CIP) reduces μ*_a (fewer dollar assets) and increases μ*_d (more dollar debt) → country’s net dollar exposure falls. - Net implication: Supply-side driven variation can produce a positive cross-sectional correlation between CIP deviations and net dollar exposures (illustrated in second panel of Figure (12) in source). - Institutional interpretation: - Advanced economies: CIP deviations largely shaped by hedging demand (shared global bank suppliers), producing a negative correlation between net dollar exposures and CIP deviations. - Emerging markets: Cross-border arbitrage constrained by capital controls, taxation, legal/regulatory differences, and higher haircuts; supply-side differences dominate, producing a positive correlation. ### Empirical findings (selected quantitative results from tables and figures in source) - Cross-sectional relationships (Table 3 in source; regression of CIP deviations on DollarExp): - All countries (column (1)): coefficient = −0.15 (standard error 0.245) - Advanced economies (column (2)): coefficient = −1.06*** (standard error 0.243) - Emerging markets (column (3)): coefficient = 3.43*** (standard error 0.639) - Observations: 210 (all), 139 (advanced), 71 (emerging). Time Fixed Effects: Yes. - Banks vs non-banks (Table 4 in source; regressions of CIP deviations on bank and non-bank DollarExp): - DollarExp_bk (banks): - All: −0.78* (0.366) - Advanced: −1.36** (0.514) - Emerging: 1.14 (0.942) - DollarExp_nb (non-banks): - All: 0.32 (0.512) - Advanced: −0.76*** (0.078) - Emerging: 6.10*** (0.961) - Observations: 210, 139, 71. Time Fixed Effects: Yes. - Robustness with CDS controls (Tables 5 and 6 in Appendix B): - Table 5 (CIP deviations on DollarExp, control=CDS): - Advanced economies coefficient = −1.06*** (0.243); Emerging = 4.69*** (0.985) - Observations: 210, 139, 57 (note: sample differences when adding controls). - Table 6 (banks/non-banks with CDS control): - DollarExp_bk (advanced) = −1.36** (0.514); (emerging) = 1.18 (1.148) - DollarExp_nb (advanced) = −0.76*** (0.078); (emerging) = 7.23*** (1.200) - Observations: 210, 139, 57. - Summary statistics of CIP deviations (Table 2 in source; annual CIP deviations 2013–2024): - All countries (Obs 236): P10 = −90.11, P25 = −59.48, P50 = −33.09, P75 = −8.32, P90 = 3.18 - Advanced economies (Obs 142): P10 = −69.58, P25 = −50.56, P50 = −23.51, P75 = −9.73, P90 = −4.3 - Emerging markets (Obs 94): P10 = −108.42, P25 = −75.49, P50 = −44.73, P75 = −2.09, P90 = 53.21 ### Main conclusions and policy-relevant implications (from model and empirical results) - Cross-sectional sign of correlation between net dollar exposures and CIP deviations depends on whether CIP deviations’ variations originate from demand-side or supply-side factors. - Demand-side dominated (typical of advanced economies with common global bank suppliers): negative correlation between net dollar exposures and CIP deviations. - Supply-side dominated (typical of emerging markets with heterogeneous arbitrage capacity, capital controls, and differing haircuts γ): positive correlation between net dollar exposures and CIP deviations. - Practical implication: Understanding whether observed CIP deviations across countries are driven primarily by hedging demand or by local supply frictions (e.g., higher haircuts, limited cross-border arbitrage) is crucial for interpreting the link between CIP deviations and countries’ dollar exposures, and for designing policies that affect dollar funding and hedging markets. *Source: Excerpt from "4.3 Divergent relations between CIP deviations and dollar exposures" (wpiea2026169-source-pdf).* ### References ### wpiea2026169-source-pdf - References ### Covered interest parity, CIP deviations, and international capital market frictions - Avdjiev, S., Du, W., Koch, C., and Shin, H. S. (2019). The dollar, bank leverage, and deviations from covered interest parity. American Economic Review: Insights, 1(2):193–208. - Baba, N. and Packer, F. (2009). Interpreting deviations from covered interest parity during the financial market turmoil of 2007–08. Journal of Banking & Finance, 33(11):1953–1962. - Cerutti, E. M., Obstfeld, M., and Zhou, H. (2021). Covered interest parity deviations: Macrofinancial determinants. Journal of International Economics, 130:103447. - Cerutti, E. and Zhou, H. (2024). Uncovering cip deviations in emerging markets: Distinctions, determinants, and disconnect. IMF Economic Review, 72(1):196–252. - Du, W. and Schreger, J. (2016). Local currency sovereign risk. The Journal of Finance, 71(3):1027–1070. - Du, W. and Schreger, J. (2022). Cip deviations, the dollar, and frictions in international capital markets. In Handbook of International Economics, volume 6, pages 147–197. Elsevier. - Du, W., Tepper, A., and Verdelhan, A. (2018). Deviations from covered interest rate parity. The Journal of Finance, 73(3):915–957. - Iida, T., Kimura, T., and Sudo, N. (2018). Deviations from covered interest rate parity and the dollar funding of global banks. 55th issue (September 2018) of the International Journal of Central Banking. - Liao, G. Y. (2020). Credit migration and covered interest rate parity. Journal of Financial Economics, 138(2):504–525. - Du, W., Hèbert, B., and Huber, A. W. (2023). Are intermediary constraints priced? The Review of Financial Studies, 36(4):1464–1507. - Du, W., Huber, A. W., et al. (2024). Dollar asset holdings and hedging around the globe. Technical report, National Bureau of Economic Research. ### Dollar funding, bank leverage, and global banking behavior - Avdjiev, S., Du, W., Koch, C., and Shin, H. S. (2019). The dollar, bank leverage, and deviations from covered interest parity. American Economic Review: Insights, 1(2):193–208. - Ivashina, V., Scharfstein, D. S., and Stein, J. C. (2015). Dollar funding and the lending behavior of global banks. The Quarterly Journal of Economics, 130(3):1241–1281. - McCauley, R. N., McGuire, P., and Sushko, V. (2015). Dollar credit to emerging market economies. BIS Quarterly Review December. - Sushko, V., Borio, C. E., McCauley, R. N., and McGuire, P. (2016). The failure of covered interest parity: Fx hedging demand and costly balance sheets. - Du, W., Hèbert, B., and Huber, A. W. (2023). Are intermediary constraints priced? The Review of Financial Studies, 36(4):1464–1507. ### Currency composition of reserves, external balance sheets, and sovereign debt - Allen, C. and Juvenal, L. (2025). The role of currencies in external balance sheets. Journal of International Economics, page 104105. - Benetrix, A., Gautam, D., Juvenal, L., and Schmitz, M. (2019). Cross-border currency exposures. International Monetary Fund. - Chinn, M. D., Ito, H., and McCauley, R. N. (2022). Do central banks rebalance their currency shares? Journal of International Money and Finance, 122:102557. - Ito, H. and McCauley, R. N. (2020). Currency composition of foreign exchange reserves. Journal of International Money and Finance, 102:102104. - Du, W. and Schreger, J. (2016). Local currency sovereign risk. The Journal of Finance, 71(3):1027–1070. - Fang, X., Hardy, B., and Lewis, K. K. (2025). Who holds sovereign debt and why it matters. The Review of Financial Studies, page hhaf031. - Krishnamurthy, A. and Lustig, H. N. (2019). Mind the gap in sovereign debt markets: The us treasury basis and the dollar risk factor. In 2019 Jackson Hole Economic Symposium, Stanford University Graduate School of Business Research Paper, number 3443231. - Jiang, Z., Krishnamurthy, A., and Lustig, H. (2021). Foreign safe asset demand and the dollar exchange rate. The Journal of Finance, 76(3):1049–1089. ### Measurement, data sources, and cross-border positions - Judson, R. and Kim, N. (2026). Measuring cross-border securities positions: Explaining asymmetries between us treasury tic and imf pip (formerly cpis) data. - Benetrix, A., Gautam, D., Juvenal, L., and Schmitz, M. (2019). Cross-border currency exposures. International Monetary Fund. - Du, W., Huber, A. W., et al. (2024). Dollar asset holdings and hedging around the globe. Technical report, National Bureau of Economic Research. *Global Dollar Exposure and Its Relationship with CIP Deviations Working Paper No. WP/2026/169* --- _Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026169-source-pdf.pdf_