## wpiea2025057-print-pdf ## Source details **Canonical URL:** [wpiea2025057-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025057-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025057-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025057-print-pdf.pdf.json) --- ### Theoretical framework: constrained intermediation and CIP deviations - Model focuses on US dollar–funded intermediary j with USD and local currency (LC) assets/liabilities; USD positions denoted with an asterisk. - Balance sheet (USD terms) and off-balance-sheet forwards: - Balance sheet identity: W^*_{j,t} + D^*_{j,t} + D_{j,t}/E_t = B^*_{j,t} + B_{j,t}/E_t. - Off-balance-sheet forwards F^*_{j,t} sold at price F_t, gross USD return (F_t / E_{t+1} − 1). - Intermediary next-period wealth (USD) (equation (3)) decomposes returns into: - return on net worth at USD interbank rate R^*_t; - excess return on USD reserves (˜R^*_t / R^*_t − 1); - excess return on local-currency assets vs liabilities (˜R_t / R_t − 1); - risk-free return on covered interest strategy (R^*_t / R_t F_t / E_t − 1) — the CIP basis; - return on net local-currency exposure (˜R_t B_{j,t} − R_t D_{j,t} + F_t F^*_{j,t}). - No open LC positions constraint (equation (4)): ˜R_t B_{j,t} + F_t F^*_{j,t} = R_t D_{j,t}; yields riskless wealth equation (5). - Regulatory/internal liquidity constraint (equation (6)): B^*_{j,t} ≥ α_t | F^*_{j,t} |; α_t depends on aggregate intermediary exposure (equation (7)). - First-order condition implies intermediary j’s CIP basis (equation (8)): - R^*_t / R_t F_t / E_t − 1 = μ_{j,t} a ( | ̄F^*_t | / ̄W^*_t )^α sign(F^*_{j,t}). - Aggregation (equation (9)) gives common price: - R^*_t / R_{i,t} F_{i,t} / E_{i,t} − 1 = μ_t a ( | ̄F^*_{i,t} | / ̄W^*_{i,t} )^α sign( ̄F^*_{i,t} ). - Key model implications: - If constraint non-binding (μ_t = 0 or a = 0) then CIP holds (basis = 0). - Sign of basis = sign(aggregate net forward dollar supply ̄F^*_{i,t}); ̄F^*_i > 0 ⇒ positive basis; ̄F^*_i < 0 ⇒ negative basis. - Size of basis scales with | ̄F^*_{i,t} | / ̄W^*_{i,t} and funding tightness μ_t. - Explains asymmetry: AE currencies (net long dollar exposure, sign(̄F^*_{i,t}) < 0) spike to negative CIP in stress; EMs (net short dollar exposure, sign(̄F^*_{i,t}) > 0) spike to positive CIP. ### Definition and decomposition of CIP deviations - Standard CIP deviation formula: - τ_{i,n,t} = y^{rf}_{USD,n,t} + ρ_{i,n,t} − y^{rf}_{i,n,t}. - y^{rf}_{i,n,t}: country i risk-free interest rate (interbank or zero-coupon sovereign yield). - ρ_{i,n,t}: annualized forward premium converting currency i into USD and back over n. - Accounting identity driving variations: - Interest rate differential: y_{USD,n,t} − y_{i,n,t} - Forward premium: ρ_{i,n,t} (log difference between forward and spot) - Empirical patterns: - Since the GFC, in AEs the LIBOR CIP basis became mostly negative; time-series volatility largely driven by forward premium (correlations: change in 3-month LIBOR basis vs forward premium R^2 = 77 percent; vs interest differential R^2 = 7 percent). - Across AE currencies average correlation between change in LIBOR basis and change in forward premium is high; change vs interest differential low (monthly change correlations reported: 0.96 with forward premium; 0.15 with interest differential). ### Stylized facts — Advanced Economies (AEs) - Instruments and sample: daily LIBOR basis for G-7 currencies: EUR, CHF, JPY, GBP, AUD, CAD. - Observations: - LIBOR basis widened after the GFC with regulatory constraints (Basel III). - CIP deviations mostly negative (except AUD under paper’s convention). - Negative spikes in stress episodes: synthetic dollar rate > cash dollar rate; basis becomes more negative during crises. ### Stylized facts — Emerging Markets (EMs) - EM sample (deliverable offshore forwards): MXN, ZAR, TRY, ISL; BRL treated with deliverable-forward group due to active onshore forwards. - Observations: - EM CIP deviations much more volatile than AEs (vertical axis scale roughly ten times larger for EMs, especially BRL and TRY). - EM CIP deviations vary between positive and negative across and within countries; no clear regime shift post-GFC (EMs have long been volatile). - Post-GFC average interest rate differential sign differs: positive for AEs, negative for EMs; forward premium often offsets differential with opposite sign. ### Purified (Supranational-based) CIP deviations — motivation and construction - Problem with naive government-bond CIP in EMs: - Local government yields embed credit spread I_Gov_i,n,t and convenience yield λ_Gov_i,n,t; φ_Gov_i,n,t = ˆλ_Gov_i,n,t − ˆI_Gov_i,n,t + τ_i,n,t (equation (11)). - Sovereign credit risk and liquidity premia contaminate φ_Gov_i,n,t; sovereign CDS do not fully hedge local-currency credit risk. - Solution: use supranational (SSA) bonds issued in EM local currencies to proxy risk-free local-currency rates: - SSA bonds typically AAA, low default risk, often zero risk weighting and HQLA eligibility; settle offshore easing investor access. - Construct supra-specific z-spreads over entire yield curve to compute supra-specific CIP basis: - φSupra i,j,t,t+n ≡ ySupra USD,j,t,t+n + ρi,t,t+n − ySupra i,j,t,t+n (Eq. (12)). - Z-spread relation under no-arbitrage: φ i,j,t,t+n = s$ j,t,t+n − s i j,t,t+n (Eq. (15)). - Extract currency-specific pure CIP τ_i,t from φ i,j,t by regressing φSupra i,j,t + y rf USD,t − ySupra USD,j,t = τ i,t + λSupra i,j + α_j BidAsk i,j,t + ε i,j,t (Eq. (17)) using at least two supranational issuers j per currency. - Tenor choice: - 1 year chosen as baseline tenor because modal remaining time-to-maturity across issuer–currency pairs is 1 year. ### Empirical implementation and sample constraints - Supranational issuers: ADB, EBRD, EIB, IBRD, IFC, KfW; IBRD, KfW, EIB most active. - Data requirement: same issuer traded in both USD and local currency with adjacent residual maturities (within five months) and observed secondary market prices to compute z-spreads; limits usable bonds and currencies. - For three EM currencies with most liquid supra trading (BRL, TRY, ZAR), purified CIP series stretch back before the GFC; six EM currencies provide sufficient data for time-series analysis. ### Key empirical findings — purified CIP vs conventional measures (Table 5, summary) - 1-year tenor mean CIP deviations (mean (std. dev.), in bps): - BRL: LIBOR -113.11 (72.12); Gov. bond -196.80 (103.14); Purified 9.39 (39.82) - CNY: LIBOR -24.43 (144.18); Gov. bond 36.02 (128.12); Purified -30.92 (60.13) - IDR: LIBOR -50.67 (136.46); Gov. bond 63.73 (117.85); Purified 35.10 (35.69) - INR: LIBOR -51.63 (134.08); Gov. bond -82.08 (116.79); Purified 3.83 (34.86) - MXN: LIBOR — (1-year LIBOR not available); Gov. bond -39.07 (47.67); Purified 8.63 (21.8) - RUB: LIBOR -56.82 (43.05); Gov. bond 21.23 (57.86); Purified -39.08 (196.62) - TRY: LIBOR -21.74 (232.70); Gov. bond -3.51 (261.21); Purified 152.53 (82.98) - ZAR: LIBOR 0.42 (31.81); Gov. bond -0.94 (54.07); Purified -28.38 (36.17) - Cross-sectional pattern: - EMs with large net external liabilities tend to have positive purified CIP (Türkiye, Mexico, Brazil). - EMs with net external assets tend to have negative purified CIP (China, Russia). - Purified CIP is often orders of magnitude smaller than conventional measures and sometimes flips sign relative to conventional CIP. ### Empirical decomposition and frequency/tenor effects - Bivariate regression of change in 3-month LIBOR basis on change in forward premium: R-squared = 77 percent; on change in interest differential: R-squared = 7 percent (AEs). - Monthly average correlation coefficients for 3-month LIBOR basis changes: - Correlation with forward premium: 0.96. - Correlation with interest rate differential: 0.15. - Forward premium dominates monthly variation in 3-month CIP basis; interest differential becomes more correlated with CIP as tenor increases to 6-months and 12-months. - Tenor effects: magnitude and volatility differences across EM vs AE pronounced at 12-month maturity (roughly one order of magnitude larger for EMs). ### Regression evidence: time-series dynamics and hedging channel (baseline framework and results) - Model-implied approximation (α = 1): CIP_i ≈ ̄μ a ( ̄F*_i / ̄W*_i ). - Dollar gap measures: - USDGAP_aug_t = Ext.DebtAssets_USD_t − Ext.DebtLiabilities_USD_t − Ext.DebtLiabilities_LC_t / GDP_t (equation (19)). - Baseline first-difference regression (equation (24)): - ∆CIP_it = α_i + β_1 ∆dollar_t + β_2 ∆dollar_t × (−USDGAP_i) + γ′ X_it + ε_it. - Expect β_1 < 0; β_2 > 0. - Conventional CIP regressions (Table 6) often fail to deliver predicted signs for EMs. - Supranational CIP regressions (Table 7, six EMs; monthly; Feb 2006–Dec 2020) — selected estimates (exact reported): - Column (1), 1-year Supra (six EMs): - CIP_t−1 = -0.137*** (0.020) - ∆dollar_t = -3.654** (1.825) - ∆dollar_t ∗ −(USDGAP_i) = 0.198** (0.087) - −(USDGAP_i) = 0.260* (0.135) - Observations = 801; Number of EM currencies = 6; Within R2 = 0.0849. - Column (2), top-3 liquid EM supras: - CIP_t−1 = -0.121*** (0.021) - ∆dollar_t = -5.806*** (1.807) - ∆dollar_t ∗ −(USDGAP_i) = 0.261*** (0.075) - −(USDGAP_i) = 0.223** (0.113) - Observations = 493; Number of EM currencies = 3; Within R2 = 0.110. - Interpretation: - Using purified supra CIP yields ∆dollar_t coefficients negative (consistent with dollar funding channel). - Interaction ∆dollar_t ∗ −(USDGAP_i) positive and statistically significant (hedging channel confirmed); stronger for most liquid supranational markets (Brazil, Türkiye, South Africa). - Conventional LIBOR and government bond CIP measures in EMs contaminate results via credit and convenience premia. ### Magnitudes and illustrative impacts (Figure 15 and related text) - For a country with zero dollar gap, a typical dollar appreciation estimated to lower CIP by 8 bps. - If net hedging demand positive with dollar liabilities and local currency external debt at 30 percent of GDP, CIP estimated to increase by 7.5 bps instead. - (−USDGAP) ranged from -41 to 67 percent of GDP in EM sample. - Median monthly change in purified CIP basis = 8 bps. - A 5 percent appreciation in the broad dollar predicts a 12.5 bps larger increase in CIP for a 10 pct of GDP higher dollar gap; differential is -7.3 bps when dollar depreciates by same amount. - Standard deviation in dollar gap: 15 pct of GDP across all EMs; 7 pct for EMs in estimation sample. ### Robustness: alternative global-dollar-cycle measures (Table 8) - Alternative μ_t proxies used: VIX, −GFCy (Miranda-Agrippino and Rey), Treasury basis x_Treas. - Selected coefficients (exact reported): - ∆ log VIX: coefficient -0.380***; interaction ∆ log VIX * (−USDGAP_i) = 0.014** (col.1). - ∆−GFCy: coefficient -25.56***; interaction ∆−GFCy * (−USDGAP_i) = 1.016*** (col.3). - ∆x_Treas: coefficient 53.41***; interaction ∆x_Treas * (−USDGAP_i) = -2.284** (col.5). - ∆dollar coefficients across specifications: -4.629***, -3.554**, -4.656***; interactions ∆dollar * (−USDGAP_i) = 0.228***, 0.175***, 0.206***. - Observations and panel details reported: Observations = 548, 493, 548, 493, 495, 493; Number of currencies = 33; Within R2 range reported (e.g., 0.101, 0.124). - Broad dollar remains most statistically significant driver when jointly controlled. ### Intermediary net worth (leverage-capital) interaction (Table 9 and Figure 16) - Model expansion: CIP_it = a κ_i μ_t η_t ̄F*_it / Y_it (eq. (25)), where intermediary leverage η_t = debt / net worth. - Estimating equation includes ∆η_t and interactions (equation (26)). - Proxy for η_t: market capital ratio of primary dealer counterparties (He et al. (2017)). - Selected empirical findings (sample: BRL, TRY, ZAR; exact reported coefficients): - Column 2 full specification highlights: - y_{t-1} = -0.131***. - ∆η_t = -2.092***. - ∆η_t × (−USDGAP_i) = 0.114*. - ∆dollar_t = -5.535**. - ∆dollar_t × (−USDGAP_i) = 0.379. - ∆dollar_t ∆η_t × (−USDGAP_i) = -0.106***. - Observations = 493; Number of currencies = 3; Within R2 = 0.154. - Economic magnitudes (Figure 16): - 1 standard deviation broad dollar appreciation = 1.35 percent month on month. - With unchanged intermediary leverage-capital ratio, broad dollar appreciation lowers CIP for low dollar-gap countries and increases it for high dollar-gap countries. - If intermediary leverage-capital ratio increases by 1.5 standard deviations, negative impact on CIP is 3–4 times larger for low dollar-gap countries; for larger dollar liability gaps the differential shrinks and can turn positive as hedging channel dominates. - Conventional CIP measures fail to capture these predicted leverage interactions due to credit-risk contamination. ### Cross-sectional evidence: dollar gap correlations (figures and summary) - Conventional 3-month LIBOR basis correlation with dollar gap: - AE currencies: sample correlation = -0.59. - EM currencies: no correlation observed. - Purified supra CIP vs USD Gap (eight EMs; figures): - 1-year Treasury CIP vs USD Gap: Corr = -0.07. - 1-year Supra CIP vs USD Gap: Corr = -0.78 (figure note) / reported turn from zero to -0.8 in fig.13 note. - 3-year Supra CIP vs USD Gap: Corr reported as -0.93 (panel d) and annotations report alternative Corr = -0.77 in panel (d). - Augmented USD Gap (includes external local-currency debt) materially affects measured gap for ZAR, MXN, TRY. - 1y Supra CIP vs Augmented USD Gap: Corr = -0.68 (figure 14b); alternative reported Corr = -0.29 in figure caption (both provided in figure). - Excluding outliers (RUB, CNY) steepens slopes; RUB and CNY display largest supranational volatility. ### Summary statistics and appendices (selected exact figures) - Appendix B, Table 12 (summary statistics, selected): - AE 3-month CIP deviation (Obs. 1,860): Mean: -19.8; Std. Dev.: 26.7. - EM 3-month CIP deviation (Obs. 3,133): Mean: -30.1; Std. Dev.: 217.8. - EM 12-month CIP deviation (Obs. 2,493): Mean: -153.8; Std. Dev.: 228.6. - Table 14 (absolute value of 1-year CIP, means and std. dev.): - BRL: Libor 113.11 (72.12); Bond 196.80 (103.14); Purified 32.22 (25.07). - CNY: Libor 105.72 (100.43); Bond 104.11 (74.08); Purified 43.98 (51.25). - IDR: Libor 506.78 (136.46); Bond 99.39 (89.50); Purified 28.37 (22.07). - INR: Libor 516.31 (134.08); Bond 115.99 (82.86); Purified 27.16 (22.04). - RUB: Libor 61.22 (36.47); Bond 40.26 (41.39); Purified 154.93 (126.40). - TRY: Libor 137.60 (188.66); Bond 168.86 (198.96); Purified 61.69 (76.39). - ZAR: Libor 25.53 (18.92); Bond 40.90 (35.28); Purified 33.67 (31.28). ### Policy implications and applications - Purified CIP basis provides cleaner empirical measure of intermediary pricing wedge (equation (8)) driven by balance sheet constraints, FX market shallowness, capital and liquidity regulations, and the “dollar premium”. - Uses: - Gauge potential impacts of spot and forward FX interventions, capital controls, and macroprudential policies on hedging costs and dollar funding. - Evaluate central bank swap lines’ effects: swap lines can lower dollar funding and hedging costs; benefits may be larger for EMs with shallower FX markets and large hedging needs. - Targeting CIP wedge can strengthen monetary transmission in EMs by narrowing wedge between domestic borrowing costs and synthetic local-currency borrowing rates implied by dollar financing. - Empirical conclusions: - When measured with purified supranational-based CIP, EM deviations from CIP are not necessarily larger post-GFC; proper measurement shows long-term hedging demand explains cross-sectional variation and time-series fluctuations driven by global dollar funding conditions and intermediary balance sheet capacity. *Extracted from content unit “wpiea2025057-print-pdf.”* ### section 3 presents our theoretical framework. This may be skipped on a first read if the reader ### wpiea2025057-print-pdf - section 3 presents our theoretical framework. This may be skipped on a first read if the reader ### Overview - Section introduces stylized facts about CIP deviations (CIP basis) in Emerging Markets (EMs) and Advanced Economies (AEs), their trend evolution, cyclical variation, cross-country differences, composition, and statistical properties. - Section structure referenced: theoretical framework in section 3 (may be skipped), estimation method in section 4, empirical results on drivers of CIP for EMs in section 5, conclusions and policy implications in section 6. ### Definition of the CIP deviation (CIP basis) - CIP deviation (aka the CIP basis) for currency of country i at maturity n and time t is measured according to the standard formula: τ_{i,n,t} = y^{rf}_{USD,n,t} + ρ_{i,n,t} − y^{rf}_{i,n,t} - Definitions within formula: - y^{rf}_{i,n,t} is the country i risk-free interest rate (often measured by an inter-bank money market rate or annualized yield on a zero-coupon sovereign bond in country i in its own currency) with maturity n at time t. - ρ_{i,n,t} is the annualized forward premium that converts currency of country i into US dollar at time t and back at time t+n. - Interpretation: the basis measures the return from a zero wealth investment strategy that borrows in the local currency and invests in the dollar cash market, hedging currency risk. Positive when cash dollar interest rate exceeds synthetic dollar rate constructed from domestic interest rate and currency hedges. In a frictionless and riskless setting CIP implies τ_{i,n,t} = 0. - Caveat: CIP deviations can reflect variations in relative risk priced in different currencies if y^{rf}_{USD,n,t} and/or y^{rf}_{i,n,t} are not truly risk-free. ### Advanced Economies (AEs) — stylized facts - Data and instruments: - Daily LIBOR basis for G-7 currencies comprising EUR, CHF, JPY, GBP, AUD, CAD (Figure 1). - Key observations: - The LIBOR basis has widened after the GFC, with the introduction of regulatory constraints such as Basel III leverage and liquidity coverage ratio. - CIP deviations are mostly negative, except for the AUD (given the paper's convention). - Negative CIP basis interpretation: direct/cash dollar interest rate is lower than the synthetic dollar rate (cost of borrowing in country i’s currency, exchanging for USD in the spot market and covering exchange risk with a forward). - CIP basis widens sharply / becomes more negative during stress episodes and crises — synthetic dollar rate is higher than cash dollar rate for most AE currencies, and this difference widens in stress episodes. ### Emerging Markets (EMs) — stylized facts - EM sample with deliverable offshore forwards shown includes: MXN, ZAR, TRY, ISL; Brazil (BRL) treated with deliverable-forward group due to active onshore FX forwards/futures priced in close lockstep with offshore NDFs. - Key observations: - CIP deviations are much more volatile in EMs than AEs, with the scale of the vertical axis being ten times as large for the former than the latter, especially for BRL and TRY. - EM CIP deviations vary more between positive and negative values, both across and within countries. - No visible regime shift after the GFC for EMs; CIP deviations in EMs have always been relatively large and volatile. - If anything, CIP deviations in AEs have become more similar to those in EMs after the GFC. - On average, the post-GFC interest rate differential has opposite signs in AEs and EMs (positive for AEs and negative for EMs). In each group the interest rate differential is partly or fully offset by a forward premium of opposite sign: negative for AEs (priced to appreciate against the US dollar) and positive for EMs (priced to depreciate against the US dollar). ### CIP basis components and decomposition - Accounting identity: variations in the CIP basis are driven by two components: - Interest rate differential between US dollar and local currency rates: y_{USD,n,t} − y_{i,n,t} - Forward premium: ρ_{i,n,t} equal to the log difference between the forward and the spot exchange rate - Empirical decomposition (Figure 3): - Decomposition of daily 3-month CIP deviation into interest differential and forward premium components, averaged over AEs with negative CIP basis (simple average across EUR, CHF, JPY, GBP, in bps). - Up until the GFC, forward premium completely offset the interest rate differential between USD money market and corresponding AE money market, so CIP largely held. - Since the GFC, the basis in most AEs became negative as the interest differential was not sufficiently offset by a large enough forward premium. - Since the GFC, time-series volatility of the LIBOR CIP deviation has been primarily driven by volatility in the forward premium; interest differentials tend to be slow-moving and smoother. - High-frequency volatility in the LIBOR basis is mainly driven by variation in the forward premium, especially during stress episodes (typical negative spike in the CIP basis of AE currencies in stress episodes is driven by a spot dollar appreciation/forward dollar discount in excess of what is predicted by the increase in interest differential). - Across AE currencies, the correlation coefficient between the change in LIBOR basis and change in forward premium is on average (figure reference; exact correlation value not provided in excerpt). ### Frequency and tenor effects - Stylized facts for daily CIP deviations carry over to: - Monthly frequency - Alternative government-based CIP basis - Longer tenor of 12-months - Magnitude differences: - Difference in magnitude and volatility of CIP deviations in EM relative to AE is particularly pronounced at 12-month maturity, where it is roughly one order of magnitude larger. *Source: Excerpt from wpiea2025057-print-pdf.* ### 0.87 while it is only 0.27 for the interest differential. Similarly, a bivariate regression of the ### wpiea2025057-print-pdf - 0.87 while it is only 0.27 for the interest differential. Similarly, a bivariate regression of the ### Empirical decomposition of the 3-month LIBOR CIP basis - A bivariate regression of the change in 3-month LIBOR basis in AE’s on the change in the forward premium delivers an R-squared of 77 percent, while the same regression on the change in interest differential yields an R-squared of 7 percent. - In EM’s, the volatility and co-movement patterns are more pronounced. - For EMs with deliverable offshore forwards (MXN, ZAR, ILS, TRY and RUB, panel (a)) and EMs with offshore NDFs (BRL, IDR, CNY, MYR, panel (b)), the decomposition shows: - Volatility in the CIP basis overwhelmingly driven by the forward premium. - The interest differential is relatively smooth. - Average correlation coefficients (monthly change) reported: - Correlation between change in the 3-month LIBOR basis and change in the forward premium: 0.96. - Correlation between change in the 3-month LIBOR basis and change in interest rate differential: 0.15. - The forward premium remains the dominant source of monthly variation in the 3-month CIP basis, but the interest differential becomes more correlated with the CIP basis change as the tenor increases to 6-months and 12-months. - The same correlation pattern holds for the CIP basis based on government bond yields. ### Stylized facts across AE and EM currency markets - Deviation from CIP across currencies is driven by time-variation in the forward premium (statistical decomposition). - Stylized asymmetry between AE and EM: - AE currencies typically see the CIP basis spike negatively in times of stress (Figure 3). - EM currencies spike into large positive values and are much more pronounced in absolute values (Figure 4). - In stress periods (such as the GFC), the forward premium for the dollar rises in excess of what is implied by the change in interest rate differential in EMs (positive spike), while it becomes large and negative in AEs. - Interpretation: excess demand for dollar forwards in EM’s and excess supply of dollar forwards (or a spot dollar shortage) in AE’s when dollar liquidity becomes scarce. - The paper provides a theoretical framework for this asymmetry and an empirical test using a purified measure of the CIP basis in emerging markets. ### Table 1: Change in Libor CIP basis and components — correlation coefficients - Table reports correlation coefficients between the monthly change of the Libor-based CIP basis and its components at the 3-month, 6-month, and 12-month maturities: the change in interest differential and the forward premium. - Panel A: AE (10 AE currencies) - Tenor headings: 3m 6m 12m - ∆CIP ∆(y_$ − y_i): 1 1 1 - ∆(y_$ − y_i): 0.096 0.117 0.179 1 - ∆ρ: 0.398 -0.875 1 0.353 -0.888 1 0.309 -0.880 1 - Panel B: EM (19 EM currencies) - Tenor headings: 3m 6m 12m - ∆CIP ∆(y_$ − y_i): 1 1 1 - ∆(y_$ − y_i): 0.014 0.011 0.235 1 - ∆ρ: 0.929 -0.357 1 0.886 -0.453 1 0.677 -0.556 1 - Notes: Panel A shows the correlation pattern for the 10 AE currencies and panel B the corresponding pattern for 19 EM currencies for which monthly Libor CIP at the various maturities can be constructed. ### Theoretical framework: constrained intermediation and CIP deviations - The model sketches a special case of Dao et al. (2025) where covered and uncovered interest parity deviations are jointly determined; here focus is on the balance sheet of a US dollar funded financial intermediary j with USD and local currency (LC) assets and liabilities. - Notation and structure: - USD positions (assets or liabilities) denoted with an asterisk. - At time t, intermediary has initial USD net worth W^*_j,t > 0. - Can issue D^*_j,t US dollar liabilities with gross borrowing rate R^*_t. - Can issue D_j,t local currency liabilities with gross borrowing rate R_t. - Holds B^*_j,t in USD with associated gross return ˜R^*_t, and B_j,t in local currency with gross return ˜R_t. - Nominal exchange rate E_t is the price of the USD in local currency (increase = depreciation of local currency). - Balance sheet in USD at time t: - W^*_j,t + D^*_j,t + D_j,t / E_t = B^*_j,t + B_j,t / E_t (equation (2)) - Off-balance-sheet forwards: - Sells F^*_j,t one-period ahead US dollars forward at price F_t in local currency. - Gross USD return on these forwards: (F_t / E_{t+1} − 1). - F^*_j,t > 0: intermediary short dollars forward (rest of market long dollars forward). - F^*_j,t < 0: intermediary long USD forward (rest of market short dollars forward). - Next period wealth expression (USD): - W^*_{j,t+1} = R^*_t W^*_{j,t} + ( ˜R^*_t / R^*_t − 1 ) R^*_t B^*_{j,t} + ( ˜R_t / R_t − 1 ) R^*_t B_{j,t} E_t + [ R^*_t / R_t F_t / E_t − 1 ] F^*_{j,t} + [ 1 − R^*_t E_{t+1} / (R_t E_t) ] 1 / E_{t+1} ( ˜R_t B_{j,t} − R_t D_{j,t} + F_t F^*_{j,t} ) (equation (3)) - Interpretation of terms: - First term: gross return on net worth at USD interbank rate R^*_t. - Second term: excess return on USD reserves (˜R^*_t / R^*_t − 1) — opportunity cost when ˜R^*_t ≤ R^*_t. - Third term: excess return on local currency assets vs liabilities (˜R_t / R_t − 1). - Fourth term: risk-free return on Covered Interest Rate strategy (the CIP basis): (R^*_t F_t / (R_t E_t) − 1). - Last term: return on net local currency exposure (˜R_t B_{j,t} − R_t D_{j,t} + F_t F^*_{j,t}), taking into account off-balance-sheet position F^*_{j,t}. - Constraint: intermediary cannot take open local currency positions, so - ˜R_t B_{j,t} + F_t F^*_{j,t} = R_t D_{j,t} (equation (4)) - Interpretation: intermediary’s local currency payable at t+1 equals local currency receivable at t+1. - With eq. (4) imposed, wealth simplifies to riskless form: - W^*_{j,t+1} = R^*_t W^*_{j,t} + ( ˜R^*_t / R^*_t − 1 ) R^*_t B^*_{j,t} + ( ˜R_t / R_t − 1 ) R^*_t B_{j,t} E_t + [ R^*_t / R_t F_t / E_t − 1 ] F^*_{j,t} (equation (5)) - Regulatory/internal liquidity constraint on off-balance-sheet positions: - B^*_{j,t} ≥ α_t | F^*_{j,t} | (equation (6)) - α_t captures intensity of the constraint: when α_t high, off-balance-sheet positions consume more reserves. - Tightness of the constraint depends on aggregate intermediary exposure: - α_t = a ( | ̄F^*_t | / ̄W^*_t )^α (equation (7)) - ̄F^*_t = ∑_j F^*_{j,t} (aggregate net forward dollar supply of all intermediaries). - ̄W^*_t = ∑_j W^*_{j,t} (aggregate balance sheet space allocated to the currency). - α ≥ 0 captures convexity of the constraint. - First-order conditions and equilibrium yield formula for CIP basis for intermediary j: - R^*_t / R_t F_t / E_t − 1 = μ_{j,t} a ( | ̄F^*_t | / ̄W^*_t )^α sign(F^*_{j,t}) (equation (8)) - μ_{j,t} ≥ 0 is Lagrange multiplier on constraint (6) — shadow cost of reserves. - Aggregation and common prices imply identical side of market and shadow cost across intermediaries: - R^*_t / R_t F_t / E_t − 1 = μ_t a ( | ̄F^*_t | / ̄W^*_t )^α sign( ̄F^*_t ) (equation (9)) - Key features of eq. (9): - If regulatory constraint does not bind (μ_t = 0 or a = 0), CIP holds and basis is zero: R^*_t / R_t F_t / E_t = 1. - Sign of basis equals sign of aggregate off-balance-sheet position ̄F^*_t. - ̄F^*_t > 0 (intermediaries sell dollars forward) ⇒ positive basis. - ̄F^*_t < 0 (intermediaries buy dollars forward) ⇒ negative basis. - Size of CIP basis related to magnitude | ̄F^*_t |, scaled by ̄W^*_t and funding tightness μ_t. - Tighter funding conditions (higher μ_t) or weaker intermediary capitalization (lower ̄W^*_t) amplify absolute basis for a given hedging demand. ### Cross-sectional and time-series implications; testable predictions - Extension to many currencies with balance sheet segmentation across currency desks: - For currency i, CIP basis satisfies: - R^*_t / R_{i,t} F_{i,t} / E_{i,t} − 1 = μ_t a ( | ̄F^*_{i,t} | / ̄W^*_{i,t} )^α sign( ̄F^*_{i,t} ) (equation (10)) - ̄F^*_{i,t} = aggregate net supply of dollar forwards from intermediaries against currency i. - Implications: - Cross-section: long-term differences in net demand for dollar forwards across countries should pin down the sign and average level of the CIP deviation. - Time-series: shifts in dollar funding costs and global intermediary balance sheet capacity should move the CIP basis jointly across currencies, with sign and magnitude depending on underlying forward demand position. - Explains asymmetry: AE currencies (net long dollar exposure, sign(̄F^*_{i,t}) < 0) spike to negative CIP basis in stress, while EMs (net short dollar exposure, sign(̄F^*_{i,t}) > 0) spike to positive CIP basis. - Explains negative correlation between CIP basis and local currency interest rates: countries with large net USD liability positions have higher interest rates and a positive basis; countries with net USD asset positions have lower domestic interest rates and a negative basis. - Empirical links to literature: - Liao and Zhang (2020) document correlation between CIP basis and net USD debt position in cross section of major AEs. - Avdjiev et al. (2019) document differing loadings on the dollar index across G-10 countries consistent with net USD positions and eq. (10). - No such correlation previously documented for EM currencies; framework offers unifying explanation and extends analysis to EM currency markets. *Source: wpiea2025057-print-pdf* ### 4.1 What does the CIP basis measure for EMs? ### 4.1 What does the CIP basis measure for EMs? ### Conceptual decomposition of the naive government-bond CIP basis - Define the ‘naive’ CIP constructed using local government yields: - φ_Gov_i,n,t = y_Gov_USD,n,t + ρ_i,n,t − y_Gov_i,n,t (equation (11) in source) - Rewriting adds and subtracts possibly unobserved risk-free rates y_rf_i,n,t and decomposes local-currency government yield spreads into: - credit spread I_Gov_i,n,t and convenience yield λ_Gov_i,n,t, so that the local currency spread y_Gov_i,n,t − y_rf_i,n,t = I_Gov_i,n,t − λ_Gov_i,n,t. - The decomposition in the source yields: - φ_Gov_i,n,t = (I_USD,n,t − λ_USD,n,t) − (I_Gov_i,n,t − λ_Gov_i,n,t) + τ_i,n,t = ˆλ_Gov_i,n,t − ˆI_Gov_i,n,t + τ_i,n,t, - where ˆλ_Gov_i,n,t = λ_Gov_i,n,t − λ_Gov_USD,n,t (relative convenience yield), - and ˆI_Gov_i,n,t = I_Gov_i,n,t − I_Gov_USD,n,t (relative credit risk). - Interpretations from the decomposition: - The observed government-bond CIP deviation decreases with relative credit spreads (ˆI_Gov_i,n,t), - increases with relative convenience yield (ˆλ_Gov_i,n,t), - and increases with the true n‑year risk-free CIP deviation τ_i,n,t. ### Sources of true CIP deviation and model linkage - The “true” risk-free CIP deviation τ_i,n,t could stem from a relative convenience yield on risk-free assets themselves (e.g., the ‘specialness’ of the dollar). - In the model notation of the source, an increase in the dollar convenience yield is captured by the parameter μ_t, which affects the CIP basis in all countries proportional to net forward demand (see equation (10) in the source). ### Different assumptions in the literature about components in eq. (11) - Du and Schreger (2016) assumption: - Main source of CIP deviations in EMs is local-currency credit risk; they assume I_Gov_USD,n,t = ˆλ_Gov_i,n,t = τ_i,n,t = 0, implying φ_Gov_i,n,t = −I_Gov_i,n,t. - Jiang et al. (2020) assumption: - The naive CIP basis reflects the relative convenience of US Treasuries, i.e. φ_Gov_i,n,t = ˆλ_Gov_i,n,t. - The source emphasizes that all three terms (relative credit risk, relative convenience yield, true CIP deviation) could potentially matter and that these structural sources can be reflected in either the interest rate differential and/or the forward premium or both. ### Empirical challenges using local-government bonds for EM CIP measures - For EMs, using government bonds is problematic because: - substantial and time-varying credit risk makes ˆI_Gov_i,n,t non-negligible; - sovereign CDS do not cover local-currency bonds (CDS typically apply to foreign-currency/foreign-jurisdiction issuance), so CDS do not fully hedge local-currency government credit risk; - capital controls, differential tax treatment, and financial market segmentation mean international and domestic investors/borrowers do not have equal access to the same currency and money markets and face differential tax burdens across onshore/offshore investments. - These factors complicate recovering τ_i,n,t from the observed φ_Gov_i,n,t. ### Proposed solution: use supranational bonds issued in EM currencies - Rationale: - Supranational bonds issued in EM local currencies offer exposure to EM currencies without exposure to EM sovereign credit risk and typically settle offshore (e.g., Euroclear), easing investor access. - Large supranational issuers typically enjoy AAA ratings, are considered of very low default risk, and often have zero risk weighting for Basel II and III capital requirements and HQLA eligibility. - Supranationals issue in multiple currencies (including EM currencies) and primary issuances are placed through global dealer banks that manage liquidity in secondary markets. - Advantages relative to local government bonds: - Supranational issuance in EM currency provides a cleaner proxy for local-currency risk-free interest rates and thus for τ_i,n,t by avoiding EM sovereign credit risk captured in ˆI_Gov_i,n,t. - They allow construction of CIP deviations adjusted for differential and time-varying liquidity premia. ### Literature precedent and implementation notes - The idea builds on Du and Schreger (2016), who constructed cross-currency bases using bonds issued by the EIB and KfW in Turkish Lira and Brazilian Real. - The source expands the construction to six issuers in eight EM currencies and six advanced economies and offers a framework to adjust for liquidity premia (Schwarz (2019) noted the role of liquidity premia for supranational yields). - Operational details relevant to constructing longer-tenor forwards: - As LIBOR rates are not available for tenors above 12 months, and there is no liquid FX forward market beyond 3 months, longer-term FX forwards are constructed by combining prices for nondeliverable forwards, interest rate swaps and cross-currency swaps (as discussed in Du et al. (2018)). ### Tenor choice for constructing the purified CIP basis - Empirical guidance for tenor n: - Choose the tenor for which supranational bonds are traded most frequently. - From the distribution of remaining time to maturity across issuer–currency pairs in the sample, 1 year is the modal tenor and is chosen as the baseline tenor to construct the purified CIP basis. *Source: wpiea2025057-print-pdf - 4.1 What does the CIP basis measure for EMs?* ### 4.3 Using Supranational bonds to construct ‘purified’ CIP deviations ### 4.3 Using Supranational bonds to construct ‘purified’ CIP deviations ### 4.3.1 Constructing the supra-specific basis - Observed yields: - ySupra i,j,t,t+n (bond issued by supranational j in currency i) - ySupra USD,j,t,t+n (bond issued by supranational j in USD) - Supra-specific CIP basis (excess return) defined as: - φSupra i,j,t,t+n ≡ ySupra USD,j,t,t+n + ρi,t,t+n − ySupra i,j,t,t+n (Eq. (12)) - Practical measurement: - Single-maturity yield differences can be imprecise because they ignore the spot-rate profile across coupon payments and implicitly assume a flat zero-coupon yield curve. - Use z-spread method (Du et al. (2018b)) to compute spread over the entire yield curve for remaining maturity; z-spread is the parallel shift in the dollar IRS curve that equalizes discounted cash flows to market price. - Z-spread in USD for issuer j with residual maturity n: - P$ j,t,t+n = Σ_{τ=1}^{n/q} c$ [1 + yj,IRS t,t+τ + s$ j,t,t+n]^{-τ} + [1 + yj,IRS t,t+τ + s$ j,t,t+n]^{-n} (Eq. (13)) - where q is annual coupon payment frequency, c$ the dollar coupon rate, P$ j,t,t+n the market price at time t. - Z-spread for same issuer in currency i (cash flows swapped into USD) uses dollar IRS curve shifted by s̄ i j,t,t+n: - P i j,t,t+n = Σ_{τ=1}^{n/q} c i [1 + yi,IRS t,t+τ + χi,xccy t,t+τ + s i j,t,t+n]^{-τ} + [1 + yi,IRS t,t+τ + χi,xccy t,t+τ + s i j,t,t+n]^{-n} (Eq. (14)) - yi,IRS t,t+τ and χi,xccy t,t+τ obtained by fitting a Nelson-Siegel model to estimate zero-coupon curves for IRS and xccy rates. - Practical notes: - Estimated zero curves used only to fill rates at coupon dates, not to impute bond prices. - Z-spread and supranational basis calculated only for issuer-dates with observed secondary market bond prices. - Match pairs of supranational bonds in dollar and currency i with residual maturities within five months to compute z-spreads and basis. - Under no-arbitrage, difference between dollar and swapped z-spreads reflects true CIP deviation for the issuer: - φ i,j,t,t+n = s$ j,t,t+n − s i j,t,t+n (Eq. (15)) ### 4.3.2 Extracting the currency-specific CIP deviation - Focus on one-year purified CIP (n = 1), drop t+n subscript. - Basis decomposition: - φ i,j,t = s$ j,t − s i j,t = τ i,t + ˆλSupra i,j,t (Eq. (16)) - where ˆI Supraj,i,t = 0 and ˆλSupra i,j,t = λSupra i,j,t − λSupra USD,j,t denotes relative convenience/liquidity yield for issuer j in local currency vs. US dollar. - Liquidity and convenience yields: - Supranational convenience yields in EMs likely very small relative to dollar/euro; issuance volumes and secondary liquidity tend to be low. - Supra convenience yield in EMs can be negative when liquidity is lower than for other local assets. - Example: KfW convenience premium in US averages around 7 bps over the last 10 years, with interquartile range at 2.5 to 11 bps. - Proxy convenience yield with linear function of bid-ask spread for issuer j: - λSupra i,j,t = λSupra i,j + α j × BidAsk i,j,t + ε i,j,t , where α j < 0. - λSupra i,j is issuer-market fixed effect (average convenience yield depending on issuance volume/frequency and market characteristics). - α j captures issuer-specific sensitivity to bid-ask spread; ε i,j,t captures idiosyncratic liquidity fluctuations. - Replace USD supra convenience yield with definition λSupra USD,j,t = y rf USD,t − ySupra USD,j,t and rearrange to: - φSupra i,j,t + y rf USD,t − ySupra USD,j,t = τ i,t + λSupra i,j + α j BidAsk i,j,t + ε i,j,t . (Eq. (17)) - Identification: - All variables except residual and τ i,t in Eq. (17) are measurable. - Estimate “pure CIP basis” τ i,t by extracting market-time fixed effects in regression equations using at least two different supra issuers in a given market over the sample period. - Identification requires j > 1 for each currency i and period t. - Advantages of “pure” CIP basis: - Conceptually free of relative credit risk premia, adjusted for differential liquidity premia. - Overcomes market segmentation and legal/fiscal differences affecting conventional measures. - Using bonds by a single SSA issuer across currencies overcomes heterogeneity in credit/counterparty risk; CIP deviation reflects violation of law of one price for that issuer. ### 5 Purified CIP deviations - Sanity checks in G-10 markets (section 5.1): - For reserve currencies (EUR, GBP), supranational CIP basis computed with supranationals should coincide with naive CIP basis because interbank and sovereign credit risks are negligible. - For EUR and GBP, expect ˆI i,j,t = 0 and ˆλ i,j,t ≈ 0; purified supra basis should closely match LIBOR and Treasury/government bond bases. - Purified supra basis computed using z-spread method on EIB and KfW bonds in EUR, GBP and USD; 1-year maturity series show supra CIP bases closely track Treasury basis and most of the time LIBOR basis, but supra series are much more volatile due to lower liquidity. - For currencies with lower supranational issuance (AUD, CHF), relative liquidity premium likely significantly negative (ˆλSupra i,j,t < 0); purified basis tracks conventional bases less well and displays volatility spikes, often lower than Treasury basis. - Purified CIP deviations in emerging markets (section 5.2): - Able to estimate pure CIP basis for eight EM currencies; six provide sufficient data for time-series analysis. - Six supranational issuers active in EM currencies over last two decades: ADB, EBRD, EIB, IBRD, IFC, KfW. IBRD, KfW and EIB are most active. - Data requirements: daily transaction price for same issuer in EM currency and USD with same adjacent residual maturity to compute z-spread limits usable bonds. - For three EM currencies with most liquid supranational trading—Brazilian Real, Turkish Lira, South African Rand—four to five supranational issuers traded since early/mid-2000s, enabling purified CIP deviations stretching back before the GFC. - Procedure: - Compute daily z-spreads for each supranational issuer with 1-year residual maturity traded. - Aggregate to monthly averages to smooth daily volatility. - Use at least two supranational issuers with same/adjoining residual maturity in the same month to extract common time fixed effect (τ i,t). - Net out differential liquidity of dollar bond vs. local currency bond using bid-ask spread. - Main observations: 1. Conventional government bond and LIBOR bases are often lower than the riskless supra CIP deviation (which is free of credit risk and adjusted for liquidity differentials), reflecting higher priced credit risk in EM sovereigns and banks and negative relative treasury convenience yields vis-à-vis the US. 2. The discrepancy between conventional and purified measures is less stark for South African Rand, though purified series is still more stable. 3. Government bond basis typically exhibits largest swings (higher and more volatile sovereign credit risk plus negative relative convenience yield of EM government bonds relative to US Treasuries). 4. LIBOR contains substantial risk premium compared with supranational measure; using LIBOR basis in EMs can be especially misleading. - Practical constraint: - To construct a time-series supra basis, need at least one transaction price in both EM currency and dollar on the same day for same issuer with adjacent residual maturities (e.g., 1-year), which limits usable supra bonds. *Source: wpiea2025057-print-pdf - 4.3 Using Supranational bonds to construct ‘purified’ CIP deviations* ### 2. As the pure CIP basis is purged from risk-and liquidity premia, it is more stable and likely ### 2. As the pure CIP basis is purged from risk-and liquidity premia, it is more stable and likely reflects slower-moving intermediation frictions. It is precisely the empirical measure of the intermediary pricing wedge in equation (8) which in turn is due to balance sheet constraints, FX market shallowness, as well as capital and liquidity regulations. As part of the balance sheet constraint, the riskless CIP basis also reflects the so-called dollar premium (see Obstfeld and Zhou (2022)) and can rise and fall with the “specialness” of the dollar, in turn reflecting the shadow cost of dollar funding of the intermediary. ### Purified CIP: stability, sign, and interpretation - Purified (supranational-based) CIP purges risk and convenience yields and therefore: - Is often by orders of magnitude smaller than conventional CIP measures (relative absolute values referenced). - Frequently flips sign relative to conventional CIP measures, revealing a hedging-channel consistent with the model-implied CIP equation. - Empirical interpretation: - The purified CIP is the empirical measure of the intermediary pricing wedge in equation (8), driven by balance sheet constraints, FX market shallowness, and capital and liquidity regulations. - The riskless CIP basis also reflects the “dollar premium” and the shadow cost of dollar funding of intermediaries. ### Key statistics: Table 5 — Mean CIP deviations (and standard deviations) at 1-year tenor (in bps) - Table 5 reports mean CIP deviations and standard deviations for eight EM currencies (1-year tenor, overlapping sample periods) using three measures: LIBOR CIP basis, Government bond CIP basis, Purified (Supranational) CIP basis. - Exact values (mean with standard deviation in parentheses) from Table 5: - BRL: LIBOR CIP basis -113.11 (72.12); Gov. bond CIP basis -196.80 (103.14); Purified CIP basis 9.39 (39.82) - CNY: LIBOR CIP basis -24.43 (144.18); Gov. bond CIP basis 36.02 (128.12); Purified CIP basis -30.92 (60.13) - IDR: LIBOR CIP basis -50.67 (136.46); Gov. bond CIP basis 63.73 (117.85); Purified CIP basis 35.10 (35.69) - INR: LIBOR CIP basis -51.63 (134.08); Gov. bond CIP basis -82.08 (116.79); Purified CIP basis 3.83 (34.86) - MXN: LIBOR CIP basis — (1-year LIBOR interest rates not available for MXN); Gov. bond CIP basis -39.07 (47.67); Purified CIP basis 8.63 (21.8) - RUB: LIBOR CIP basis -56.82 (43.05); Gov. bond CIP basis 21.23 (57.86); Purified CIP basis -39.08 (196.62) - TRY: LIBOR CIP basis -21.74 (232.70); Gov. bond CIP basis -3.51 (261.21); Purified CIP basis 152.53 (82.98) - ZAR: LIBOR CIP basis 0.42 (31.81); Gov. bond CIP basis -0.94 (54.07); Purified CIP basis -28.38 (36.17) - Cross-sectional pattern noted: - EMs with large net external liabilities tend to have a positive purified CIP basis (Türkiye, Mexico, Brazil). - EMs with net external assets on average have negative purified CIP basis (China, Russia). ### Cross-sectional variation and model link (equations and expectations) - Specialized model (α = 1) average expression (equation (18)): - CIP_i ≈ ̄μ a ( ̄F*_i / ̄W*_i ) - ̄μ (average shadow cost of reserves) is common across currencies — cross-sectional variation comes from ̄F*_i / ̄W*_i (average demand for hedging relative to intermediation capacity). - Sign implications: - Countries with net dollar liabilities ( ̄F*_i > 0 ) → predicted CIP_i > 0. - Countries with net dollar assets ( ̄F*_i < 0 ) → predicted CIP_i < 0. - Measurement of hedging demand: - Use Benetrix et al. (2019) dataset updated by Allen and Juvenal (2024) to construct a dollar gap: difference between external dollar debt asset and external dollar debt liabilities as a share of GDP. - Augmented dollar gap (USDGAP_aug) also subtracts external local currency debt liabilities to capture foreign investor local-currency debt hedging demand: - USDGAP_aug_t = Ext.DebtAssets_USD_t − Ext.DebtLiabilities_USD_t − Ext.DebtLiabilities_LC_t / GDP_t (equation (19)). ### Empirical correlations (figures and sample correlations) - Conventional 3-month LIBOR basis: - AE currencies: sample correlation with dollar gap = -0.59. - EM currencies: no correlation observed. - Conventional 1-year LIBOR basis: - AE currencies: sample correlation with dollar gap = -0.65. - EM currencies: weak positive correlation (no expected sign). - Purified (supranational) CIP basis for eight EMs: - Replacing treasury/government CIP with supranational CIP reveals a clear negative correlation with USD Gap: - 1-year Treasury CIP in EMs vs USD Gap: Corr = -0.07 (panel a of fig.13). - 1-year Supra CIP in EMs vs USD Gap: Corr = -0.78 (panel b of fig.13); reported sample correlation turns from zero to -0.8 in fig.13 note. - 3-year Treasury CIP vs USD Gap: Corr = -0.93? (panel c reports Corr = -0.93 for 3y Supra in panel (d) and additional notes say sample correlation turns from zero to -0.9 for 3-year supra). - 3y Supra CIP in EMs vs USD Gap: Corr = -0.93 (panel d) and stated Corr = -0.77 in panel (d) annotation (see figure captions and notes). - Excluding outliers (RUB and CNY) steepens slope; supranational bases show most volatility for RUB and CNY. - Augmented dollar gap effects (Figure 14): - Augmenting USD gap with external local-currency debt makes dollar gap substantially more negative for several EMs, notably ZAR, MXN and TRY. - 1y Supra CIP vs Augmented USD Gap: Corr = -0.68 (figure 14b); reported alternative Corr = -0.29 in figure 14b caption (both correlations provided in figure). ### Carry Trade, foreign investors, and interpretation of augmented dollar gap - Foreign investors funding local-currency assets in dollars can reverse sign of net forward demand over the dollar cycle. - Augmented dollar gap is constructed to include foreign investors’ external local currency debt liabilities. - Foreign investors’ positions can be volatile (“Original Sin Redux”), and demand for hedging increases in times of financial stress. - Cross-section vs time-series: - Cross-sectional correlation between foreign investors’ local currency assets and CIP basis is weak because positions and hedging demand shift over time. - Time-series analysis (within currencies) is used to capture how CIP fluctuation is shaped by the dollar cycle and demand for dollar funding/hedging. ### Time variation and baseline regression framework - The model-implied time-series expression (α = 1, intermediaries’ net worth proportional to GDP) (equations (20)–(23)): - CIP_it = a/κ μ_t ( ̄F*_i / Y*_i ) (equation (20) and approximation (21)) - Measurement equation for hedging demand: ̄F*_i / Y*_i = ρ_0 + ρ_1 (−USDGAP_i) (equation (22)), with expected ρ_1 > 0 and ρ_0 < 0. - Plugging in yields CIP_it ≈ aρ_0/κ μ_t + aρ_1/κ μ_t (−USDGAP_i). - Baseline first-difference regression (equation (24)): - ∆CIP_it = α_i + β_1 ∆dollar_t + β_2 ∆dollar_t × (−USDGAP_i) + γ′ X_it + ε_it - ∆dollar_t is log-change in the Broad Dollar Index (proxy for μ_t). - Expectation: β_2 > 0 (tighter global funding conditions make CIP more positive for countries with larger negative USD gaps), β_1 < 0. - Controls: currency fixed effects, lagged CIP to capture mean reversion, slow-moving change in dollar-gap (annual), and other robustness variables. - Estimation uses OLS fixed effects with Driscoll-Kraay standard errors. ### Regression evidence (Tables 6 and 7): conventional vs. supranational CIP - Table 6 (baseline regressions with LIBOR and Government Bond CIP basis; monthly frequency; sample Feb 2006–Dec 2020): - AE and EM samples, 3-month and 12-month tenors. - Selected coefficient patterns (exact estimates and standard errors reported in table): - 3-month AE LIBOR: CIP_t−1 = -0.196*** (0.034); ∆dollar_t = -1.702*** (0.451); ∆dollar_t ∗ (−USDGAP_i) = 0.007 (0.015). - 3-month EM LIBOR: CIP_t−1 = -0.200*** (0.040); ∆dollar_t = -3.593* (1.905); ∆dollar_t ∗ (−USDGAP_i) = 0.015 (0.010). - 12-month AE GOV: CIP_t−1 = -0.190*** (0.016); ∆dollar_t = 5.549*** (0.634); ∆dollar_t ∗ (−USDGAP_i) = -0.122* (0.071). - 12-month EM GOV: CIP_t−1 = -0.170*** (0.019); ∆dollar_t = 8.501*** (0.973); ∆dollar_t ∗ (−USDGAP_i) = -0.220** (0.105). - Main takeaway: using conventional CIP bases for EMs does not deliver the theoretically predicted signs for β_1 and β_2; in some specifications the signs are opposite. - Table 7 (baseline regression with Supranational CIP basis in EMs; six EM currencies with sufficient data): - Column highlights (exact estimates and standard errors reported): - Column (1), 1-year Supra (six EMs): CIP_t−1 = -0.137*** (0.020); ∆dollar_t = -3.654** (1.825); ∆dollar_t ∗ −(USDGAP_i) = 0.198** (0.087); −(USDGAP_i) = 0.260* (0.135). Observations = 801; Number of EM currencies = 6; Within R2 = 0.0849. - Column (2), top-3 liquid EM supras: CIP_t−1 = -0.121*** (0.021); ∆dollar_t = -5.806*** (1.807); ∆dollar_t ∗ −(USDGAP_i) = 0.261*** (0.075); −(USDGAP_i) = 0.223** (0.113). Observations = 493; Number of EM currencies = 3; Within R2 = 0.110. - Columns (3)-(6) repeat regressions for GOV and Libor CIP for the same subsample and do not detect the hedging-channel effect as clearly (interaction terms smaller or insignificant). - Main takeaway: replacing conventional CIP with supranational (purified) CIP in EMs yields: - ∆dollar_t coefficients that are negative (as theory predicts for the funding channel). - ∆dollar_t ∗ −(USDGAP_i) coefficients that are positive and statistically significant (β_2 > 0), consistent with the model. - Stronger and more precisely estimated interaction terms when focusing on the most liquid supranational markets (Brazil, Türkiye, South Africa). ### Empirical interpretation and implications (from the empirical results) - Purifying CIP (using supranational bonds) uncovers the hedging-channel predicted by theory: - The purified CIP basis co-varies with global dollar funding conditions and country-specific dollar gap exposures in the direction implied by the model. - Conventional CIP measures (LIBOR, government bond) are contaminated by credit and convenience yields and can mask or reverse the hedging-channel signals, particularly in EMs. - Augmented dollar gap matters: - Including foreign investors’ external local-currency debt (augmented USD gap) materially affects the measured dollar-gap and its correlation with CIP; the correlation is weaker than the domestic-only gap but remains negative in the cross-section (figure 14b Corr = -0.68 / alternative reported Corr = -0.29). - Time-series dynamics: - Assuming hedging demand is slow-moving, time-series variation in CIP is primarily driven by changes in intermediary dollar funding costs (μ_t) and capital. - The purified CIP responds to monthly appreciations of the broad dollar as predicted by the funding-channel model (regression coefficients negative for ∆dollar_t and positive for the interaction with −USDGAP_i). *Italic source attribution line: Extracted from content unit “wpiea2025057-print-pdf - 2. As the pure CIP basis is purged from risk-and liquidity premia, it is more stable and likely reflects slower-moving intermediation frictions.”* ### 1.3 percent (equal to its sample standard deviation) at different values of the negative dollar ### wpiea2025057-print-pdf - 1.3 percent (equal to its sample standard deviation) at different values of the negative dollar gap ### Impact of dollar gap on the purified CIP basis - For a country with zero dollar gap where net hedging demand versus the dollar is balanced, a typical dollar appreciation is estimated to lower the CIP basis by 8 basis points. - If net hedging demand for dollar forwards is positive, with dollar liabilities and local currency external debt at 30 percent of GDP, the CIP is estimated to increase by 7.5 bps instead. - (−USDGAP) ranged from -41 to 67 percent of GDP in EM during the sample period. - The median monthly change in purified CIP basis in the sample is 8 bps. - A 5 percent appreciation in the broad dollar predicts a 12.5 bps larger increase in the CIP basis for a 10 pct of GDP higher dollar gap. - The differential is a negative 7.3 bps when the dollar depreciates by the same amount. - The standard deviation in the dollar gap is 15 pct of GDP across all EM’s and 7 pct for the EM’s in the estimation sample. ### Visualization notes (Figure 15) - Panel (a): Predicted change in the purified CIP basis in response to 1.3 percent broad dollar appreciation, computed for varying degrees of hedging demand proxied with the negative dollar gap. - Panel (b): Marginal impact of an increase in 1 ppt in the (negative) dollar gap ratio (in percent of GDP) on the predicted change in the CIP basis, evaluated at different rates of prevailing broad dollar depreciation/appreciation rates. - Predicted values are based on estimates in column (1) of Table 7 with confidence intervals at 90 percent level. ### Measures of the global dollar cycle (robustness) - Alternative measures used as proxies for global dollar funding conditions: - VIX index - Global financial cycle factor from Miranda-Agrippino and Rey (2020) (GFCy, multiplied by -1 so an increase signifies tightening) - Treasury basis from Jiang et al. (2020) (xTreas) - Regression results (Table 8) summary: - ∆ log VIX: coefficient -0.380*** (col. 1); interaction ∆ log VIX * (−USDGAP_i) = 0.014** (col. 1). - ∆−GFCy: coefficient -25.56*** (col. 3); interaction ∆−GFCy * (−USDGAP_i) = 1.016*** (col. 3). - ∆x_Treas: coefficient 53.41*** (col. 5); interaction ∆x_Treas * (−USDGAP_i) = -2.284** (col. 5). - ∆dollar: coefficients -4.629*** (col. 1), -3.554** (col. 2), -4.656*** (col. 3); interactions ∆dollar * (−USDGAP_i) = 0.228*** (col. 1), 0.175*** (col. 2), 0.206*** (col. 3). - (−USDGAP_i) standalone coefficients: 0.148 (col.1), 0.193* (col.2), 0.147 (col.3), 0.177* (col.4), 0.146 (col.5), 0.203* (col.6). - Observations: 548, 493, 548, 493, 495, 493 across columns reported. - Number of currencies: 33. - Within R2 values reported: 0.101, 0.124, 0.109, 0.125, 0.095, 0.123. - Interpretation: Increases in global risk aversion (VIX), retreat in global financial cycle (GFCy), or increases in synthetic Dollar borrowing rates (more negative Treasury basis) have a negative impact on the purified CIP basis when the Dollar gap is zero or positive (dollar funding channel). The same increases trigger an increase in the purified CIP basis proportional to the underlying Dollar liability gap (hedging demand channel). - When jointly controlling for the broad dollar index and alternatives, the broad dollar remains the most statistically significant driver of the dollar cycle. ### Interaction with intermediary net worth (leverage-capital ratio) - Model expansion: - Intermediary leverage η_t defined as ratio of their debt to net worth. - Under assumption debt is proportional to output (D*_ijt + D_ijt)/E_it = κ_ij Y_it, the basis becomes CIP_it = a κ_i μ_t η_t ̄F*_it / Y_it (eq. (25)). - Expanded estimating equation (26) for changes: ∆CIP_it = α_i + β_1 ∆dollar_t + β_2 ∆dollar_t × (−USDGAP_i) + β_3 ∆η_t + β_4 ∆η_t × (−USDGAP_i) + β_5 ∆dollar_t ∆η_t + β_6 ∆dollar_t ∆η_t × (−USDGAP_i) + γ′ X_it + ε_it - Proxy for η_t: market capital ratio of primary dealer counterparties of the New York Federal Reserve from He et al. (2017). - Predicted sign patterns: - β_1, β_3, β_5 < 0 (dollar funding channel) - β_2, β_4, β_6 > 0 (hedging channel) - Empirical results (Table 9) summary (sample: three EM currencies with most liquid supranational bond markets: BRL, TRY, ZAR): - Column 1: ∆η_t = -4.018*** (0.1 significance); ∆η_t × (−USDGAP_i) = 0.172**. - Column 2 (full specification): y_{t-1} = -0.131***; ∆η_t = -2.092***; ∆η_t × (−USDGAP_i) = 0.114*; ∆dollar_t = -5.535**; ∆dollar_t × (−USDGAP_i) = 0.379; ∆dollar_t ∆η_t = -0.0157; ∆dollar_t ∆η_t × (−USDGAP_i) = -0.106***; (−USDGAP_i) = 0.109. - Observations: 548 (col.1), 493 (col.2), 493 (col.3), 487 (col.4). - Number of currencies: 3 across reported columns. - Within R2: 0.097, 0.154, 0.088, 0.082. - Economic magnitudes visualized (Figure 16): - A 1 standard deviation appreciation of the broad dollar equals 1.35 percent month on month. - With unchanged intermediary leverage-capital ratio, a broad dollar appreciation lowers the CIP basis for low dollar liability gap countries and increases it for high dollar gap countries. - If intermediary leverage-capital ratio increases by 1.5 standard deviations, the negative impact of the same dollar appreciation on the CIP basis is 3-4 times larger than with unchanged intermediary net worth. - For larger dollar liability gaps the differential becomes smaller and turns positive as the hedging channel starts to dominate. - Conventional CIP measures (government bond yields and LIBOR-based) do not yield predicted effects: coefficients often not statistically significant and triple interaction term may have opposite sign, likely reflecting credit risk influences not captured by the purified CIP basis. ### Conclusion and policy implications - A novel “purified” CIP basis is constructed for major Emerging Markets using supranational bonds to adjust for relative credit risk and convenience yields. - Key empirical findings: - Deviation from CIP in EMs has not widened post GFC when measured with the purified CIP basis; deviations in AEs have become more similar to those in EMs when EMs are computed properly. - Cross-section: long-term demand for dollar forwards from hedging is a strong predictor for average CIP deviation across EM currencies. - Time-series: shocks and policies affecting supply of dollar funding, global intermediary’s balance sheet capacity, and demand for dollar assets interact with country-specific dollar hedging needs to shape CIP deviations. - Policy relevance: - The purified CIP basis can gauge potential impacts of policies such as spot and forward FX intervention, capital controls, and macro-prudential policies. - Empirical analysis of central bank swap lines between the Federal Reserve and emerging market central banks could show how swap lines lower dollar funding and hedging costs and promote local currency investment; benefits in EMs could be larger due to shallower FX markets and large funding/hedging needs from local dollar borrowing and foreign local currency investments. - Deviations from CIP introduce a wedge between domestic borrowing costs and synthetic local currency borrowing rates through dollar financing; targeting this wedge can strengthen monetary transmission in emerging markets, complementing control of domestic policy rates. *Italic: Content summarized from the supplied PDF chapter/section.* ### References ### References ### Major topical clusters in the bibliography - Covered interest parity (CIP) deviations and related arbitrage limits - Augustin, Patrick, Mikhail Chernov, Lukas Schmid, and Dongho Song (2022) “The Term Structure of Covered Interest Rate Parity Violations,” HKIMR Research Paper WP, No. 01/2022. - Avdjiev, Stefan, Wenxin Du, Catherine Koch, and Hyun Song Shin (2019) “The Dollar, Bank Leverage, and Deviations from Covered Interest Parity,” American Economic Review: Insights, 1(2), 193–208. - Baba, Naohiko and Frank Packer (2009) “Interpreting deviations from covered interest parity during the financial market turmoil of 2007–08,” Journal of Banking & Finance, 33 (11), 1953–1962. - Du, Wenxin, Alexander Tepper, and Adrien Verdelhan (2018b) “Deviations from Covered Interest Rate Parity,” Journal of Finance, 73(3). - Moskowitz, Tobias J, Chase P Ross, Sharon Y Ross, and Kaushik Vasudevan (2024) “Quantities and Covered-Interest Parity,” Working Paper 4820243, SSRN. - Cerutti, Eugenio and Haonan Zhou (2024) “Uncovering CIP deviations in Emerging Markets: Distinctions, Determinants, and Disconnect,” IMF Economic Review, 72 (1), 196–252. - Zeev, Nadav Ben and Daniel Nathan (2024) “The widening of cross-currency basis: When increased FX swap demand meets limits of arbitrage,” Journal of International Economics, 152, 103984. - Global dollar role, dollar risk premia, and global dollar cycle - Benetrix, Agustin, Deepali Gautam, Luciana Juvenal, and Martin Schmitz (2019) “Cross-Border Currency Exposures,” IMF Working Papers, 2019/299. - Du, Wenxin and Jesse Schreger (2016) “Local Currency Sovereign Risk,” Journal of Finance, 71(3). - Du, Wenxin, Joanne Im, and Jesse Schreger (2018a) “The U.S. Treasury Premium,” Journal of International Economics, 112, 167–181. - Jiang, Zhengyang, Arvind Krishnamurthy, and Hanno Lustig (2020) “Foreign safe asset demand and the dollar exchange rate,” Journal of Finance, 76(3), 1049–89. - Krishnamurthy, Arvind and Hanno N Lustig (2019) “Mind the gap in sovereign debt markets: The U.S. treasury basis and the dollar risk factor,” in 2019 Jackson Hole Economic Symposium, Stanford University Graduate School of Business Research Paper(3443231). - Obstfeld, Maurice and Haonan Zhou (2022) “The Global Dollar Cycle,” Brookings Papers on Economic Activity(Fall). - Miranda-Agrippino, Silvia and Hélene Rey (2020) “US monetary policy and the global financial cycle,” The Review of Economic Studies, 87 (6), 2754–2776. - Du, Wenxin and Amy Huber (2024) “Dollar Asset Holdings and Hedging Around the Globe,” Working Paper No. 32453, National Bureau of Economic Research. - Cross-border banking, funding, and intermediary constraints - Bruno, Valentina and Hyun Song Shin (2015) “Cross-border banking and global liquidity,” The Review of Economic Studies, 82 (2), 535–564. - (2017) “Global Dollar Credit and Carry Trades: a Firm-level Analysis,” Review of Financial Studies, 30(3), 703–749. - Ivashina, Victoria, David Scharfstein, and Jeremy Stein (2015) “Dollar Funding and the Lending Behavior of Global Banks,” The Quarterly Journal of Economics, 130(3), 1241–1281. - He, Zhiguo, Bryan Kelly, and Asaf Manela (2017) “Intermediary asset pricing: New evidence from many asset classes,” Journal of Financial Economics, 126 (1), 1–35. - Benetrix et al. (2019) and Allen and Juvenal (2024) on External Dollar debt exposures and the Dollar Gap. - Asset pricing, convenience yields, and collateral considerations - Devereux, Michael B, Charles Engel, and Steve Pak Yeung Wu (2023) “Collateral advantage: Exchange rates, capital flows and global cycles,” Working Paper No. 31164, National Bureau of Economic Research. - Diamond, William and Peter Van Tassel (2021) “Risk-free rates and convenience yields around the world,” working paper, Jacobs Levy Equity Management Center for Quantitative Financial Research. - Garleanu, Nicolae and Lasse Heje Pedersen (2011) “Margin-based asset pricing and deviations from the law of one price,” The Review of Financial Studies, 24 (6), 1980–2022. - Schwarz, Krista (2019) “Mind the gap: Disentangling credit and liquidity in risk spreads,” Review of Finance, 23 (3), 557–597. - Policy tools, reserves, and capital controls - Bahaj, Saleem and Ricardo Reis (2022) “Central bank swap lines: Evidence on the effects of the lender of last resort,” The Review of Economic Studies, 89 (4), 1654–1693. - Bianchi, Javier and Guido Lorenzoni (2022) “The prudential use of capital controls and foreign currency reserves,” in Handbook of International Economics, 6, 237–289: Elsevier. - Basu, Suman S, Emine Boz, Gita Gopinath, Francisco Roch, and Filiz Unsal (2023) Integrated monetary and financial policies for small open economies: International Monetary Fund Working Paper 23/161. - Empirical and methodological contributions - Campbell, John Y, Karine Serfaty-De Medeiros, and Luis M Viceira (2010) “Global currency hedging,” The Journal of Finance, 65 (1), 87–121. - Brunnermeier, Markus K, Stefan Nagel, and Lasse H Pedersen (2008) “Carry trades and currency crashes,” NBER Macroeconomics Annual, 23 (1), 313–348. - McBrady, Matthew R, Sandra Mortal, and Michael J Schill (2010) “Do firms believe in interest rate parity?” Review of Finance, 14 (4), 695–726. - Garleanu & Pedersen (2011); He, Kelly & Manela (2017); and others provide intermediary-based pricing frameworks and empirical evidence across asset classes. ### Notable single-author or institution contributions cited - Gourinchas, Pierre Olivier (2022) “International Macroeconomics: From the Great Financial Crisis to COVID-19, and Beyond,” IMF Economic Review, 1–34. - Dao, Mai Chi, Pierre-Olivier Gourinchas, and Oleg Itskhoki (2025) “Breaking Parity: Equilibrium Exchange Rates and Currency Premia,” Working Paper forthcoming, IMF. - De Leo, Pierre, Lorena Keller, and Dongchen Zou (2024) “Speculation, Forward Exchange Demand, and CIP Deviations in Emerging Economies,” research papers, The Wharton School. --- ### Appendix A — Data Appendix (Table 10: Variables and data sources) - Variables and their data sources with descriptions: - Spot exchange rate — Bloomberg — Daily spot exchange rate - Forward exchange rate — Bloomberg — Daily forward point (3 month/ 6 month) - Interbank rate — Bloomberg — Daily interbank rate (3 month/ 6 month) - Cross-currency basis swap (XCCY) — Bloomberg — Daily cross-currency basis swap rate (1y, 2y, 3y, 5y, 7y, 10y) - Interest rate swap (IRS) — Bloomberg — Daily vanilla interest rate swap rate (1y, 2y, 3y, 5y, 7y, 10y) - Non-deliverable swap (NDS) — Bloomberg — Daily non-deliverable swap (1y, 2y, 3y, 5y, 7y, 10y): In replacement of XCCY and IRS for BRL, INR, CNY, IDR, KRW, PHP, COP, PEN, RUB - Government bond yields — Bloomberg — Daily government bond yields (1y, 2y, 3y, 5y, 7y, 10y) - SSA bonds — Bloomberg — Daily bond prices (bid, ask) and bond yields - Dollar Gap — Benetrix et al. (2019), Allen and Juvenal (2024) — External Dollar debt assets net of external Dollar debt liabilities, in percent of GDP, annual averages. - Augmented Dollar Gap (baseline) — Benetrix et al. (2019), Allen and Juvenal (2024) — External Dollar debt assets plus external local currency debt liabilities, net of external Dollar debt liabilities, in percent of GDP, annual averages. - Broad U.S. Dollar Index — Federal Reserve Board — Nominal trade-weighted Dollar exchange rate - Treasury basis — Bloomberg — computed as in Krishnamurthy and Lustig (2019), using zero-coupon 12-months government bond yields, forward and spot exchange rates - Primary dealer leverage ratio — He et al. (2017) — Reciprocal of primary dealers’ capital ratio (market net worth in pct of book debt and market net worth) ### Bloomberg tickers and description — structure and selected entries - Note: Bloomberg tickers for FX Spot, FX Forward, Basis Swap, IRS, NDS, CCS are with suffix CURNCY and abbreviated here. - Note: Bloomberg tickers for interbank rate and Government bond rate are with suffix INDEX and abbreviated here. - Example currency entries (currency / type / ticker / name) — exact tickers and names as provided: - AUD / Spot / AUD / USDAUD Spot Exchange Rate - Price of 1 USD in AUD - AUD / Forward / AUD3M, AUD6M, AUD12M / Australian Dollar Forward Points (3M, 6M, 12M) - AUD / Interbank Index / BBSW3M, BBSW6M, BBSW1Y / ASX Australian Bank Bill Short Term Rates Mid (3M, 6M, 12M) - AUD / Basis Swap (vs US LIBOR) / ADBS1 - ADBS10 / AUD-USD Basis Swap (90D Bank Bill vs 3M Libor) (1Y - 10Y) - AUD / Basis Swap (vs SOFR) / ADBSQQ1 - ADBSQQ10 / AUD-USD Basis Swap (BBSW vs SOFR) (1Y - 10Y) - AUD / Interest Rate Swap (3M) / ADSWAP1Q - ADSWAP10Q / AUD Quarterly (vs. 3M Bank Bills) (1Y - 10Y) - AUD / Interest Rate Swap (6M) / ADSWAP1 - ADSWAP10 / AUD Semi Annual (vs. 6M Bank Bills) (1Y - 10Y) - AUD / Government bond index / C1273M, C1276M, C1271Y - C12710Y / BFV AUD AUSTRALIA SOVEREIGN (3M, 6M, 1Y - 10Y) - Additional selected currencies and their tickers (exact entries preserved): - CAD / Spot / CAD / USDCAD Spot Exchange Rate - Price of 1 USD in CAD - CHF / Spot / CHF / USDCHF Spot Exchange Rate - Price of 1 USD in CHF - EUR / Spot / EUR / USDEUR Spot Exchange Rate - Price of 1 USD in EUR - GBP / Spot / GBP / USDGBP Spot Exchange Rate - Price of 1 USD in GBP - JPY / Spot / JPY / USDJPY Spot Exchange Rate - Price of 1 USD in JPY - NOK / Spot / NOK / USDNOK Spot Exchange Rate - Price of 1 USD in NOK - NZD / Spot / NZD / USDNZD Spot Exchange Rate - Price of 1 USD in NZD - SEK / Spot / SEK / USDSEK Spot Exchange Rate - Price of 1 USD in SEK - USD / Interbank Index / US0003M, US0006M, US0001Y / ICE LIBOR USD (3M, 6M, 12M) - USD / Interest Rate Swap (3M) / USSW1 - USSW10 / USD Semi Anl 30/360(vs3MLIBOR) (1Y - 10Y) - BRL / Spot / BRL / USDBRL Spot Exchange Rate - Price of 1 USD in BRL - CNY / Spot / CNY / USDCNY Spot Exchange Rate - Price of 1 USD in CNY - INR / Spot / INR / USDINR Spot Exchange Rate - Price of 1 USD in INR - KRW / Spot / KRW / USDKRW Spot Exchange Rate - Price of 1 USD in KRW - MXN / Spot / MXN / USDMXN Spot Exchange Rate - Price of 1 USD in MXN - ZAR / Spot / ZAR / USDZAR Spot Exchange Rate - Price of 1 USD in ZAR - Country- and instrument-specific notes included in the table: - For BRL, INR, CNY, IDR, KRW, PHP, COP, PEN, RUB: NDS is used in replacement of XCCY and IRS (see Non-deliverable swap entry). - Several basis swap series are reported both versus US LIBOR and versus SOFR (e.g., AUD: ADBS1 - ADBS10 and ADBSQQ1 - ADBSQQ10). - Government bond indices are listed with multiple tenors (3M, 6M, 1Y - 10Y) using BFV-coded tickers for many currencies. *Content extracted from the References and Appendix A (Data Appendix) of the source PDF.* ### Appendix B Additional Tables and Figures ### Appendix B Additional Tables and Figures ### Summary statistics for conventional CIP deviations and components (Table 12) - Panel A: Advanced Economies (monthly values aggregated daily averages; units in basis points annualized) - 3-month CIP deviation (Obs. 1,860) - Mean: -19.8 - LIBOR $−y Mean: -10.7 - LIBOR i Mean: -19.4 - Gov. bond basis $−y Mean: -10.5 - Gov. bond basis i Mean: -8.9 - Median: -17.1, 5.8, -13.7, 1.2, -22.7 - Std. Dev.: 26.7, 156.4, 35.2, 159.8, 164.4 - 12-month CIP deviation (Obs. 1,209) - Mean: -12.1, 19.4, -17.7, 11, -28.8 - Median: -12.4, 21.8, -12.5, 12.2, -34 - Std. Dev.: 36.9, 141.4, 29.3, 149.5, 155.2 - Panel B: Emerging Markets - 3-month CIP deviation (Obs. 3,133) - Mean: -30.1, -377.1, -30.9, -379.3, 348.3 - Median: -35.2, -346.6, -42.8, -335.2, 300.8 - Std. Dev.: 217.8, 339.8, 223, 335.9, 413.8 - 12-month CIP deviation (Obs. 2,493) - Mean: -153.8, -361.3, -98, -397.3, 299.3 - Median: -79.9, -300.8, -61.8, -334.2, 215.1 - Std. Dev.: 228.6, 344.2, 209.8, 349, 392.7 - Notes: - Monthly LIBOR and Government bond based CIP deviations at 3-month and 12-month maturities; corresponding interest rate/yield differentials and dollar forward premium ρ = f − s (log difference in forward and spot exchange rates). - Sample period is post-GFC for all currencies. Overall 10 AE currencies and 16 EM currencies included. Source: Bloomberg. ### Time-series figures: Government bond and LIBOR-based CIP deviations (Figures 17–20, 18–20 series) - Figure 17: Government bond-based CIP deviation in AEs shown over 01jan2000–01jan2022 for currencies including EUR, CHF, JPY, GBP, AUD, CAD. - Figure 18: 3-months LIBOR and Treasury-based CIP bases for EUR and JPY over 01jan2000–01jan2022 (panel a: EUR; panel b: JPY). - Figure 19: 1-year LIBOR and Treasury-based CIP bases for EUR and JPY over 01jan2000–01jan2022 (panel a: EUR; panel b: JPY). - Figure 20: 3-month LIBOR and Treasury-based CIP bases for MXN (panel a) and BRL (panel b) across their available sample windows (MXN: 01jan2000–01jan2022; BRL: 01jan2005–01jan2024 shown). ### Number of bonds per supranational agency by currency (Table 13) - AE’s (number of bonds with observed secondary market prices, January 2000 to September 2023) - USD: IBRD 1223, KFW 721, EIB 526, IFC 515, EBRD 236, ADB 344, IADB 278, AFDB 139 - AUD: 265, 755, 7, 562, 711, 647, 40 (table layout indicates these are counts per SSA but preserve values as listed) - CAD: 312, 930, 721, 3104 - CHF: 521, 762, 274, 3 - EUR: 181, 646, 315, 331, 123, 572, 8 - GBP: 849, 1117, 212, 931, 239 - JPY: 232, 8729, 689, 13713 - NZD: 1184, 121, 388, 331, 618 - EM’s - BRL: 197, 5461, 127, 1096, 15568 - CNY: 418, 598, 11339.10 - HKD: 469, 216, 451, 310, 1717 - HUF: 108, 284, 1032. - IDR: 3910, 18159, 25756 - INR: 227, 1920, 628, 4045 - MXN: 102, 1319, 6447, 312826 (values preserved as in table) - PLN: 3710361219.. - RUB: 58517, 514, 313315 - TRY: 965066, 142, 1471602473 - TWD: ..37..20.. - ZAR: 3137975, 951071482085 - Notes: Not all bonds can be used to compute the CIP basis due to data requirements explained in the text. Source: Bloomberg. ### Supra vs. conventional CIP bases and purifying effect (Table 14; Figures 21, 23) - Table 14: Means (and standard deviations) of the absolute value of CIP basis at 1-year tenor using conventional (LIBOR and Government Bond) vs. purified (Supranational bond) measures for overlapping sample periods (in bps). Summary for listed currencies: - BRL: Libor CIP 113.11 (72.12); Bond CIP 196.80 (103.14); Purified CIP 32.22 (25.07) - CNY: Libor CIP 105.72 (100.43); Bond CIP 104.11 (74.08); Purified CIP 43.98 (51.25) - IDR: Libor CIP 506.78 (136.46); Bond CIP 99.39 (89.50); Purified CIP 28.37 (22.07) - INR: Libor CIP 516.31 (134.08); Bond CIP 115.99 (82.86); Purified CIP 27.16 (22.04) - MXN: Libor CIP -49.55 (36.58); Bond CIP 17.25 (15.83) (note: 1-year LIBOR interest rates are not available for MXN) - RUB: Libor CIP 61.22 (36.47); Bond CIP 40.26 (41.39); Purified CIP 154.93 (126.40) - TRY: Libor CIP 137.60 (188.66); Bond CIP 168.86 (198.96); Purified CIP 61.69 (76.39) - ZAR: Libor CIP 25.53 (18.92); Bond CIP 40.90 (35.28); Purified CIP 33.67 (31.28) - Figures: - Figure 21: 1-year Supra and Treasury/LIBOR basis for CAD and JPY across shown date ranges. - Figure 23: Purified versus conventional CIP deviations (1-year tenor) shown for MXN, INR, IDR, CNY, RUB across their sample windows. ### Regression evidence on alternative measures of μ_t (Table 15) - Dependent variable: Δτ_t^τ_(t−1) - Key coefficients (standard errors in parentheses; significance indicated) - Δτ_t^τ_(t−1) coefficient across columns: -0.128*** (0.0197), -0.133*** (0.0209), -0.131*** (0.0183), -0.134*** (0.0202), -0.128*** (0.0210), -0.132*** (0.0215) - Δ log VIX_t: -0.173 (0.113); -0.0755 (0.0976) in alternative specifications - Δ log VIX_t * (−USDGAP_i): 0.00595 (0.00587); 0.000525 (0.00579) - Δ−GFCy_t: -19.28*** (7.354); -15.45** (6.983) - Δ−GFCy_t * (−USDGAP_i): 0.853** (0.384); 0.599 (0.394) - Δx_Treas_t: 44.26*** (15.80); 34.67*** (12.86) - Δx_Treas_t * (−USDGAP_i): -2.121** (1.037); -1.618 (1.018) - Δdollar_t: -3.245* (1.720); -1.759 (1.459); -2.759* (1.515) - Δdollar_t * (−USDGAP_i): 0.195** (0.0822); 0.119 (0.0750); 0.150** (0.0741) - (−USDGAP_i): 0.210* (0.124); 0.250* (0.131); 0.203* (0.110); 0.219* (0.121); 0.203 (0.131); 0.243* (0.127) - Observations and panel details: - Observations: 856, 801, 856, 801, 803, 801 (by column) - Number of currencies: 6 in all columns - Within R2: 0.0786, 0.0891, 0.0918, 0.0970, 0.0852, 0.0950 - Notes: - Driskoll-Kraay heteroskedasticity and autocorrelation robust standard errors in parentheses. ***, **, * denote p<0.01, p<0.05, p<0.1 respectively. - Regressions use Supra CIP basis at 1-year tenor as dependent variable for six EM currencies with available supranational bond prices; μ is measured by: VIX index (cols. 1–2), Global Financial Cycle (GFCy) dynamic factor (cols. 3–4), and Treasury basis x_Treas (cols. 5–6). GFCy multiplied by -1 so an increase signifies tightening of financial conditions. dollar stands for log Broad dollar Index. USDGAP_i denotes the augmented hedging demand proxy: dollar net external asset position plus external debt in local currency position (in percent country i GDP). Currency fixed effects included. ### Bid-ask spreads and cross-sectional relationships (Figures 22, 24) - Figure 22: Daily Bid-Ask spreads for supranational and government bonds in major EM’s (1-year tenor) displayed (visual time series not tabulated here). - Figure 24: Net Dollar debt asset position (USD Gap) and average 3-month and 1-year Treasury (Government Bond) CIP basis in AE’s and EM’s: 2010-2017 averages by country/currency. - Scatterplots presented for: - (a) 3m Treasury CIP basis in AE’s: horizontal axis 3m Gov Bond basis (bps) range shown -60 to 20; vertical axis Dollar Gap (pct of GDP) range -200 to 200 - (b) 3m Treasury CIP basis in EM’s: horizontal axis 3m Gov Bond basis (bps) range -200 to 100; vertical axis Dollar Gap (pct of GDP) range -100 to 30 - (c) 1y Treasury CIP basis in AE’s: horizontal axis 1y Gov Bond basis (bps) range -60 to 20; vertical axis Dollar Gap (pct of GDP) range -200 to 200 - (d) 1y Treasury CIP basis in EM’s: horizontal axis 1y Gov Bond basis (bps) range -250 to 0; vertical axis Dollar Gap (pct of GDP) range -100 to 30 *Source: wpiea2025057-print-pdf - Appendix B Additional Tables and Figures (IMF).* --- _Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025057-print-pdf.pdf_