## wpiea2023028-print-pdf - Introduction ## Source details **Canonical URL:** [wpiea2023028-print-pdf - Introduction](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023028-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023028-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023028-print-pdf.pdf.json) --- ### Introduction: scope and motivation - Covered Interest Parity (CIP) defined: forward premium of a currency equals its nominal interest-rate advantage absent counterparty default risk and financial frictions. - Prior facts: - CIP held closely for advanced economies (AEs) until the Global Financial Crisis (GFC); deviations have persisted since the GFC. - Emerging markets (EMs) received less attention despite EM international trading reaching scales comparable to AE currencies. - Paper objectives: - Systematically discuss measurement of short-term CIP deviations. - Study macro-financial determinants of CIP deviations in EMs. - Highlight differences between EMs and AEs. - Analyze differing sensitivities of offshore and onshore FX markets to global macro-financial determinants. ### Rationale: macro-financial importance of EM CIP deviations - Short-term CIP deviations matter for: - International mutual funds rolling short-term currency forwards to hedge local-currency fixed income and equity portfolios. - Costs of accumulating foreign reserves. - Firms hedging with FX forwards, particularly at short tenors. - Long-term CIP deviations as indicators of local-currency sovereign risk. - Presence of dollar basis implies synthetic domestic-currency borrowing/lending rates can differ from domestic central bank rates and depend on core-country monetary policies. ### Measurement approach and practical choices - Benchmark construction principles: - Focus on 1-month and 3-month CIP deviations. - Include both offshore and onshore forwards where segmented. - Dollar interest rate choices: - One version uses canonical Libor. - Preferred U.S. dollar interest rate: A2/P2 non-financial commercial paper interest rate. - Data and implementation: - Spot/forward exchange rates and money-market interest rates from Bloomberg and Refinitiv; include offshore (often NDF) and onshore forwards when indicated. - Dollar interest rate series = simple average of: - Federal Reserve A2/P2 commercial paper interest rate (FRED ticker RIFSPPNA2P2D90NB for 90-day issuances). - Bloomberg ticker DCPD090Y for 90-day issuances. - Use continuously compounded rates; account for day count conventions and maturity differences following Du, Tepper and Verdelhan (2018) and Cerutti, Obstfeld and Zhou (2021). - Sample: wide EM definition covering 20 non-G10 currencies; sample spans 2002 to 2021. - Currencies with significant offshore/onshore forward wedges identified: BRL, CNY, IDR, INR, MYR, PHP, THB, TWD. ### Measurement caveats: segmentation, default risk, CDS adjustments - Market segmentation and regulatory limits (examples): limits on onshore net open positions, prohibition of resident participation in offshore FX derivatives, documentation requirements for non-resident access — these create forward-rate gaps between offshore non-residents and onshore residents. - CDS adjustments: - Conceptual CDS-adjusted deviation adj^x_{t,t+n} defined but practical problems include illiquidity/non-availability of short-tenor CDS and common global factors in CDS spreads that could introduce spurious dynamics. - Paper uses unadjusted benchmark basis and treats default risk as a separate regression determinant. ### Theoretical formulation (measurement notation) - Key definitions and equations preserved: - i_{t,t+n}: annualized log return on local money-market deposit. - s_t: log spot exchange rate (local currency per USD). - f_{t,t+n}: log n-period forward exchange rate. - Hedged return: i_{t,t+n} − (f_{t,t+n} − s_t). - CIP condition: i^$_{t,t+n} = i_{t,t+n} − (f_{t,t+n} − s_t). (Equation (1)) - CIP deviation: x_{t,t+n} = i^$_{t,t+n} − [i_{t,t+n} − (f_{t,t+n} − s_t)]. (Equation (2)) - Negative x_{t,t+n} implies hedged EM investment return exceeds direct dollar funding cost. - Alternative borrower perspective: x_{t,t+n} = [i^$_{t,t+n} + (f_{t,t+n} − s_t)] − i_{t,t+n}. (Equation (3)) - CDS-adjusted deviation: adj^x_{t,t+n} = i^$_{t,t+n} − [(i_{t,t+n} − l_t) − (f_{t,t+n} − s_t)]. (Equation (4)), where l_t is CDS spread. ### Key empirical observations and contrasts with AEs - EM CIP deviations are: - More heterogeneous and segmented across currencies and over time than AEs. - More volatile than AE currency CIP deviations, with significant deviations from zero even before the GFC. - Cross-sectional correlations (preserve reported values): - AEs: correlation between AE CIP deviations and interest rates ≈ 0.72; correlation with net international investment position ≈ -0.6. - EMs: correlation with interest rate level ≈ 0.35; correlation with NIIP position ≈ 0.30. - Global intermediaries’ time-varying risk-bearing capacity strongly influences EM offshore CIP deviations. - For Emerging European currencies with relatively integrated FX markets: - CIP deviations spike during quarter-ends and year-ends after 2016 when FX intermediation declines due to regulatory window-dressing. - Average size of the jump is "more than 20 basis points" — nearly twice amount estimated for G10 currencies by Cerutti, Obstfeld and Zhou (2021). ### Evolution of CIP deviations (benchmarks and patterns) - Benchmark 3-month offshore CIP deviations: - Widened during the GFC; notable peaks at the 2013 Taper Tantrum and January 2020 COVID-19 onset. - EM offshore CIP deviations are larger and more volatile than G-10 counterparts. - Offshore deviations often switch from negative to positive during crises; post-GFC AE deviations remain persistently negative for most AEs except Australia and New Zealand. - Offshore vs onshore and NDF role: - Countries with NDF markets tend to have larger CIP deviations than countries with deliverable forwards. - Panel summaries: - Panel (a): Eight currencies show wide offshore-onshore differences; all but Thailand correspond to NDF jurisdictions. - Malaysian 2016 tightening reduced offshore trading by 48 percent. - Panel (b): CLP, COP, KRW, PEN have comparable offshore and onshore deviations due to resident participation enabling arbitrage. - Panel (c): Remaining 8 EMs (convertible currencies) show no significant offshore-onshore differences. - CIP deviations using A2/P2 commercial paper generally larger than IBOR-calculated ones due to embedded credit risk. ### Costly financial intermediation and period-end dynamics - Regulation-driven constraints on global banks reduce intermediation capacity near regulatory reporting dates (quarter-ends, year-ends). - Event-study findings: - After 2016 (GSIB capital surcharge effective Jan 1, 2016): - AE basis jump: downward by 13 basis points on average. - Eastern European currencies: average CIP deviation response of 23 basis points (panel (a)), nearly twice the amount for G10 currencies. - Conclusion: regulatory constraints on global FX dealers are especially binding for EMs; arbitraging capital retreats more from EM markets than AE counterparts. ### Theoretical underpinnings and hypotheses - Intermediary risk-bearing capacity linkage: - For AEs: CIP deviations enlarge when broad dollar appreciates or intermediary leverage tightens. - For EM net debtors (typical excluding reserves): tightening of dealer leverage can make offshore CIP deviations more positive (opposite sign to G-10). - Costly intermediation mechanism implications: - Net creditors → more negative CIP deviations in stress (dealer requires lower forward dollar price). - Net debtors → positive change in offshore CIP deviations when dealer intermediation deteriorates (forward dollar becomes more expensive than spot movement). - Market segmentation: - Perfect arbitrage → onshore and offshore respond similarly. - Segmented markets → offshore CIP deviations more sensitive to global risk factors; onshore deviations less sensitive or may respond negatively. - Testable hypotheses: - Hypothesis 1: Sensitivity of onshore CIP deviations to global factors is smaller than offshore counterparts. - Hypothesis 2: Sensitivity of offshore CIP deviations to global risk factors is stronger for segmented FX markets than for integrated ones. ### Regression evidence (methodology and baseline variables) - Panel regressions of 3-month CIP deviations (offshore and onshore) on global macro-financial factors using monthly first differences. - Data treatment: CIP and forward liquidity winsorized at 1% and 99% tails; currency fixed effects; two-way clustered standard errors by time and currency. - Time coverage: 2002–2021 and 2010–2021 subsamples; subgroup analyses for integrated (Group I) versus segmented (Group II) forward markets. - Forward market liquidity measure: (Fask − Fbid) / Fmid × 10000. - Aggregate EM FX dealer leverage ratio: ICRi,t = Market equityit / (Market equityit + Book debtit); dealer leverage = inverse; log used in regressions. - Baseline regressors: - ∆(rUS − r): change in USD A2/P2 commercial paper rate minus local nominal money market rate. - ∆ log dealer leverage. - ∆ fwd bid-ask. - In Panel (b): safe-haven common factor and safe-haven residual. ### Key regression results (selected highlights, preserve numeric coefficients and diagnostics) - Table 2 Panel (a) — baseline (selected columns): - Column (1): 02-21, ∆ offshore: - ∆(rUS − r): 0.231* (0.122) - ∆ log dealer leverage: 1.076*** (0.344) - ∆ fwd bid-ask: 0.872* (0.438) - Observations: 4,128; R-squared: 0.069 - Column (2): 10-21, ∆ offshore: - ∆(rUS − r): 0.164** (0.072) - ∆ log dealer leverage: 0.912*** (0.261) - ∆ fwd bid-ask: 0.828** (0.321) - Observations: 2,706; R-squared: 0.042 - Column (5): Group II segmented 10-21, ∆ offshore: - ∆(rUS − r): 0.350*** (0.095) - ∆ log dealer leverage: 1.348** (0.428) - ∆ fwd bid-ask: 1.631* (0.779) - Observations: 1,069; R-squared: 0.061 - Column (6): Group II segmented 10-21, ∆ onshore: - ∆(rUS − r): 0.212** (0.078) - ∆ log dealer leverage: 0.379 (0.292) - ∆ fwd bid-ask: 0.185 (0.285) - Observations: 1,087; R-squared: 0.057 - Table 2 Panel (b) — add safe-haven dollar factor (selected columns): - Column (1): 02-21, ∆ offshore: - safe haven common factor: 71.086** (28.614) - safe haven residual: 8.291*** (2.819) - Observations: 4,128; R-squared: 0.093 - Column (5): Group II segmented 10-21, ∆ offshore: - safe haven common factor: 78.277* (35.639) - safe haven residual: 21.325** (6.199) - Observations: 1,069; R-squared: 0.121 ### Substantive regression findings (numeric magnitudes preserved) - ∆(rUS − r) positive association with basis across samples. - ∆ fwd bid-ask positive association with CIP deviations. - ∆ log dealer leverage positive correlation with offshore CIP deviations: - Entire sample: one percentage point increase in leverage ratio → 1.1 basis points increase in offshore basis (interpretation preserved from text). - Post-crisis sample: 0.9 basis points increase. - Group II segmented: one percentage point increase → 1.35 basis point increase in offshore basis; onshore: 0.38 basis point (statistically insignificant). - Group I integrated: one percentage point increase → 0.61 basis point increase offshore (significant at 5%). - Safe-haven factors: - Safe-haven common factor and residual positive and significant correlates of offshore CIP deviations, especially for segmented currencies (e.g., common factor 78.277*, residual 21.325** in Column (5)). - Onshore basis responses to safe-haven factors typically statistically insignificant. - Overall: offshore CIP deviations respond more strongly and positively to intermediary deleveraging and safe-haven appreciation; onshore bases show zero-to-negative or insignificant responses. - Robustness: results robust to winsorization, exclusion of post-2020 COVID-19 period, 1-month tenor and IBOR-based measures (referenced Appendix Tables A5, A6). ### CIP deviations and country-specific correlates (country channels) - Country-specific factors examined: - Sovereign default risk: 5-year USD sovereign CDS spreads residualized by first principal component. - FX intervention (FXI): monthly broad measure as percent of GDP; lagged by one month. - Main findings — sovereign default risk (Panel (a)): - Example (Panel (a), column (2), 10-21): - safe haven common factor: 37.994* (20.012) - safe haven residual: 9.772*** (3.313) - ∆ 5y residualized cds spread: 0.482** (0.221) - ∆ log dealer leverage: 0.593** (0.245) - ∆ fwd bid-ask: 0.587** (0.254) - Observations: 2,439; R-squared: 0.076 - General evidence weak; 5-year CDS significance does not persist across currency groups. - Main findings — lagged FX intervention (Panel (b)): - Example (Panel (b), column (2), 10-21): - safe haven common factor: 42.667* (20.396) - safe haven residual: 10.945*** (3.336) - FXI: 3.426 (2.070) — positive but insignificant - ∆ log dealer leverage: 0.588** (0.234) - ∆ fwd bid-ask: 0.704** (0.278) - Observations: 2,706; R-squared: 0.071 - For Group II segmented currencies: - A one percentage point increase in foreign-currency asset purchases (% of GDP) associated with a 3.6 basis point decline in onshore CIP deviations (text statement). - Offshore CIP deviations correlate with FX intervention in opposite sign (insignificant at 10%), suggesting barriers to transmission. - Panel (b), column (6) Group II onshore: FXI coefficient -3.642* (1.547), Observations 1,087, R-squared 0.063. ### Policy implications and recommendations (preserved content) - Two policy instrument types for "basis control": - Pre-emptive capital flow management and macroprudential measures (including participation constraints in FX forward markets) to reduce onshore sensitivity to global risk and dampen external finance premia. - Use segmentation as policy space: central banks can intervene in forward markets (e.g., buy U.S. dollar forward at a high price) to provide downside protection to domestic dollar borrowers and cap CIP deviations by offering cheaper hedges. - Tradeoffs and risks: - Implicit taxes on hedging can discourage foreign participation in local-currency markets. - Tightening global intermediaries' risk-absorbing capacity can amplify offshore CIP sensitivity to global factors. - Segmentation limits onshore interventions’ ability to stabilize offshore expectations; high NDF hedging costs can trigger international investor liquidation of local bond holdings. - Operational considerations: - NDF operations may be cost-effective and pose little threat to FX reserve stability because ex ante cost equals the gap between forward rate and expected exchange rate; stabilizing expectations could make operations profitable ex ante and ex post. - Recommended research priority: evaluate macroeconomic impacts and welfare benefits of capital flow measures and FX forward market intervention. ### Appendix, data, and supplemental evidence (selected statistics and robustness) - Currency coverage examples: BRL, CNY, IDR, INR, MYR, PHP, THB, TWD, CLP, COP, CZK, HUF, ILS, KRW, MXN, PEN, PLN, RUB, TRY, ZAR. - FX dealer leverage measure constructed from large FX dealer banks (list provided in Appendix). - Summary statistics (preserve reported values): - log FX dealer leverage ratio: Obs 240; Mean 3.097; Std. Dev. .389; Min 2.332; Max 4.521; P50 3.08. - 3-month offshore CIP deviations (bps): Obs 4368; Mean -28.144; Std. Dev. 130.112; Min -537.634; Max 646.177; P50 -20.942. - 3-month onshore CIP deviations (bps): Obs 2056; Mean -39.599; Std. Dev. 93.264; Min -553.043; Max 280.301; P50 -22.853. - rUS − r (%): Obs 4500; Mean -3.5964.599; Min -47.3794.505; P50 -2.785. - offshore fwd bid-ask: Obs 4546; Mean 25.864; Std. Dev. 18.715; Min 5.165156.905; P50 21.023. - 5y residualized cds spread (bps): Obs 4158; Mean -8.228; Std. Dev. 135.216; Min -804.393; Max 2642.105; P50 -8.558. - FXI (% GDP): Obs 4798; Mean .135; Std. Dev. .853; Min -7.8910.82; P50 .05. - Appendix robustness tables show: - G-10 regressions: tightening dealer leverage associated with negative CIP deviation responses (contrasting EM positive responses). - 1-month CIP regressions and IBOR-based CIP regressions broadly corroborate main findings, with larger sensitivities for segmented currencies (selected coefficients reproduced in appendix tables A5, A6). *Source: wpiea2023028-print-pdf - Introduction (PDF chapter/section).* ### Introduction ........................................................................................................... ### wpiea2023028-print-pdf - Introduction ........................................................................................................... ### FX Market Development and CIP Deviation Measurement - Located on page 9. ### Time-Series and Cross-Sectional Stylized Facts - Located on page 13. ### Evolution of CIP Deviations - Located on page 14. ### Cross-Sectional Correlations with Macro-Financial Variables - Located on page 16. ### Costly Financial Intermediation and Period End Dynamics - Located on page 18. ### Global Factors and CIP Deviations: On-Shore and Off-Shore Disconnect - Located on page 20. ### Theoretical Underpinnings and Hypotheses: Basis Sensitivity, Costly Financial Intermediation, and Segmented Markets - Located on page 21. ### Regression Evidence - Located on page 24. ### CIP Deviations and Country Specific Correlates - Located on page 30. ### Discussion and Policy Implications - Located on page 33. *Source: wpiea2023028-print-pdf - Introduction (PDF chapter/section).* ### References ............................................................................................................. ### References ......................................................................................................................................................... 36 ### Introduction: scope and motivation - Covered Interest Parity (CIP) is defined as the relationship equating the forward premium of a currency to its nominal interest-rate advantage absent counterparty default risk and financial frictions. - CIP held closely for advanced economies (AEs) until the Global Financial Crisis (GFC); deviations have persisted since the GFC. - Emerging markets (EMs) have received less attention on CIP deviations despite EM international trading reaching scales comparable to AE currencies. - The paper’s objectives: - Systematically discuss measurement of short-term CIP deviations. - Study macro-financial determinants of CIP deviations in EMs. - Highlight differences between EMs and AEs. - Present a novel analysis of differing sensitivities of offshore and onshore FX markets to global macro-financial determinants. ### Rationale: macro-financial importance of EM CIP deviations - Short-term CIP deviations matter for: - International mutual funds rolling short-term currency forwards to hedge local-currency fixed income and equity portfolios. - Costs of accumulating foreign reserves (Amador, Bianchi, Bocola and Perri, 2019). - Firms hedging with FX forwards, particularly at short tenors (Alfaro, Calani and Varela, 2021). - Long-term CIP deviations indicating local-currency sovereign risk (Du and Schreger, 2016). - Presence of dollar basis implies synthetic domestic-currency borrowing/lending rates can differ from domestic central bank rates and depend on core-country monetary policies (Cerutti, Obstfeld and Zhou, 2021). ### Key empirical observations and contrasts with AEs - EM CIP deviations are: - More heterogeneous and segmented across currencies and over time than AEs. - More volatile than AE currency CIP deviations, with significant deviations from zero even before the GFC. - Cross-sectional correlations between EM CIP deviations and interest rate differential or net international investment position have opposite signs to those in AEs. - Global financial intermediaries’ time-varying risk-bearing capacity strongly influences EM offshore CIP deviations. - For Emerging European currencies with relatively integrated FX markets: - CIP deviations spike during quarter-ends and year-ends after 2016 when FX intermediation declines due to regulatory window-dressing by European banks and G-SIBs. - The average size of the jump is "more than 20 basis points"—nearly twice the amount estimated for G10 currencies by Cerutti, Obstfeld and Zhou (2021). ### Theoretical interpretation and testable predictions - Costly financial intermediation theory predicts CIP deviations respond differently for net debtor versus net creditor economies: - Most major EMs are net international debtors (excluding reserves); CIP deviations should respond opposite to net creditors. - Constructed measure of major foreign exchange dealer banks’ leverage ratio (in the spirit of He, Kelly and Manela (2017)) shows: - As the leverage ratio increases (dealers’ average capital ratio declines), offshore CIP deviations become more positive, indicating a relative decrease of implied dollar interest rate in the swap market—contrasting with G-10 currency patterns. - Market segmentation implications: - Currencies with significant offshore/onshore segmentation show offshore CIP deviations that are sensitive to global risk factors, while onshore CIP deviations do not comove with global risk factors. - Segmentation insulates onshore markets but can increase offshore FX market sensitivity due to worsening efficient risk-sharing and limits to arbitrage. ### Policy implications for EM authorities - Disconnect between onshore and offshore CIP deviations matters for policy design: - Capital flow management used preemptively can lower foreign-currency FX debt, reduce sensitivity to global risk factors, and reduce risks of sudden stops and crises. - FX market segmentation can provide monetary authorities wider policy space to intervene in onshore markets to keep funding cost low and provide hedges to foreign-currency borrowers. - Segmentation imposes a “tax” on hedging that can discourage foreign participation in local bond markets if sufficiently large (example noted: Malaysia’s 2016 tightening of resident participation in offshore FX markets). - Onshore interventions may have limited ability to alleviate offshore adverse sentiments and capital outflow pressure because limited global dealer risk-bearing capacity can amplify offshore CIP volatility and hedging costs for global investors. ### Contributions to the literature - Threefold contributions: 1. Extends analysis of evolution and macro-financial determinants of short-term CIP deviations to EMs, documenting heterogeneity and volatility relative to AEs. 2. Complements UIP analysis in EMs and AEs—highlights forward FX markets’ role in differences across EMs and AEs (ties to Kalemli-Özcan and Varela (2021)). 3. Contributes to onshore/offshore FX market literature (Patel and Xia (2019); Schmittmann and Chua (2020)) by documenting how macro-financial factors affect CIP deviations computed from onshore and offshore forwards. ### Measurement approach and practical choices - Benchmark CIP deviations construction principles: - Short tenors: focus on 1-month and 3-month CIP deviations. - Include both offshore and onshore forward exchange rates where markets are segmented. - Dollar interest rate choice partially reflecting EM default risk: - One version uses canonical Libor. - Preferred U.S. dollar interest rate is the A2/P2 non-financial commercial paper interest rate (reflecting credit risk of EM dollar borrowers with fundamentals close to A2/P2). - Data sources and implementation details: - Spot and forward exchange rates, and money-market interest rates from Bloomberg and Refinitiv; both offshore (often non-deliverable) and onshore forward rates included when explicitly indicated. - Benchmark short-term interbank domestic interest rates chosen based on availability. - Dollar interest rate series is a simple average of: - Federal Reserve A2/P2 commercial paper interest rate (FRED ticker RIFSPPNA2P2D90NB for 90-day issuances). - Bloomberg ticker DCPD090Y for 90-day issuances. - Continuously compounded interest rates used; account for day count conventions and maturity date differences following Du, Tepper and Verdelhan (2018) and Cerutti, Obstfeld and Zhou (2021). - Spot/forward series selected for best coverage; a wide EM definition covering 20 non-G10 currencies; sample spans 2002 to 2021 (currency data availability varies). - For currencies with significant offshore/onshore forward wedges the currencies identified include: BRL, CNY, IDR, INR, MYR, PHP, THB, and TWD. ### Measurement caveats: segmentation, default risk, and CDS adjustments - Market segmentation and regulatory limits to arbitrage: - Examples: limits on onshore net open positions of forwards/swaps; prohibition of resident participation in offshore FX derivatives; documentation requirements for non-resident access. - These create gaps between forward rates faced by offshore non-residents and onshore resident borrowers. - Default risk and CDS issues: - Default risk undermines riskless arbitrage; investors demand compensation for loss-upon-default and covariance with pricing kernel. - CDS spreads could be used to adjust CIP deviations (adj formula (4)), but practical problems arise: - Short-tenor CDS contracts (e.g., three-month) are often not available, illiquid, or unobserved. - Sovereign CDS share common factors tied to global risk aversion; adjusting with CDS may mechanically reduce offshore basis sensitivity to global factors and increase onshore sensitivity, potentially introducing spurious dynamics. - The paper thus uses unadjusted benchmark basis and treats country default risk as a separate potential determinant in regressions. ### Theoretical formulation (measurement notation) - Definitions and key equations: - Annualized log return on local money-market deposit: i_{t,t+n}. - Log spot exchange rate: s_t (local currency per USD). - Log n-period forward exchange rate: f_{t,t+n}. - Hedged return using forward: i_{t,t+n} − (f_{t,t+n} − s_t). - CIP condition (ideal, riskless): i^$_{t,t+n} = i_{t,t+n} − (f_{t,t+n} − s_t). (Equation (1)) - CIP deviation convention x_{t,t+n} = i^$_{t,t+n} − [i_{t,t+n} − (f_{t,t+n} − s_t)]. (Equation (2)) - Negative x_{t,t+n} implies hedged EM investment return exceeds direct dollar funding cost. - Alternative borrower perspective: x_{t,t+n} = [i^$_{t,t+n} + (f_{t,t+n} − s_t)] − i_{t,t+n}. (Equation (3)) - CDS-adjusted deviation (conceptually): adj^x_{t,t+n} = i^$_{t,t+n} − [(i_{t,t+n} − l_t) − (f_{t,t+n} − s_t)]. (Equation (4)), where l_t is CDS spread. ### Empirical focus and paper structure (forward look) - Initial analysis centers on offshore EM CIP deviations for comparability with AE literature; subsequently covers offshore vs onshore differences. - The paper next presents measurement of CIP deviations, evaluates evolution across time and currencies, estimates influence of global and country-specific factors on EM CIP deviations, and concludes with policy implications. *IMF Working Paper Uncovering CIP Deviations in Emerging Markets — INTERNATIONAL MONETARY FUND* ### 3.1 Evolution of CIP deviations ### 3.1 Evolution of CIP deviations ### Benchmark offshore 3-month CIP deviations - The benchmark 3-month offshore CIP deviations are presented across EMs, split into two groups: (i) countries with NDF in the first row of charts, and (ii) countries with deliverable FX forwards in the second row of charts. The right-hand chart in each row contains countries with larger CIP deviations. - General patterns and event peaks: - CIP deviations widened during the GFC and fluctuate after the crisis with recognizable common peaks such as the Taper Tantrum in 2013 and the beginning of the COVID-19 crisis in January 2020. - Offshore CIP deviations of EM currencies are considerably larger and more volatile than their G-10 currency counterparts. - Offshore CIP deviations often switch from negative to positive territories during crisis periods in many EMs; this switching is less prevalent in AEs, where post-GFC CIP deviations are persistently negative for most AEs except Australia and New Zealand. - Policy and market structure factors: - Pervasive use of capital flow management measures in EMs may affect the relative scarcity of US dollars to different investors (example reference: Keller (2021) analysis of Peruvian banks). - The growing appetite of global investors for EM assets has driven offshore FX trading, led by robust growth in NDF trading (Patel and Xia, 2019). ### Offshore vs onshore differences and role of NDFs - Countries with NDF markets tend to have larger CIP deviations than countries with deliverable forwards. - FX market segmentation and capital restrictions on foreign participation in domestic FX markets influence the existence and size of offshore NDF markets (McCauley and Shu, 2016). - Table 1 highlights: - Panel (a): Eight currencies show wide differences between offshore and onshore CIP calculations; all but Thailand correspond to jurisdictions with NDF. Thailand uses an offshore deliverable forward market and imposes limitations on non-residents engaging with onshore financial institutions (e.g., requirement of providing proof of underlying for each transaction; end of the day outstanding position limit for non-residents). Some of these limitations on non-residents were relaxed in January 2021 according to the 2022 AREAER. - Malaysian authorities tightened restrictions in 2016 by effectively banning offshore trading of NDFs by domestic entities, resulting in a fall of 48 percent in offshore trading (see Patel and Xia (2019); Schmittmann and Chua (2020)). - Panel (b): Some EMs with NDFs (CLP, COP, KRW, PEN) have comparable offshore and onshore CIP deviations; Korea’s large NDF market and ability of Korean residents to participate in NDF markets enable arbitrage that ensures close integration between offshore and onshore FX forward markets. - Panel (c): Remaining 8 EMs are mostly convertible currencies (including through deliverable forwards), with no significant differences between offshore and onshore calculations. For these, the difference between A2/P2 commercial rate paper and the US dollar Libor rate (IBOR) is presented; CIP deviations using commercial rate paper are generally larger than IBOR-calculated ones due to embedded credit risk in the dollar CP rate. ### Cross-sectional correlations with macro-financial variables (summary from Section 3.2) - Advanced economies (AEs): - Strong positive correlation of about 0.72 between AE CIP deviations and the level of interest rates in the cross-section. - Strong negative correlation of about -0.6 between AE CIP deviations and net international investment position (excluding reserves). - Examples: Australia and New Zealand have high-rate currencies and positive CIP deviations, indicating direct USD borrowing is more expensive than synthetic dollar interest in the FX forward market. Net creditor countries (e.g., Japan and Norway) tend to have more negative cross-currency basis. - Emerging markets (EMs): - Correlations are smaller and signs differ from AEs: about 0.35 for interest rate level and about 0.30 for NIIP position. - Heterogeneity across EMs is large, partly due to frequent segmentation between offshore and onshore FX markets and the sample’s composition (only net debtor countries and countries with positive interest spreads over US rates). - Note on magnitudes and measures: - Correlation values preserved as reported: 0.72, -0.6, 0.35, 0.30. - CIP deviations are defined using USD A2/P2 commercial paper rate as the dollar interest rate so that a negative CIP deviation corresponds to a lower direct dollar interest rate relative to the synthetic dollar interest rate. ### Costly financial intermediation and period-end dynamics (summary from Section 3.3) - Regulation-driven financial constraints on global banks link to CIP deviations, especially near regulatory reporting dates when window-dressing reduces intermediation capacity. - Quarter-end and year-end dynamics: - Figure 3 plots evolution of benchmark 1-month CIP deviations at quarter ends; impacts are larger during the last quarter of the year but present at other quarter-ends as well. - Eastern European EM currencies (CZK, HUF, PLN, RUB) show notable quarter-end movements due to the role of European banks and their universal business model. - The GSIB capital surcharge introduced on January 1, 2016, has a strong effect in driving three-month benchmark CIP deviations in the fourth quarter when U.S. and euro area regulators evaluate GSIB balance sheets. - Event-study findings: - Regressions of daily 3-month offshore/NDF CIP deviations on dummies for dates before and after the day when 3-month forward contracts begin to settle (usually priced at end-September): - Before GSIB regulation, 3-month bases for EME European currencies, AE currencies, and other deliverable EM currencies experienced little to no action during these dates. - For non-Eastern European EM NDFs, estimated coefficients before year-end dates are large yet statistically insignificant (panel (c)). - After 2016, observed effects include: - A downward jump of the AE basis by 13 basis points on average (panel (b)). - An average response of CIP deviations of 23 bps for Eastern European currencies (panel (a)), nearly twice the amount observed for G10 currencies. - Little to no action observed for other EM currencies. - Conclusion: Regulatory constraints on global FX dealers are especially binding for emerging markets, with arbitraging capital retreating from EM markets by more than from AE counterparts. *IMF Working Paper Uncovering CIP Deviations in Emerging Markets* ### 4.1 Theoretical underpinnings and hypotheses: Basis sensitivity, costly ### 4.1 Theoretical underpinnings and hypotheses: Basis sensitivity, costly financial intermediation, and segmented markets ### Relationship between CIP deviations and intermediary risk-bearing capacity - Empirical literature finds a close relationship between CIP deviations and the risk-bearing capacity of intermediaries (Avdjiev, Du, Koch and Shin, 2019; Augustin, Chernov, Schmid and Song, 2020; Liao and Zhang, 2020; Cerutti, Obstfeld and Zhou, 2021). - For advanced economies, CIP deviations are enlarged when the broad dollar index appreciates, or global financial intermediaries’ intermediary leverage ratio (He, Kelly and Manela, 2017) tightens. - Event-study evidence is reported for two sample periods: 2010-2015 (when G-SIB regulation were not in place) and 2016-2021 (after regulations were enacted). ### Costly intermediation mechanism (Liao and Zhang, 2020 framework) - Net creditors with a positive external investment position hedge currency exposure by selling dollar forward; in times of financial stress, swap dealers with limited risk-bearing capacity require a lower price of the forward dollar to absorb demand pressure, producing: - More negative CIP deviations (local currency overvalued in forward market relative to spot, after adjusting for interest rate differentials). - For international net debtors (typical for many EMEs when excluding reserve accumulation), the international investor funds in dollars, converts to local currency to buy local-currency assets and demands dollar-forward hedges from off-shore FX swap dealers. - A rise in global risk aversion reduces FX swap dealers’ intermediation capacity; the price of the dollar in the forward market rises by even more than spot appreciation because dealers require a higher spread (more expensive forward dollar) to intermediate hedging demand. - Hedging demand itself may rise as previously unhedged carry traders protect downside, further increasing demand for dollar forward. - Net effect: an overreaction of local currency depreciation in the forward market relative to spot — a larger f − s leads to a positive change in off-shore CIP deviations. ### Market segmentation and differential transmission to on-shore markets - With perfect arbitrage, off-shore pressures transmit fully to on-shore markets; both on-shore and off-shore CIP deviations would respond similarly to global risk-aversion shocks. - In practice, for a number of EM currencies, significant limits to arbitrage exist due to restrictions on domestic FX intermediation and derivative market position limits and participation constraints, segmenting FX swap markets. - On-shore FX dealers (usually domestic banks) are less directly affected by global risk-off shocks and thus reduce intermediation capacity by less; forward local currency may underreact relative to spot, so: - The on-shore basis would be less sensitive to global financial shocks; a negative response of the on-shore basis is also possible. ### Empirical patterns: wedges between on-shore and off-shore CIP deviations - Significant wedges between on-shore and off-shore CIP deviations are observed in the data. - For a group of eight currencies with a considerably large wedge (BRL, CNY, IDR, INR, MYR, PHP, THB, TWD), off-shore minus on-shore CIP deviations: - Fluctuate around zero during normal times. - Become positive in times of economic stress, driven by over-depreciation of local currency in the off-shore forward market relative to on-shore market. - Two recent global risk-off episodes highlighted: - The Taper Tantrum of 2013 (Panel (a)). - The onset of COVID-19 pandemic in early 2020 (Panel (b)). - In these episodes, the difference amounts to 300-400 basis points. ### Role of market integration and hedging supply - The sensitivity difference of offshore CIP deviations to global factors depends on the degree of segmentation in the forward market. - International banks are often dominant counterparties for non-deliverable FX forward through market-making roles and face increasing hedging demand from international mutual funds’ investment in local-currency bond markets. - For currencies with integrated forward markets, on-shore and off-shore participants can share risk, reducing segmentation and flattening the hedging supply curve. - For currencies with segmented markets, a steeper supply curve for hedging services amplifies the response of offshore CIP deviations when global risk-aversion tightens. ### Hypotheses to be tested empirically - Hypothesis 1. The sensitivity of CIP deviations computed using onshore forward exchange rates to global factors is smaller than their counterparts computed using offshore forward exchange rates. - Hypothesis 2. The sensitivity of offshore CIP deviations to global risk factors for currencies with segmented FX markets is stronger compared to their counterparts with integrated FX markets. *IMF Working Paper — 4.1 Theoretical underpinnings and hypotheses: Basis sensitivity, costly financial intermediation, and segmented markets* ### 4.2 Regression evidence ### 4.2 Regression evidence ### Methodology - Panel regressions of 3-month CIP deviations (off-shore and on-shore) on global macro-financial factors using monthly averages and first differences. - CIP deviations and forward market liquidity measure winsorized at 1% and 99% tails. - Currency fixed effects included; two-way clustered standard errors by time and currency. - Use same interest rate and spot exchange rate to compute off-shore and on-shore CIP deviations for each currency. - Time coverage: regressions reported for 2002–2021 and 2010–2021 subsamples; subgroup analyses for integrated versus segmented FX forward markets. - Forward market liquidity measure: (Fask − Fbid) / Fmid × 10000, where Fmid = (Fask + Fbid)/2. - Aggregate EM currency FX dealer leverage ratio constructed from largest EM currency dealer banks’ intermediary capital ratio ICRi,t = Market equityit / (Market equityit + Book debtit); dealer leverage is the inverse (log used in regressions). ### Variables included (baseline) - ∆(rUS − r): change in the (nominal) USD A2/P2 commercial paper rate minus local nominal money market rate. - ∆ log dealer leverage: change in log of aggregate EM currency FX dealer leverage ratio. - ∆ fwd bid-ask: change in normalized forward bid-ask spread. - In Panel (b): safe haven common factor (first principal component of nominal effective exchange rates of USD, CHF, JPY) and safe haven residual (dollar residual after projecting broad dollar onto the common factor). ### Key regression results (Table 2 — Panel (a): Baseline panel regressions) - Column (1): 02-21, ∆ offshore - ∆(rUS − r): 0.231* (0.122) - ∆ log dealer leverage: 1.076*** (0.344) - ∆ fwd bid-ask: 0.872* (0.438) - Observations: 4,128 - R-squared: 0.069 - Column (2): 10-21, ∆ offshore - ∆(rUS − r): 0.164** (0.072) - ∆ log dealer leverage: 0.912*** (0.261) - ∆ fwd bid-ask: 0.828** (0.321) - Observations: 2,706 - R-squared: 0.042 - Column (3): Group I: integrated 10-21, ∆ offshore - ∆(rUS − r): 0.098 (0.081) - ∆ log dealer leverage: 0.610** (0.218) - ∆ fwd bid-ask: 0.442 (0.267) - Observations: 1,637 - R-squared: 0.036 - Column (4): Group I: no NDF 10-21, ∆ offshore - ∆(rUS − r): 0.093 (0.088) - ∆ log dealer leverage: 0.630** (0.254) - ∆ fwd bid-ask: 0.451 (0.325) - Observations: 1,110 - R-squared: 0.043 - Column (5): Group II: segmented 10-21, ∆ offshore - ∆(rUS − r): 0.350*** (0.095) - ∆ log dealer leverage: 1.348** (0.428) - ∆ fwd bid-ask: 1.631* (0.779) - Observations: 1,069 - R-squared: 0.061 - Column (6): Group II: segmented 10-21, ∆ onshore - ∆(rUS − r): 0.212** (0.078) - ∆ log dealer leverage: 0.379 (0.292) - ∆ fwd bid-ask: 0.185 (0.285) - Observations: 1,087 - R-squared: 0.057 ### Key regression results (Table 2 — Panel (b): Add safe haven dollar factor) - Column (1): 02-21, ∆ offshore - ∆(rUS − r): 0.211* (0.113) - ∆ log dealer leverage: 0.654** (0.290) - ∆ fwd bid-ask: 0.786* (0.421) - safe haven common factor: 71.086** (28.614) - safe haven residual: 8.291*** (2.819) - Observations: 4,128 - R-squared: 0.093 - Column (2): 10-21, ∆ offshore - ∆(rUS − r): 0.156** (0.065) - ∆ log dealer leverage: 0.599** (0.235) - ∆ fwd bid-ask: 0.698** (0.276) - safe haven common factor: 40.409* (20.225) - safe haven residual: 10.710*** (3.257) - Observations: 2,706 - R-squared: 0.070 - Column (3): Group I: integrated 10-21, ∆ offshore - ∆(rUS − r): 0.096 (0.080) - ∆ log dealer leverage: 0.491* (0.241) - ∆ fwd bid-ask: 0.392 (0.243) - safe haven common factor: 17.656 (17.769) - safe haven residual: 3.614 (2.599) - Observations: 1,637 - R-squared: 0.044 - Column (4): Group I: no NDF 10-21, ∆ offshore - ∆(rUS − r): 0.093 (0.088) - ∆ log dealer leverage: 0.547* (0.272) - ∆ fwd bid-ask: 0.390 (0.292) - safe haven common factor: 12.608 (19.050) - safe haven residual: 2.757 (2.645) - Observations: 1,110 - R-squared: 0.048 - Column (5): Group II: segmented 10-21, ∆ offshore - ∆(rUS − r): 0.317*** (0.050) - ∆ log dealer leverage: 0.734* (0.360) - ∆ fwd bid-ask: 1.368* (0.628) - safe haven common factor: 78.277* (35.639) - safe haven residual: 21.325** (6.199) - Observations: 1,069 - R-squared: 0.121 - Column (6): Group II: segmented 10-21, ∆ onshore - ∆(rUS − r): 0.210** (0.076) - ∆ log dealer leverage: 0.341 (0.268) - ∆ fwd bid-ask: 0.177 (0.278) - safe haven common factor: 4.363 (20.453) - safe haven residual: 1.398 (2.366) - Observations: 1,087 - R-squared: 0.059 ### Substantive findings and numeric magnitudes - A rise in the U.S. interest rate relative to the EM country (∆(rUS − r)) is associated with an increase in the basis across samples and market types (positive coefficients reported across columns). - FX forward bid-ask spread (∆ fwd bid-ask) is positively associated with CIP deviations, consistent with reduced forward-market liquidity widening the basis. - FX dealer leverage (∆ log dealer leverage) is positively correlated with offshore CIP deviations: - A one percentage point increase in the leverage ratio corresponds to 1.1 basis points increase in the off-shore basis for the entire sample, and 0.9 basis points for the post-crisis sample. - For segmented currencies (Group II), a one percentage point increase in the leverage ratio corresponds to a 1.35 basis point increase in the off-shore basis, but only a 0.38 basis point (statistically insignificant) increase in the on-shore basis. - For integrated currencies (Group I), a one percentage point increase in the leverage ratio corresponds to 0.61 basis point increase in the off-shore basis (statistically significant at the 5% level), more than 50 percent smaller than the effect for segmented currencies. - Safe-haven currency movements: - Safe-haven common factor and safe-haven residuals are positive and significant correlates of off-shore CIP deviations (Panel (b)), particularly for segmented currencies (e.g., common factor 78.277* and residual 21.325** in Column (5)). - On-shore basis responses to safe-haven factor and residuals are typically statistically insignificant. - Disconnect between on-shore and off-shore bases: - Evidence consistently shows off-shore CIP deviations respond more strongly and positively to intermediary deleveraging and safe-haven appreciation, while on-shore bases exhibit zero-to-negative or statistically insignificant responses. - Time-series regressions for individual segmented currencies corroborate panel results: offshore bases respond positively to tightening leverage and appreciation of safe-haven common component; on-shore bases typically have zero-to-negative coefficients. ### Interpretation and hypotheses tests - Findings are consistent with the costly financial intermediation / limited risk-bearing capacity framework: - Tightening of dealer leverage (declining risk-bearing capacity) reduces intermediaries’ willingness/ability to supply currency hedging, widening offshore CIP deviations for EM currencies. - Market liquidity deterioration (wider forward bid-ask spreads) and relative U.S. funding cost increases (∆(rUS − r)) amplify the basis. - Hypothesis 1 (offshore-onshore disconnect due to segmented markets): supported — segmented currencies show larger offshore responses to intermediary leverage shocks than on-shore. - Hypothesis 2 (greater sensitivity for segmented versus integrated markets to intermediary constraints and safe-haven moves): supported — coefficients for segmented currencies exceed those for integrated currencies. ### Robustness and additional checks - Results robust to winsorization and exclusion of post-2020 COVID-19 period in untabulated exercises. - Appendix robustness: 1-month CIP deviations and CIP deviations constructed using 3-month USD Libor shown in Appendix Tables (A5, A6) — referenced but not reproduced here. - Comparison with G-10: For G-10 currencies, tightening dealer leverage is associated with negative CIP deviation responses (cited literature), contrasting with EM results where offshore CIP deviations jump up during risk-off. *IMF Working Paper Uncovering CIP Deviations in Emerging Markets — Section 4.2 Regression evidence* ### 4.3 CIP deviations and country-specific correlates ### 4.3 CIP deviations and country-specific correlates ### Country-specific channels and empirical approach - Country-specific factors considered: sovereign default risk (5-year USD sovereign CDS spreads, residualized by first principal component), and FX intervention (monthly broad measure as percentages of GDP; positive = accumulation of FX reserves; lagged by one month). - Motivation: sovereign default risk can co-move with currency risk and spill over to onshore and offshore currency markets; EM central banks may intervene in spot and forward markets or accumulate FX reserves to stabilize spot exchange rates. - Estimation: monthly regressions of first differences of 3-month CIP deviations (offshore/NDF for most columns; onshore in one column). Two-way clustered standard errors by currency and time. CIP deviations winsorized at 1% and 99%. Country fixed effects included. ### Main empirical findings — sovereign default risk (Panel (a)) - General evidence is weak that sovereign risk is a significant correlate of short-term CIP deviations. - Statistical details (selected coefficients and diagnostics from Panel (a), column (2) — sample 10-21): - safe haven common factor: 37.994* (standard error 20.012) - safe haven residual: 9.772*** (standard error 3.313) - ∆ 5y residualized cds spread: 0.482** (standard error 0.221) - ∆ log dealer leverage: 0.593** (standard error 0.245) - ∆ fwd bid-ask: 0.587** (standard error 0.254) - Observations: 2,439 - R-squared: 0.076 - Although the 5-year CDS spread (residualized) is significant at 5% level for offshore CIP deviations after 2010, this statistical significance does not persist when splitting by specific currency groups. - Untabulated results: little association between CIP deviations and rating downgrades by international rating agencies. ### Main empirical findings — lagged FX intervention (Panel (b)) - Evidence linking CIP deviations and (lagged) FX intervention is relatively weak and mixed in sign across samples. - Selected coefficients and diagnostics from Panel (b), column (2) — sample 10-21: - safe haven common factor: 42.667* (standard error 20.396) - safe haven residual: 10.945*** (standard error 3.336) - FXI: 3.426 (standard error 2.070) — positive but statistically insignificant - ∆ log dealer leverage: 0.588** (standard error 0.234) - ∆ fwd bid-ask: 0.704** (standard error 0.278) - Observations: 2,706 - R-squared: 0.071 - Group-specific patterns: - For currencies with segmented forward markets (Group II), a one percentage point increase in the size of foreign-currency asset purchase (as a fraction of GDP) is associated with a 3.6 basis point decline in on-shore CIP deviations (text statement). - For the same segmented group, offshore CIP deviations correlate with FX intervention in the opposite sign (insignificant at 10% level), suggesting potential barriers to policy transmission. - Panel (b), column (6) (Group II onshore): FXI coefficient -3.642* (standard error 1.547), Observations 1,087, R-squared 0.063 — indicating a statistically significant negative association for that onshore specification. ### Interpretation and caveats - Signs of correlations between country-specific factors and CIP deviations are ambiguous because: a rise in sovereign risk can trigger a flight to safety and unwinding of hedges affecting forward premia, while FX interventions can impact both forward and spot markets. - FX intervention may be endogenous to market conditions; the intervention measure is included lagged by one month to mitigate endogeneity concerns. - Market segmentation can make onshore and offshore dynamics diverge: onshore FX interventions may have limited effect on offshore expectations and hedging costs, potentially amplifying offshore CIP deviation sensitivity to global factors. ### Policy implications and recommendations - Two types of policy instruments to implement “basis control” in EMs vulnerable to the global financial/dollar cycle: - Pre-emptive capital flow management and macroprudential measures, including constraints on participation in FX forward markets, to reduce onshore market sensitivity to global risk sentiments and dampen external finance premia. - Use segmentation in forward markets as policy space: central banks can intervene in forward markets (e.g., enter forward contracts to buy U.S. dollar forward at a high price) to provide downside protection to domestic dollar borrowers and set a ceiling on CIP deviations by offering cheaper currency hedges. - Tradeoffs and risks: - Imposing implicit taxes on currency hedges can discourage foreign participation in local-currency markets. - Tightening risk-absorbing capacities of global intermediaries could amplify offshore CIP deviations’ sensitivity to global factors. - Market segmentation prevents onshore interventions from fully stabilizing offshore investor expectations; high hedging costs in NDFs can trigger international investors to liquidate local currency bond holdings. - Operational considerations: - Non-deliverable forward (NDF) operations may be cost-effective and pose little threat to FX reserve stability because ex ante cost equals the gap between forward rate and expected exchange rate at settlement; if interventions stabilize expectations, operations could be profitable ex ante and ex post. - Evaluate macroeconomic impacts and welfare benefits of capital flow management measures and FX forward market intervention — recommended as a high-priority area for future research. *IMF Working Paper Uncovering CIP Deviations in Emerging Markets — section 4.3* ### 103447. NBER International Seminar on Macroeconomics 2020. ### NBER International Seminar on Macroeconomics 2020 (IMF Working Paper: Uncovering CIP Deviations in Emerging Markets) — Appendix extracts ### Literature and context - The appendix lists related literature on CIP deviations, global liquidity, FX interventions, non-deliverable forwards, sovereign risk, and intermediary leverage, including but not limited to: - Cerutti, Claessens, and Ratnovski (2017); Das, Gopinath, and Kalemli-Özcan (2021); Domanski, Kohlscheen, and Moreno (2016); Du (2019); Du, Tepper, and Verdelhan (2018); Du and Schreger (2016, 2022); Du, Im, and Schreger (2018); Gabaix and Maggiori (2015); Gilchrist et al. (2022); and others listed in the source. ### Data sources and measures (Appendix A, Table A1–A3) - Coverage: - Currencies included (examples shown): BRL, CNY, IDR, INR, MYR, PHP, THB, TWD, CLP, COP, CZK, HUF, ILS, KRW, MXN, PEN, PLN, RUB, TRY, ZAR. - Domestic interest rate tickers and onshore/offshore definitions are provided per currency (see Table A1). - FX dealer leverage measure: - Aggregate EM currency FX dealer leverage computed from large FX dealer banks listed in Table A2 (BNP Paribas, Barclays, Bank of America, Citigroup, Credit Suisse, Deutsche Bank, Goldman Sachs, HSBC, JP Morgan, Morgan Stanley, Societe Generale, Standard Chartered, State Street, UBS). - FX dealer leverage ratio defined as (Market equity + book debt)/Market equity for largest FX dealers for EM currencies (Table A3, Panel (c)). - Key variable definitions (Table A3 Panel (c)): - FXI: Adler, Mano, Chang and Shao (2021) measure of FX intervention (spot+forward); unit: percent GDP. - Forward bid-ask: 10000×(ask - bid)/mid (forward market liquidity). - Residualized CDS spread: residuals from regressing 5-year USD CDS on the first principal component of CDS spreads (source: Markit). - Summary statistics (Table A3): - log FX dealer leverage ratio: Obs 240; Mean 3.097; Std. Dev. .389; Min 2.332; Max 4.521; P50 3.08. - safe haven currency common factor: Obs 240; Mean .153; Std. Dev. .963; Min -1.577; Max 2.266; P50 -.172. - safe haven residual: Obs 240; Mean -1.9976.288; Min -18.194; Max 14.847; P50 -1.505. - 3-month offshore CIP deviations (bps): Obs 4368; Mean -28.144; Std. Dev. 130.112; Min -537.634; Max 646.177; P50 -20.942. - 3-month onshore CIP deviations (bps): Obs 2056; Mean -39.599; Std. Dev. 93.264; Min -553.043; Max 280.301; P50 -22.853. - rUS − r (%): Obs 4500; Mean -3.5964.599; Min -47.3794.505; P50 -2.785. - offshore fwd bid-ask: Obs 4546; Mean 25.864; Std. Dev. 18.715; Min 5.165156.905; P50 21.023. - onshore fwd bid-ask: Obs 4397; Mean 22.905; Std. Dev. 17.512; Min 3.262144.808; P50 18.414. - 5y residualized cds spread (bps): Obs 4158; Mean -8.228; Std. Dev. 135.216; Min -804.393; Max 2642.105; P50 -8.558. - FXI (% GDP): Obs 4798; Mean .135; Std. Dev. .853; Min -7.8910.82; P50 .05. - Notes: - CIP deviations and liquidity (bid-ask) are winsorized at 1% and 99% tails. - Sample runs from 2002M1 to 2021M12 (figures and regressions use various subperiods noted in tables). ### Figures (descriptions) - Figure A1: Daily onshore (red) and offshore (blue) 3-month CIP deviations for BRL, CNY, IDR, INR, MYR, PHP, THB and TWD (2004–2021); gray horizontal lines at zero. - Figure A2: Time-series β of monthly change of offshore/NDF 3-month CIP deviations on monthly changes in log FX dealer leverage ratio (Panel (a)) and on safe-haven currency common factor (Panel (b)). Regression controls for interest rate differential and forward bid-ask spread. Error bands: 90% confidence interval with Newey-West standard errors with 12 lags. CIP deviations winsorized at 1% and 99%. - Figure A3: Cross-country scatter plots (2010–2021) of benchmark 3-month offshore CIP deviations against: - (a1) Interest rate differential: EM (with TWD), correlation -26% shown for one panel. - (b1) Net IIP (minus reserves) to GDP ratio: EM (with TWD), correlation -16% shown for one panel. - Note: Interest rate spread for EM currencies = 3-month money market rate − 3-month US A2/P2 commercial paper rate. - Figure A4: Czech Kruna (CZK) 3-month CIP deviations and forward premia during exchange rate floor (11/07/2013 to 04/06/2017). Forward premium = log 3-month forward − log spot (Kruna per USD). - Figure A5: Evolution of key global factors: - (a) Log intermediary leverage (He-Kelly-Manela primary dealer leverage and FX dealer leverage constructed from selected FX dealer banks). - (b) Safe haven currency common factor (first principal component of USD, CHF, JPY nominal effective exchange rates) and USD safe-haven residuals. ### Regression evidence — G10 currencies (Table A4) - Monthly first-difference regressions of changes in 3-month Libor basis (∆ Libor basis). - Panel (a): He, Kelly and Manela (2017) primary dealer leverage factor: - Sample 02-21 (column (1)): - ∆(rUS − r): coefficient -4.635 (standard error 2.658). - ∆ log primary dealer leverage: coefficient -0.094 (standard error 0.061). - ∆ fwd bid-ask: coefficient -0.458 (standard error 0.398). - Observations 2,242; R-squared 0.020. - Sample 10-21 (column (2)): - ∆(rUS − r): coefficient -11.659 (7.869). - ∆ log primary dealer leverage: coefficient -0.201** (0.083). (** p<0.05) - ∆ fwd bid-ask: -0.290 (0.472). - Observations 1,344; R-squared 0.047. - Columns (3)-(4) drop AUD/NZD; similar patterns with significance for ∆ log primary dealer leverage in column (4) (-0.233**, 0.095). - Panel (b): FX dealer leverage factor: - Sample 02-21 (column (1)): - ∆(rUS − r): -4.879* (2.615). (* p<0.1) - ∆ log fx dealer leverage: -0.074 (0.046). - ∆ fwd bid-ask: -0.445 (0.395). - Observations 2,242; R-squared 0.023. - Sample 10-21 (column (2)): - ∆(rUS − r): -11.922 (7.815). - ∆ log fx dealer leverage: -0.132* (0.062). (* p<0.1) - ∆ fwd bid-ask: -0.307 (0.482). - Observations 1,344; R-squared 0.050. - Note: Two-way clustered standard errors by currency and time reported. Columns drop AUD/NZD in (3)-(4) as robustness. ### Regression evidence — EM currencies (1-month tenor, Table A5) - Dependent variable: changes in the 1-month CIP deviations (offshore/NDF or onshore depending on column). - Panel (a): FX dealer leverage (baseline) - Full sample 02-21 (column (1), ∆ offshore): - ∆(rUS − r): 0.154 (0.124). - ∆ log dealer leverage: 1.189*** (0.406). (*** p<0.01) - ∆ fwd bid-ask: 1.695** (0.659). (** p<0.05) - Observations 3,888; R-squared 0.038. - Subsample 10-21 (column (2), ∆ offshore): - ∆ log dealer leverage: 0.933** (0.328). (** p<0.05) - ∆ fwd bid-ask: 1.621* (0.849). (* p<0.1) - Observations 2,630; R-squared 0.021. - Group I (integrated) 10-21 (column (3), ∆ offshore): - ∆ log dealer leverage: 0.450* (0.223). (* p<0.1) - ∆ fwd bid-ask: 0.915 (1.086). - Observations 1,558; R-squared 0.015. - Group II (segmented) 10-21 (column (5), ∆ offshore): - ∆ log dealer leverage: 1.697** (0.643). (** p<0.05) - ∆ fwd bid-ask: 3.714*** (1.016). (*** p<0.01) - Observations 1,072; R-squared 0.038. - Column (6) reports ∆ onshore for Group II: ∆(rUS − r) coefficient 0.198*** (0.038). - Panel (b): Add dollar factors (safe haven common factor and safe haven residual) - Full sample 02-21 (column (1)): - safe haven residual: 10.663*** (3.635). (*** p<0.01) - safe haven common factor: 50.303 (35.191). - ∆ log dealer leverage: 0.796** (0.371). (** p<0.05) - ∆ fwd bid-ask: 1.570** (0.640). (** p<0.05) - Observations 3,888; R-squared 0.051. - Subsample 10-21 (column (2)): - safe haven residual: 13.563*** (4.272). (*** p<0.01) - ∆ log dealer leverage: 0.575* (0.305). (* p<0.1) - Observations 2,630; R-squared 0.038. - Group II (segmented) 10-21 (column (5)): - safe haven residual: 25.042** (8.229). (** p<0.05) - safe haven common factor: 67.190 (70.916). - ∆ fwd bid-ask: 3.239*** (0.845). (*** p<0.01) - Observations 1,072; R-squared 0.069. ### Regression evidence — IBOR basis and robustness (Table A6) - Dependent variable: changes in Libor 3-month deviations from CIP (offshore/NDF or onshore). - Panel (a): Baseline panel regressions - Full sample 02-21 (column (1), ∆ offshore): - ∆(rUS − r): 0.127 (0.080). - ∆ log dealer leverage: 0.670** (0.251). (** p<0.05) - ∆ fwd bid-ask: 0.504 (0.317). - Observations 4,205; R-squared 0.025. - Subsample 10-21 (column (2)): - ∆ log dealer leverage: 0.434 (0.263). - ∆ fwd bid-ask: 0.736* (0.387). (* p<0.1) - Observations 2,780; R-squared 0.018. - Group II (segmented) 10-21 (column (5)): - ∆ log dealer leverage: 1.157* (0.584). (* p<0.1) - ∆ fwd bid-ask: 1.835* (0.958). (* p<0.1) - Observations 1,098; R-squared 0.048. - Panel (b): Add safe haven dollar factor and residualized CDS spread - Full sample 02-21 (column (1)): - safe haven residual: 7.685** (2.889). (** p<0.05) - safe haven common factor: 35.045 (20.563). - ∆ log dealer leverage: 0.341 (0.210). - ∆ 5y residualized cds spread: 0.144 (0.098). - Observations 3,729; R-squared 0.046. - Subsample 10-21 (column (2)): - safe haven residual: 9.359** (3.718). (** p<0.05) - safe haven common factor: 18.071 (12.419). - ∆ 5y residualized cds spread: 0.459* (0.222). (* p<0.1) - Observations 2,505; R-squared 0.050. - Group II (segmented) 10-21 (column (5)): - safe haven residual: 23.189** (8.780). (** p<0.05) - safe haven common factor: 63.531* (26.602). (* p<0.1) - ∆ log dealer leverage: 0.689 (0.635). - ∆ 5y residualized cds spread: 0.555 (0.508). - Observations 823; R-squared 0.128. - Notes on significance reporting: - *** p<0.01, ** p<0.05, * p<0.1. - Two-way clustered standard errors by currency and time reported in regressions. ### Key empirical patterns (as reported in the appendix tables and figures) - FX dealer / intermediary leverage and forward market liquidity are consistently associated with changes in CIP deviations: - For 1-month CIP deviations (offshore), ∆ log dealer leverage is positive and statistically significant in the full sample (1.189***, 0.406) and in subsamples, especially for Group II (segmented) currencies (1.697**, 0.643). - Forward bid-ask (fwd bid-ask) increases are associated with larger changes in CIP deviations, particularly for segmented currencies (3.714***, 1.016 for 1-month offshore Group II). - Safe-haven dollar factors matter: - Safe haven residual (USD residual after projecting on common factor) is positive and often statistically significant (e.g., 10.663***, 3.635 in Table A5 Panel (b) full sample for 1-month). - Safe haven common factor coefficients are large in some segmented-currency regressions but with larger standard errors (e.g., 67.190, 70.916 in Table A5 Panel (b) column (5)). - Country-specific risk measures: - Residualized 5-year CDS spread enters positively in some IBOR regressions (e.g., 0.459* (0.222) in Table A6 Panel (b), column (2)), indicating that country-specific credit risk dynamics can be associated with CIP deviation changes for certain samples. - Heterogeneity by forward-market segmentation: - Group I (integrated or small offshore-onshore spread) versus Group II (substantial forward market segmentation) show distinct coefficient magnitudes and statistical significance patterns, with Group II often exhibiting larger sensitivity of CIP deviations to dealer leverage and forward market liquidity. *Source: IMF Working Paper — Uncovering CIP Deviations in Emerging Markets (Appendix material, tables and figures as provided).* --- _Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023028-print-pdf.pdf_