CIP deviations, the dollar, and frictions in international capital markets

Because the fragmentation of the global dollar funding market is what makes intermediaries indispensable, and therefore what makes their constraints show up in prices. The disproportion is stated with numbers: “While the United States accounts for about 15% of world trade and 25% of global GDP, the U.S. dollar accounts for about 50% of cross-border loans and international debt securities, 90% of all foreign exchange (FX) transactions, 60% of official FX reserve holdings, and 50% of trade invoicing” (Introduction, p. 1, citing BIS 2020, Maggiori-Neiman-Schreger 2020 and Gopinath-Stein 2018). The consequence: “the global market for dollar funding is highly fragmented. Many market participants in need of dollars lack access to direct dollar funding from dollar-rich lenders, such as retail and corporate depositors, U.S. money market funds (MMFs), and central bank reserve managers. As a result, financial intermediaries (i.e. large global banks) play a particularly crucial role channeling the global supply of dollars.” Since the crisis their balance sheet constraints “have tightened considerably, in part due to regulatory reforms on the banking sector. The most salient manifestation of these constraints is the breakdown of the covered interest rate parity (CIP) condition” (p. 1).

Very sharp, especially at five-year maturity, and yes – the deviations persist and are on average larger with OIS or repo rates than with IBOR (Section 3.1, pp. 9-10). “CIP deviations were close to zero before the GFC, increased significantly during the GFC, and then remained elevated post-GFC… The pre- and post-GFC dichotomy is even more stark for the 5-year cross-currency basis.” The historical baseline is drawn from the pre-crisis literature: Frenkel and Levich (1975, 1977) confirming CIP “held exceedingly well among currency pairs in the offshore market,” Popper (1993) reaching the same conclusion for long-term rates via cross-currency swaps, and Akram, Rime and Sarno (2008) showing that “short-lived CIP deviations (from 30 seconds to 40 minutes) were documented, but generally dissipated very quickly, consistent with the efficient market hypothesis” (fn. 2). The crucial post-crisis point is the absence of distress: “During the peak of GFC, the CIP condition broke down, but significant CIP deviation remained post-GFC, even under tranquil financial market conditions.” And the benchmark check: “IBOR rates are imperfect benchmarks for the risk-free rate due to their indicative and unsecured nature. However, CIP deviations exist beyond IBOR… the CIP deviations remain, and even become larger on average, if we use Overnight Index Swap (OIS) rates or repo rates,” where OIS involves “little credit risk” and the repo rates “are collateralized with government bonds and are free from credit risk” (pp. 9-10). A further check on whether benchmark rates are real transaction rates: Anderson, Du and Schlusche (2019) “use banks’ actual funding rates from the unsecured wholesale funding markets through the issuance of certificates of deposits and commercial paper, and show that more than 95% of this funding is raised at a cost below the implied dollar funding rate from the dollar-yen swap market by top-tier global banks” (fn. 8).

Generally negative for G10 currencies – the Australian and New Zealand dollars excepted – meaning the cash dollar rate is below the synthetic one. “A negative cross-currency basis means that the direct dollar interest rate from the cash market is lower than the synthetic dollar interest rate. With a negative basis, banks would earn a profit by borrowing dollars in the cash market and lending in the FX swap market” (Section 3.1, p. 9). The chapter’s summary formulation is that “the failure of the CIP condition has become the new normal,” with a basis that “is highly correlated with the nominal interest rates across currencies and co-moves with global risk factors” (Introduction, pp. 2-3).

A 90 percent cross-sectional correlation since 2008, which makes the hedged arbitrage the mirror image of the unhedged carry trade (Section 3.2, pp. 10-11). “High-interest-rate currencies, such as the Australian dollar (AUD) and the New Zealand dollar (NZD), tend to have positive cross-currency bases… Conversely, low-interest-rate currencies, such as the euro (EUR) and the Swiss franc (CHF), have very negative cross-currency bases… The cross-sectional correlation between the cross-currency basis and the nominal interest rate level is 90% since 2008.” The trade direction: “To arbitrage CIP deviations, an investor should go long in the low-interest-rate currencies and short in the U.S. dollar, or short in the high-interest-rate currencies such as the AUD and NZD, while fully hedging against the foreign currency risk” – “the exact opposite of the classical unhedged FX carry trade.” The worked example uses the average three-month dollar-yen basis of negative 25 basis points post-crisis, executed by borrowing dollars in the cash market, converting spot into yen, investing in yen, and selling yen forward.

About 20 basis points a year unlevered and riskless for CIP arbitrage, against roughly 5 percent a year with substantial risk for the carry trade. “The average annualized profit on an unlevered three-month CIP arbitrage earns about 20 basis points, whereas the UIP trade earns an average annualized return of about 5%. However, the key difference is that the CIP arbitrage profits are known ex-ante and involve no risk, and therefore, the Sharpe ratio of CIP arbitrage is essentially infinite. In contrast, the UIP trade has an considerable amount of risk, especially during bad times, and has a Sharpe ratio equal to 0.54 based on data from 1994 to 2010 (Hassan and Mano (2019))” (Section 3.2, p. 11). The chapter draws the practical consequence: “The small, but sure profits of CIP arbitrage underscores the importance of using leverage to scale up the return to an attractive level” – which is precisely why a leverage constraint bites.

That most of the movement is common across currencies and lines up with global risk measures, making the basis a barometer of intermediary conditions (Section 3.3, pp. 11-12). A principal component analysis on quarterly changes in the five-year G10 bases over 2008Q1-2020Q3 finds “the first principal component (PC) explains 51% of the total quarterly variation in the bases. This implies that much of the overall movement in deviations from CIP is driven by common changes in funding conditions, with the various currencies loading heterogeneously on this common variation.” The first PC “is effectively a ’level’ factor, as it is 96% correlated with the quarterly changes in the average cross-currency basis across all G10 currencies.” Bad times widen it: “the cross-currency basis becomes more negative… and generally larger CIP deviations in ‘bad times’ when the global financial conditions are tighter. During these bad times, we generally have a strong broad dollar (Avdjiev et al. (2019)), high VIX, a wide BBB-Treasury spread, low returns on the S&P index, and negative shocks to the intermediary capital ratio in He, Kelly and Manela (2017).” The interpretation offered: “the cross-currency basis is also an important barometer of leverage and risk-taking in the global capital markets, highlighting the importance of intermediary constraints,” and larger deviations “could reflect a combination of lower supply of dollar funding by financial intermediaries and higher demand for dollar funding by ultimate dollar borrowers” (p. 12).

The slope of the supply curve. Pre-crisis it was flat at a zero basis; post-crisis it slopes up, which is what allows demand to matter (Section 4.1, pp. 12-13). “Prior to the GFC, the supply of dollar funding and hedging in the FX swap market was perfectly elastic at the point where CIP holds. A flat supply curve at a cross-currency basis of zero prior to the GFC reflects the fact that there was no cost for financial intermediaries to supply dollars in the FX swap market regardless of the quantity demanded… The force of arbitrage would enforce the CIP condition, regardless of fluctuations in demand for dollar funding.” Afterwards, “the supply curve for dollar funding became upward sloping, as it became increasingly costly for banks to provide larger quantities of dollar funding and hedging services… banks now need to be compensated with a higher CIP deviation in order to be willing to increase their supply of dollar funding in the FX swap market. With an upward sloping supply curve, shifts in the demand for dollar funding (and hedging) can also cause fluctuations in the equilibrium CIP deviation.” This is the chapter’s central analytical move: the demand side only becomes price-relevant once the supply side is constrained.

Because the leverage ratio is non-risk-weighted, so a matched-book CIP trade consumes capacity despite carrying no risk. “Non-risk-weighted capital requirements, in the form of the leverage ratio requirement, mandate banks to maintain capital against all assets, regardless of their risk characteristics, and thereby restrict the ability of banks to engage in traditional CIP arbitrage (Duffie (2017)). In particular, the traditional ‘matched-book’ CIP arbitrage requires large banks to borrow dollars in the cash market and lend dollars in the FX swap market. Even though the position is riskless, it still expands the size of bank balance sheet and therefore increases the bank’s leverage ratio potentially generating a binding leverage constraint” (Introduction, p. 3). The chapter calls the leverage ratio “the pivotal regulatory constraint for short-term arbitrage as it limits the size of positions that can be taken by global banks” and records that the requirement rose “from 3%, prior to Basel III, to the current level of 5-6% for U.S. banks” (Section 4.2). Corroborating evidence is offered indirectly: “Binding leverage constraints are also consistent with the existence of a host of other near-arbitrages in the fixed-income market” (p. 3). The authors are explicit that this is a different class of explanation from the older ones: “These explanations based on the bank balance sheet constraints as a result of the post-crisis regulatory reform differ from the traditional explanations for the failure of the law of one price, including transaction costs, counterparty risk, convergence risk, margin constraints, liquidity, information, and sentiment, such as in Shleifer and Vishny (1997), Garleanu and Pedersen (2011) and Pasquariello (2014)” (p. 3).

Three client groups, each stuck with a currency mismatch they cannot fix in the cash market (Introduction, pp. 3-4, and Section 5.1). First, “non-U.S. banks that do not belong to the top-tier of global banks often have difficulty tapping into the direct dollar funding markets or can only do so at a considerably higher cost. Meanwhile, they have access to insured deposits in local currency and thus often choose to raise dollar funding in the FX swap market to finance their dollar-denominated assets.” Second, “institutional investors outside the U.S., such as pensions, insurances and mutual funds, have local currency liabilities and yet choose to invest a significant fraction of their portfolios in dollar-denominated assets. An FX swap arrangement allows them to fund their dollar portfolios and hedge the corresponding currency risk.” Third, “non-financial corporates that borrow in multiple currencies may face asymmetric funding costs relative to the local risk-free benchmark because bond markets are segmented by currency (Maggiori, Neiman and Schreger (2020)), and therefore may choose to use the cross-currency swap market to minimize their funding costs (Liao (2020)).” The chapter also notes that corporate-level deviations “could be the opposite sign of the CIP deviations faced by intermediaries based on Libor or OIS rates due to differences in credit conditions,” so that multi-currency issuers’ hedging demand “could push the Libor/OIS CIP deviations even wider” (fn. 14).

Through search for yield: low-rate-currency investors want hedged dollar and high-rate-currency exposure, and intermediaries charge the basis to supply it (Section 5.2, pp. 23-24). “The search for yield motive prompts clients in low-interest-rate countries (e.g., Japan, euro area, and Switzerland) to go long in high-interest-rate currencies (e.g. the U.S. dollar and the Australian/New Zealand dollar) to earn the interest rate carry. This cross-currency investment demand generates dollar funding and hedging demand, which has to be provided by financial intermediaries… Since the low-interest-rate currencies on a hedged basis offer higher synthetic dollar yields than high-interest-rate currencies, financial intermediaries earn CIP deviations as an intermediation fee to justify using up precious balance sheet space.” Liao and Zhang (2020) are cited for linking “the cross-country pattern of CIP deviations to net U.S. dollar asset holdings.” In the diagram’s terms, “an unexpected widening of the nominal interest rate between the U.S. and foreign currency shifts the demand curve to the right, widening the CIP deviations,” and the empirical literature confirms that monetary policy surprises move deviations – specifically “unexpected tightening in the U.S. monetary policy and unexpected accommodation in foreign monetary policy lead to a widening of the CIP deviations of the foreign currency vis-à-vis the dollar.” Yield-curve slope differences also correlate with the basis in time series (Iida, Kimura and Sudo 2018).

That offshore dollar funding conditions are now tighter than cash-market indicators suggest, and tighten further when the Fed surprises on the hawkish side. “Prior to the GFC, when the CIP condition held, the offshore dollar funding conditions in the FX swap market were nearly identical to the cash markets. Post-GFC, a negative cross-currency basis makes the offshore dollar funding condition tighter than the cash market funding condition, gauged by traditional indicators such as the Libor-OIS spread. Furthermore, unexpected U.S. monetary policy tightening pushes the cross-currency basis more negative, further tightening the offshore dollar funding conditions” (Section 5.2, p. 24). The chapter points to Amador et al. (2020) for “a framework with financial intermediaries where CIP deviations emerge endogenously when interest rates are at the zero lower bound.”

They rotate the crisis-time supply curve back out, and have been drawn heavily in all three recent crises (Section 5.3, pp. 25-26). The mechanics: “Local banks in countries with a swap line agreement with the Fed can borrow dollars from their respective central banks by pledging eligible collateral in their local currency. Their central banks then tap the central bank swap lines to borrow dollars from the Fed against their local currencies.” Before COVID the Fed had standing facilities with five central banks – Canada, England, the ECB, Japan and Switzerland – with temporary agreements added during the pandemic. Pricing is “a fixed rate over the ongoing OIS rate. The spread over the OIS rate was initially set to be 100 basis points during the peak of the GFC, and was lowered to 50 basis points in November 2011 to address the funding strain induced by the European Debt Crisis, and was again lowered to 25 basis points in March 2020 in response to the COVID pandemic.” Peak outstanding was “about $580 billion at the height of the 08-09 global financial crisis, $110 billion at the peak of the European debt crisis, and $450 billion during the COVID pandemic,” with maturities from one week to 84 days. The assessment is attributed rather than asserted: “Overall, central bank swap lines are found to be quite effective in restoring market conditions in the global dollar funding markets (Goldberg, Kennedy and Miu (2010); Bahaj and Reis (2018, 2020); Cetorelli, Goldberg, Ravazzolo et al. (2020)).” The reason a backstop is needed is spelled out: “The inability to roll-over short-term dollar funding and hedging can lead to fire sales of dollar-denominated assets and propagate financial distress to longer-maturity assets.”

Because they need not be arbitrage at all – three distinct forces can separate a government yield from a risk-free-rate proxy (Section 6.1, pp. 27-28). “While CIP deviations between risk-free rates represent an arbitrage opportunity for global banks for the reasons discussed in the previous sections, CIP deviations may exist between government bond yields for several other reasons.” The three: a convenience yield, since “the yields could be lower than the risk-free rate because there are additional benefits to holding government bonds, such as greater liquidity and better collateral value”; sovereign default risk, which pushes the yield above the risk-free rate; and segmentation, where “the government bond yield could be different from the risk-free rate facing global investors due capital controls imposed by countries and the market segmentation between the domestic bond market and international capital market.” The upshot is that “CIP can then fail between government bond because of cross-country differences in convenience yields, sovereign default risk, and varying degrees of international capital market integration” – and, as the introduction puts it, “a higher synthetic dollar yield on foreign government bonds than the U.S. Treasury yield could capture the perceived relative default risk on the foreign government bonds (Du and Schreger (2016)),” while even without default risk a wedge can open “when investors value the safety and liquidity of one country’s bonds more than another” (pp. 4-5, citing Du, Im and Schreger 2018 and Jiang, Krishnamurthy and Lustig 2018).

One treats deviations as a signal of financial constraints that drive exchange rates; the other treats bond convenience yields as a determinant of exchange rates directly (Section 7, pp. 34-35). The first: Avdjiev et al. (2019) “argue that there should be a connection between movements in the broad U.S. dollar and Libor CIP deviations, because the dollar acts as a barometer of risk-taking capacity in global capital markets and the CIP deviations capture the shadow costs on the balance sheet constraints of the financial intermediary,” consistent with Gabaix and Maggiori (2015) on constrained intermediaries; Lilley et al. document a structural break around the crisis in the relationship between exchange rates, capital flows and risk measures. The chapter’s own characterisation of this strand is careful: “This first line of thinking treats CIP deviations essentially as a signal of financial market constraints, which in turn drive exchange rate movements, rather than as a determinant of exchange rates themselves.” The second strand works through liquidity premia: Engel (2016) shows “the UIP puzzle changes signs at various horizons, with high interest rate countries experiencing high returns in the short-run, but also forecasting lower returns at longer horizons,” and suggests bond liquidity premia can rationalise it, with Valchev (2020) offering a related account via endogenous convenience yields from monetary-fiscal interaction. The conceptual framing given: “the failure of UIP means that the path of interest rates will be insufficient to explain the value of a currency, but perhaps this well-known failure is explained by the existence of a non-pecuniary benefit that investors realize by holding different assets.”

Whether the post-crisis regulations are optimally calibrated, the role of non-banks, and the macroeconomic consequences – and it declines to take a firm position on the first. The chapter frames its own vintage honestly: “Given that a persistent failure of CIP has emerged for fewer than 15 years, the literature surveyed in this paper is still in a relative early stage” (Section 8, p. 37). On regulation it lays out both sides without resolving them: “While there is an inherent tradeoff between efficiency of financial intermediation and overall financial market stability, one view is that a binding non-risk-weighted capital constraint places us in the interior of the efficiency-stability frontier (Duffie (2017)). Meanwhile, it is clear that excessive leverage is one of the main contributors to past financial crises, and ex-ante assessments of risk weights of assets can be misguided. These reasons argue in favor of maintaining a minimum non-risk-weighted capital standard.” The open design questions are posed as questions: how to calibrate risk-weighted against non-risk-weighted requirements, “Should central bank reserves and/or Treasury securities be exempted from the leverage ratio calculation? Should the public disclosure of the leverage ratio move to a daily-averaging regime in all major jurisdictions to avoid reporting-period window-dressing?” The September 2019 repo turmoil is cited as evidence of frictions “beyond Basel III regulatory metrics, most prominently the intra-day liquidity constraint.” On non-banks: “There are very few systematic studies in this area,” and the chapter lists specific gaps for institutional investors, real-money dollar investors and corporate issuers. Its closing claim is interpretive rather than empirical: “The cross-currency basis has emerged from an esoteric corner of the FX derivative market to become an important barometer of the health of global capital markets. The breakdown of CIP puts financial frictions and intermediary constraints at the heart of international finance and macroeconomics” (Section 9, pp. 40-41).