The spread of COVID-19 has led to large foreign exchange (FX) moves, as past global crises have, but both the scale of the epidemic and the speed of its global spread makes the current situation unique. In particular, the pattern of FX dynamics is fast-tracked and capital outflows from emerging markets (EMs), week on week, are much larger than in previous crises. Relative to the past, and in particular the 2007-2008 Great Financial Crisis (GFC), the recent prompt activation of central bank FX swap lines appears to have tempered dollar movements.
A recurrent pattern around economic crises
Figure 1 plots the evolution of key exchange rates against the dollar, indexed relative to 25th February 2020, when the US yield curve inverted (measured using the difference between 10-year and 1-year US zero-coupon government bond yields). For a couple of weeks after the inversion, the US dollar lost value against the euro and the Japanese yen – while sterling remained broadly stable – and EM currencies depreciated somewhat. After that, from around the second week of March, the dollar strengthened markedly against all currencies to 19th March, when the Federal Reserve announcement the establishment of temporary swap lines with a range of central banks—in addition to the extensions to its pre-existing swap line arrangements announced on 15th March.
Figure 1 COVID FX: Exchange rate dynamics around 25th February 2020 US yield curve inversion
Note: Vertical solid black line denotes date of the US yield curve inversion, on 25th February 2020, where yield curve slope defined as the 10 minus 1-year yield zero-coupon yield. Exchange rates normalised relative to this date. Vertical dashed grey lines denote dates of Federal Reserve announcements to (a) extend the maturity of its existing swap line agreements with the Bank of Canada, Bank of England, Bank of Japan, ECB and SNB on 15th March 2020 and (b) establish temporary swap line arrangements with central banks in Australia, Brazil, Denmark, Korea, Mexico, Norway, New Zealand, Singapore and Sweden on 19th March 2020. USDEM7 a PPP-weighted average of 7 EM currencies: Brazil, India, Indonesia, Mexico, Russia, South Africa, Turkey. Dates: 3rd February 2020 to 30th March 2020.
Data Source: Datastream.
The overall pattern in Figure 1 is not new and, in Figure 2, we illustrate that a similar pattern emerged during the GFC. Preceding that, the US yield curve was inverted for a protracted period of time, from June 2006 until June 2007. Following the end of this period of inversion, FX dynamics followed a pattern that, qualitatively, is very close to Figure 1. The euro and, to a lesser extent sterling, first strengthened relative to the dollar after the end of the inversion, before strong dollar appreciation in the second half of 2008. Relative to Figure 1 and the current COVID crisis, however, the FX patterns from the GFC are in comparative ‘slow motion’ (it did not feel that way at the time): the time scale for Figure 2 is months, but days in Figure 1. In the COVID-19 crisis, the dollar started to appreciate within two weeks of the initial yield curve inversion. In 2007-2008, it materialised after six months.
Figure 2 Exchange rate dynamics around the end of the 2006-2007 US yield curve inversion
Note: Vertical solid black line denotes the end date of the 2006-2007 US yield curve inversion on the 5th June 2007, where yield curve slope defined as the 10 minus 1-year yield zero-coupon yield. Exchange rates normalised relative to this date. USDEM7 a PPP-weighted average of 7 EM currencies: Brazil, India, Indonesia, Mexico, Russia, South Africa, Turkey. Dates: 1st January 2007 to 30th November 2008.
Data Source: Datastream.
Following the economic literature, think of an economic disaster as a set of events (originating from a variety of shocks including supply or financial) causing a significant drop consumption across a large number of countries and a sharp depreciation of their currencies – in practice vis-à-vis the dollar – a definition used in Barro (2006) and Gabaix (2012). In Lloyd and Marin (2019) and Corsetti and Marin (2020) disasters are preceded by an inversion of the US yield curve (unrelated to monetary policy) – although, to be clear, as is widely known not all yield curve inversions are followed by disasters or recessions. Intuitively, the yield curve reflects investors’ expectations of a crisis at some point the future. According to the theory, an inverted yield curve signals that investors believe the risk of a disaster ahead is relatively high in the short run. Thus, short-run yields must compensate investors for bearing risk. These expectations also dominate FX markets: relatively risky currencies deliver a risk premium as compensation for the large depreciation they experience when the crisis materialises (Farhi and Gabaix, 2016). We argue that this reasoning underlies the pattern in Figure 1 and 2.
Differences with today’s events
There are, of course, further differences to be highlighted across the 2020 and GFC episodes concerning behaviour of FX and bond markets:
- The most recent US yield curve inversion (on 25th February 2020) was driven by a sharp fall in longer-term yields and was brief. In contrast, in 2006-2007, the US yield curve inversion was associated with rising short-term interest rates, and elongated.
- In the current crisis, central bank FX swap lines have been promptly (re-)activated by the Federal Reserve and appear to have stemmed the rise of the dollar against the countries they cover (advanced economies and some EMs), as the most recent observations in Figure 1 indicate.
Bond yields and FX
Lloyd and Marin (2019) document that the FX dynamics of Figures 1 and 2 after yield curve inversion are recurrent and quite general, by showing that the pattern in these figures can be mapped into a time-varying estimated covariance between interest rate differentials and exchange rates. This is obtained by running standard Fama (1984) regressions, testing the well-known, uncovered interest rate parity condition (UIP).
The UIP condition predicts that investors need to be compensated with proportionally higher yields to willingly hold currencies that are expected to depreciate. By augmenting the canonical UIP regression with an interaction between interest rate differentials and an indicator of US yield curve inversions, and its lags, Lloyd and Marin (2019) show that, on average, the coefficient on the interest spread becomes smaller (typically negative) immediately following the US yield curve inversion. Here, as investors require excess returns from high-yield currencies, risky in view of the possibility of a disaster, the dollar depreciates. Then, when the disaster occurs, the coefficient increases sharply, rising above 1, consistent with a large appreciation of the dollar.
These results of the Fama regressions are shown in Figure 3 for a sample of six advanced economy currencies vis-à-vis the US dollar for the period 1980 to 2017. In this setting, the coefficient takes, on average, around eight months following a US yield inversion to switch sign and exceed 1.
Figure 3 Changes in the Fama coefficient in the months following US yield curve inversions associated with economic disasters
Source: Lloyd and Marin (2019). Note: Red dots denote the estimated interaction coefficient between 6-month interest rate differentials and a lagged US yield curve inversion indicator. The horizon axis denotes the lag in this interaction, in months. Coefficients estimated using six currencies vis-à-vis the dollar from 1980 to 2017, with country fixed effects. 95% confidence bands constructed using Driscoll and Kraay (1998) standard errors.
These findings are not specific to recent decades either. Using a century-long sample of data for the US and the UK, Corsetti and Marin (2020) show that the time-variation in the UIP coefficient also occurred during the Great Depression, and across different monetary regimes. Taken together, these findings highlight a systematic pattern in FX dynamics following US yield curve inversions that are associated with subsequent economic disasters, which cannot otherwise be attributed to, for example, quantitative easing or other unconventional monetary measures implemented in the GFC.
While the scale of the dollar appreciation to date is not as large as in 2007, the associated capital outflows from EMs, week on week, are many times larger than at the peak of the GFC
Concluding remarks: “The largest capital outflows ever recorded from the emerging markets”
Coming back to the current COVID-19 crisis, while the scale of the dollar appreciation to date is not as large as in 2007, the associated capital outflows from EMs, week on week, are many times larger than at the peak of the GFC. Real-time data also indicate that capital outflows from EMs week on week have been double the peak weekly outflows seen around the 2013 US taper tantrum – stressing the EM economies to the max. As of the end of March, nearly 80 countries are requesting IMF help, while, according to the IMF, $83 billion have left them since the beginning of the crisis.
While these outflows may subside, the evidence from Figure 3 would be consistent with further pressure on the dollar to appreciate in the near future. As many of these EMs are central to the global production network, policy action to sustain supply chains and demand for commodities (for example food) will be crucial in helping many contain the economic disruption.
Barro, R (2006), “Rare Disasters and Asset Markets in the Twentieth Century”, The Quarterly Journal of Economics 121(3): 823-866.
Corsetti, G and E A Marin (2020), “A Century of Arbitrage and Disaster Risk Pricing in the Foreign Exchange Market”, CEPR Discussion Paper No. 14497.
Fama, E F (1984), “Forward and Spot Exchange Rates”, Journal of Monetary Economics 14: 319-338.
Farhi, E and X Gabaix (2016), “Rare Disasters and Exchange Rates”, The Quarterly Journal of Economics 131(1): 1-52.
Gabaix, X (2012), “Variable Rare Disasters: An Exactly Solved Framework for Ten Puzzles in Macro-Finance”, The Quarterly Journal of Economics 127(2): 645-700.
Lloyd, S P and E A Marin (2019), “Exchange Rate Risk and Business Cycles”, Cambridge Working Papers in Economics, No. 1996.