Central bank swap arrangements in the COVID-19 crisis

Abstract

Facing acute strains in the offshore dollar funding markets during the COVID-19 crisis, the Federal Reserve (Fed) provided US dollar liquidity to the global economy by reactivating or enhancing swap arrangements with other central banks and establishing a new repo facility for financial institutions and monetary authorities (FIMA). This paper assesses motivations for the Fed liquidity lines, and the effects and spillovers of US dollar auctions by central banks using these lines. We find that the access to the Fed liquidity arrangements was driven by the recipient economies’ close financial and trade ties with the US. Access to dollar liquidity also reflected global trade exposure. We find that announcements of expansion of Fed liquidity facilities or of auctions using these facilities led to appreciation of partner currencies against the US dollar and reduced these currencies’ deviations from covered interest parity (CIP). Dollar auctions by major central banks (BoE, ECB, BoJ and SNB) had spillovers: they led to temporary appreciation of other currencies against the US dollar, reduced CIP deviations, and persistently reduced sovereign bond yields of other economies. However, dollar auctions done by non-major central banks with access to Fed facilities did not have a meaningful impact on key domestic financial variables. The impact of major central bank auctions does not differ by the economies’ financial or trade links with the US or their balance sheet currency exposure, i.e. the major central bank auctions benefitted even the more vulnerable economies.

1. Introduction

The COVID-19 crisis saw the emergence of acute strains in the offshore dollar funding markets in March 2020. These strains manifested as deviations from neo-classical arbitrage conditions, including deviations from Covered Interest Parity (CIP) (Fig. 1 ).1 As during the Global Financial Crisis (GFC) of 2008, the US Federal Reserve (Fed) took several actions to provide US dollar liquidity to global financial markets through foreign central banks. It reduced the pricing of swap operations, extended the maturity, and increased the frequency of swap operations with the five central banks with which it has standing swap lines. It also reactivated swap lines with nine other central banks with which it had established the lines during the GFC.

Fig. 1.

Cross-currency basis (3-month swaps, basis points) widened during the COVID-19 crisis Source: Bloomberg, JP Morgan.

Furthermore, for the first time, on March 31, 2020, the Fed announced the establishment of a temporary repurchase agreement facility for foreign and international monetary authorities (FIMA). With this facility, the Fed allowed existing FIMA account holders at the New York Federal Reserve (NY Fed) to borrow dollars against US Treasuries held in these accounts.2 The US dollars obtained by the central banks could then be used for onward lending to institutions in their jurisdictions, thus mitigating the US dollar liquidity shortage in these jurisdictions.

What motivates the Fed to extend dollar liquidity to selected foreign central banks? Research on the extension of swap lines during the GFC suggested that US bank exposure to the swap partner economies was an important motivator for the Fed to extend these lines (Aizenman and Pasricha, 2010). More broadly, central banks tend to extend swap lines only to central banks of economies with significant trade and financial linkages with their own economy (Aizenman et al., 2011).

Given that the economic environment in the COVID turmoil was different from that of the GFC (which was driven by a shock that originated in the US financial system) and that FIMA facility was added as another liquidity provision scheme, one may ask whether the factors motivating the US decision to provide dollar liquidity were different in this crisis. Did the Fed extend swap lines or develop the FIMA facility solely based on its self-interest as it did during the GFC? What factors determined economies’ access to dollar liquidity? One may also ask to what extent the size of US dollar auctions conducted by central banks reflected stress in domestic financial or currency markets.

The economic impacts of these liquidity facilities are also worth investigating. Were they effective in mitigating financial stress, and if so, which facilities? The announcements of swap lines during the GFC had relatively large short-run impacts on the exchange rates of the selected emerging market economics (EMs), but much smaller effect on the credit default swap (CDS) spreads, relative to that of other EMs that were not the recipients of swap-lines (Aizenman and Pasricha, 2010).

Further, EMs have long argued for broader expansion of US swap lines, and the FIMA facility reflected such an expansion. The amounts of auctions conducted by the major central banks, however, dwarfed the amounts of auctioned done by other central banks using Fed swap lines or FIMA facility (Fig. 2 ). Therefore, one can ask whether the impact of dollar auctions by the major central banks differ from those by other central banks using Fed facilities. Baba and Packer (2009) found that at the time of the GFC, US dollar term funding auctions by the ECB, SNB, and BoE, as well as the Fed’s commitment to provide unlimited dollar swap lines ameliorated the foreign exchange (FX) market dislocations. Rose and Spiegel (2012) found that US dollar auctions by major central banks disproportionately benefitted economies that had higher trade or asset exposure to the US, by reducing their CDS spreads.

Fig. 2.

Economies with access to Fed liquidity lines saw smaller decline in US treasury holdings between February and April 2020 Note: The figure plots the percentage change in the holding of US treasury securities between end-February and end-April 2020, for economies with or without access to Fed facilities as of the end of June 2020. The economies with Fed facilities include the 14 economies with Fed swap lines as well as Hong Kong (China), Indonesia, Chile, and Colombia. Economies without Fed facilities are 84 other economies on which data is available. Source: TIC data series “Foreign Holdings OF U.S. Treasury Securities, Billions of U.S. dollars, end of period”. Extracted via Haver.

We revisit these questions and investigate the announcement and implementation effects of the dollar liquidity lines on exchange rates, government bond yields, CDS spreads and cross-currency basis (CIP deviations). We distinguish between dollar auctions by an economy’s own central bank with access to Fed facilities and those by the four major central banks, i.e. the Bank of England (BOE), the Bank of Japan (BOJ), the European Central Bank (ECB), and the Swiss National Bank (SNB). We ask whether the impacts were different, i.e., whether the dollar auctions by the major central banks ameliorated financial stress in other economies and whether and how these spillovers differ from the impact of the auctions conducted by the smaller central banks with access to Fed liquidity lines, on their own economies. We use a local projection model and daily data from 15 January 2020 to 31 May 2020, on a sample of 39 economies to investigate these questions about the impact of liquidity facilities.

Our results indicate that an economy’s share in US trade and a military alliance with the US are the positive factors that determined access to a Fed swap agreement. On the determinants of the size of access to liquidity arrangements, we find that an economy with high levels of US bank exposure and stronger trade ties with the US tended to have greater access to Fed liquidity, via swap lines or FIMA. Economies with a large share of global trade, regardless of whether they are major trading partners of the US, also had greater access to US dollar liquidity via the Fed.

The access to US dollar liquidity lines was largely determined by pre-existing characteristics of the economies. However, the demand for US dollars auctioned by central banks should reflect the ongoing stress in the domestic currency and financial markets. We find that economies that faced appreciation pressures against the US dollar and whose local currency exchange rate became more volatile were more likely to auction greater amounts of US dollars in its domestic market. These effects were stronger for countries to which US had greater trade and financial exposure. While this may sound paradoxical, the record of the GFC is that from gold to the US dollar to the Japanese yen, ‘safe havens’ were greatly sought – and greatly affected – by investors reacting to the global financial crisis. Thereby, the appreciation and the volatility of the yen happened at the same time as the collapsing export/GDP of Japan during that time.3

As for the effects of the swap-related announcements, we find that these announcements led to appreciation of the currencies against the US dollar and reduced the size of the cross-currency basis. These announcements had no meaningful impact on the 10-year government bond yields, or CDS spreads. Dollar auctions by smaller central banks with Fed facilities did not have significant domestic effects, but dollar auctions by major central banks (BoE, ECB, BoJ and SNB) led to short-term appreciation of other, non-major currencies against the US dollar, reduced the size of the cross-currency basis and persistently reduced long-term government bond yields. These results do not change by these economies’ financial or trade exposure to the US. Nor do they change by the pre-existing financial vulnerabilities of the countries, measured by the foreign exchange exposure in the country’s external debt.

Putting our results in the proper context, the conceptual differences between international reserves and emergency swap lines and emergency liquidity arrangements are noteworthy. The demand for international reserves is shaped by precautionary motives, i.e., ex ante insurance against future sudden stops and trade funding challenges; and possibly concerns about trade competitiveness, dubbed sometimes as mercantilist motives (Aizenman, 2008). In contrast, emergency liquidity arrangements and swap lines provided during the GFC and the COVID-19 crisis provide liquidity relief at the time of crisis. Hence, there is no presumption that the demand for international reserves should be explained by the same factors determining emergency liquidity and swap lines. At the times of financial crisis, trade credits supplied by local banks may shrink even in economies experiencing currency appreciation, as banks may opt to reduce their risk exposure, documented by Amiti and Weinstein (2011).4 During a deep crisis with uncertain duration, like the GFC, economies may use their international reserves as a first line of defense, but may refrain from tapping too deeply to reserves out of the fear that depleting reserves may cost them dearly if the crisis would deepen in the future (Aizenman and Sun, 2012). In these circumstances, swap lines and FIMA type of arrangements may mitigate the decline of trade credit, the depletion rate of international reserves, and mitigate the rise of sovereign spreads of exposed emerging markets.

Boissay et al. (2020) indicate that the forces of these factors have only increased after the GFC due to the dominant role of dollar funding, and the growing depth of the global supply chains. Specifically, they noted:

“Trade finance, as proxied by the share of cross-border factoring in total factoring, has steadily increased over the past two decades. This long-term trend has gone hand in hand with the rise in international trade and, more specifically, the lengthening of global value chains (GVCs). While the lengths of domestic production chains estimated using world input–output tables at the country-sector level have remained constant, GVCs involving multiple border crossings have lengthened significantly between 2000 and 2017. Since financing needs increase with the length of supply chains (Bruno et al (2018), Bruno and Shin (2019)), trade finance has become more prominent in the context of GVCs.” (page 5). “The sharp appreciation of the dollar in the early stages of the Covid-19 crisis may have had knock-on effects to trade finance from stress in the banking system. Given the prevalence of the US dollar in trade financing, mitigating the impact of dollar credit fluctuations will be an important component of shielding global value chains from the pandemic’s economic fallout. In this respect, the recent expansion of central bank dollar swap lines and other measures to mitigate dollar liquidity conditions are likely to further cushion trade finance.” (page 6).

Baldwin and Freeman (2020) also highlight the magnification of the intermingling of trade and finance associated with the of the GVC deepening. The COVID pandemic vividly illustrated the dependence of the US and the EU on the GVC as the source of critical medical supplies. These considerations imply that we should take the econometric significance of the ‘trade factors’ in our regressions with a grain of salt: one needs micro data to identify more sharply the role of trade, finance and the interaction between the two.5

The results of our paper also are in line with Gourinchas and Rey, 2007, Obstfeld et al., 2009, and Gopinath et al. (2020), analyzing the “exorbitant privilege” position of the US dollar, and the dominant currency paradigm. These considerations suggest that the Fed’s swap lines and emergency liquidity provisions of the FIMA type may impact most emerging and developing economies, including nations with limited trade and financial dealings with the US. For economies with greater trade and financial integration with the US, these effects tend to be more direct, for others, in the form of spillovers.

The paper is organized as follows. In Section 2, we first review the background of the Fed’s liquidity lines and the US dollar auctions. In Section 3, we reexamine what factors led the Fed to select nine countries as swap partner countries as it did during the GFC. We then investigate the determinants of the size of liquidity lines – both swap and FIMA- in Section 4. In Section 5, we examine the determinants of the size of dollar auctions by central banks. In Section 6, we examine the impact of the liquidity lines and dollar auctions by central banks on important financial variables. In Section 7, we present the results of numerous robustness checks. In Section 8, we make concluding remarks.

2. Background on the Fed’s liquidity lines, and the US dollar auctions

In this section, we provide an overview of the institutional background of the Fed’s liquidity provision policies and the US dollar auctions conducted by other central banks. Having institutional knowledge of the Fed’s facilities and the system where other central banks auction US dollar liquidity lines will help the reader better understand the findings from the empirical analysis.

The Fed has standing swap lines with five central banks: ECB, BoJ, BoE, BoC and SNB. For each central bank, there is no limit on the volume of swaps. On March 15, 2020, these central banks agreed to reduce the pricing and extend the maturity of swap operations, and on March 20, 2020 they increased swap operations from weekly to daily frequency. On March 19, 2020, the Fed reactivated the swap lines it had established with nine central banks at the time of the Global Financial Crisis (GFC) of 2008, doubling their maximal lines (Table 1 ).6 The maturity of the swap can be between 1 day to up to 3 months.

Table 1.

Chronology of Fed swap lies and FIMA announcements.

Date	Policy Actions	Relevant Central Banks
March 15	Reduced Swap Pricing	Bank of Canada (BoC), Bank of England (BoE), Bank of Japan (BoJ), European Central Bank (ECB), Swiss National Bank (SNB)
	Introduced 84-day operations	BoE, BoJ, ECB, SNB
March 19	Reactivate temporary swap lines with nine other central banks	Reserve Bank of Australia (RBA), Banco Central do Brasil (BCB), Bank of Korea (BoK), Banco de Mexico (BdM), Monetary Authority of Singapore (MAS), and Sveriges Riksbank (SR) with the maximal lines of $60 billion. Danmarks Nationalbank (DNB), Norges Bank (NB), and Reserve Bank of New Zealand (RBNZ) with the maximal lines of $30 billion.
March 20	Increased frequency of one-week operations to daily	BoC, BoE, BoJ, ECB, SNB
March 31	Established FIMA repo facility	Foreign International Monetary Authority (FIMA) account holders

Notes: Bank of Canada, Bank of England, Bank of Japan, European Central Bank, Swiss National Bank have standing swap arrangements with the Fed. There is no limit to the swap amount for these central banks.

The Fed announced the establishment of the FIMA repo facility on March 31, 2020. To obtain US dollar liquidity through this facility, FIMA account holders had to apply for the use of the facility, and once approved, they could enter into repo agreements with the Fed. The Hong Kong Monetary Authority announced on April 22, 2020, the launch of a US dollar liquidity facility based on US dollars obtained via FIMA. In addition, central banks of Colombia, Chile and Indonesia announced on April 20, April 8, and June 24, respectively, that they had gained access to the FIMA facility.

Under the FIMA repo facility, the Fed supplies overnight US dollar liquidity in exchange for existing US Treasuries held by these institutions, and the transaction can be rolled over. The FIMA facility was expected to reduce the need for the sale of US Treasuries and mitigate pressure in this market.7 As a facility collateralized by US treasures (rather than foreign currency as in the case of swap lines), FIMA does not expose the US Fed to credit risk from the foreign central bank.

The Fed facilities appear to have been successful in limiting the sale of US Treasuries - economies with access to Fed facilities on average saw smaller percentage declines in holdings of US treasuries between February and April 2020 (Fig. 2).

The FIMA facility has ambivalent characteristics from the perspective of other economies. On the one hand, it is egalitarian because the facility is accessible not only to economies with standing swap lines or bilateral repo arrangements with the Fed, but also to economies which do not have any such agreements, or whose banks do not have direct access to the Fed’s lending window through their US subsidiaries or branches. It allows central banks to provide dollar liquidity to their markets without necessarily reporting a decline in FX reserves and potentially increasing risk perceptions for their currency.8 On the other hand, liquidity through FIMA is not equally accessible. It is more easily available for those economies which already hold a large volume of US Treasuries.

US dollar liquidity was actively provided to local markets through US dollar auctions by several central banks. The dates and terms of these auctions were pre-announced, including for those conducted by central banks other than major central banks.9 The auctions by the four major central banks - ECB, BOJ, SNB, and BOE - peaked in mid-March at about 112 billion US dollars per day and were the orders of magnitude larger than those by all other central banks (Fig. 3 ). The auctions by the other 13 central banks that published data on these auctions, used either US dollars obtained via swap lines with the Fed or their own foreign exchange reserves. Assuming that countries used only dollars obtained via the Fed for all auctions after they gained access to Fed liquidity facilities, we find that nine central banks auctioned up to 8 billion US dollars a day using Fed facilities, peaking at the end of March 2020. Some of these nine central banks conducted US dollar auctions prior to gaining access to Fed facilities, while other central banks (e.g.: India, Georgia) conducted auctions without access to Fed facilities.10 These auctions are shown in the bottom panel of Fig. 3, and peaked in mid- March, at about 23 billion USD per day. The US dollar shortage was mitigated by late June, so was the demand for US dollar liquidity from major central banks.

Fig. 3.

Many advanced and emerging-market central banks auctioned US dollars during the COVID-19 crisis Note: Auctions by major central banks are those conducted by BoE, BoJ, ECB, SNB. Auctions by other CBs with Fed facilities are those conducted by Australia, Denmark, Hong Kong (China), Indonesia, Korea, Mexico, Norway, Singapore, Sweden after the announcement of the Fed swap line or announcement of approval of access to these economies under FIMA. Auctions by other CBs without Fed facilities include those conducted by central banks of Chile, Colombia, Georgia, India, and Indonesia (prior to approval of FIMA access if relevant).

3. Probability of gaining access to Fed swap lines

The Fed reactivated swap lines with the same nine central banks with which it had established them in 2008. However, considering the difference in the economic conditions between the GFC and the COVID crisis as well as the difference in terms of the type of shock that hit the world economy (i.e., financial crisis originating in the US vs. the pandemic), the determinants for the selection of the nine central banks could differ between the two crisis episodes.

Hence, it is worthwhile to examine what factors led the nine central banks to be selected to receive swap liquidity lines from the Fed again. Following Aizenman and Pasricha (2010), we estimate the factors that affect the probability of economies included in the swap agreements. The candidate factors are: exposure of US banks to individual economies (BankExp), measured by the share of the individual market in the consolidated foreign claims of US banks; trade exposure to the US (TradeShare), measured by the share of economy i in total US goods imports and exports; financial openness (KAOPEN) measured by the Chinn-Ito index of capital account openness (Chinn and Ito, 2006, Chinn and Ito, 2008); and the existence of formal military relationship with the US (ALLIANCE).11

By extending US dollar liquidity to economies where US banks have greater asset exposure, the Fed could prevent costly deleveraging, possibly enhancing the welfare of both source- and recipient-economies. The more trade economy i conducts with the US, the greater the incentive of the Fed to secure trade flow by readily granting liquidity lines to the economy. A more financially open economy might have more need of liquidity lines, as the size of financial shock it faces may be larger due to its greater integration with global financial system. A non-economic factor may matter: the US may prefer providing swap lines with economies it has closer geopolitical relationship.

We apply the probit estimation model to a sample of 65 economies, based on data availability. This sample does not include Canada, the euro-area countries, Japan, Switzerland, and the United Kingdom because these countries have standing swap lines with the Fed. The dependent variable is a dummy variable that takes the value of unity when the economy of concern is selected as the recipient of swap lines. All the explanatory variables are sampled as of the end of 2019.

Columns of (1) through (4) of Table 2 show that all the candidate variables are significant contributors to selection into Fed’s swap agreements when they are included alone. Greater US banks’ exposure, greater US trade exposure, more open financial markets, and the existence of military alliance increase the probability of receiving a Fed swap line. Among the three variables, the variable for financial exposure of US banks alone explains the probability of providing a swap line the most (pseudo R² of 36%).

Table 2.

Probit regressions for explaining inclusion in Fed Swap arrangements.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Financial exposure	124.18				232.19			219.38
of US banks	(31.46)***				(124.30)*			(110.33)**
Share in US trade		12.28				18.26	19.85
		(6.71)*				(7.42)**	(8.08)**
De jure financial			1.63				1.09	1.74
openness			(0.70)**				(0.92)	(1.50)
Dummy for				1.99	3.95	2.33	2.05	3.47
military alliance				(0.48)***	(2.20)*	(0.58)***	(0.59)***	(2.14)
Constant	−1.80	−1.24	−2.23	−1.78	−4.62	−2.25	−2.88	−5.52
	(0.31)***	(0.22)***	(0.58)***	(0.32)***	(2.36)*	(0.47)***	(0.78)***	(2.63)**
N	65	65	63	65	65	65	63	63
pseudo R2	0.39	0.07	0.13	0.36	0.73	0.49	0.51	0.76

Notes: * p < 0.1; ** p < 0.05; *** p < 0.01. The figures reported in the table are estimated coefficients, not the marginal effects.

Columns (5) through (8) include the dummy for military alliance and one or two of the remaining three variables. Among these models, the model in column (8), with US financial exposure and de-jure financial openness of the economy along with the military alliance dummy, yields the highest pseudo R², though the estimate on financial openness becomes statistically insignificant and that on the military alliance dummy becomes marginally significant (with the p-value of 10.5%). Overall, both financial exposure and trade exposure and the existence of military alliances contribute positively to the probability of selection for the swap arrangement.

These probit models are good at predicting the probability of a swap line being extended. If we interpret a predicted probability of selection of 15% or more as a selection prediction, models (5), (6), (7), and (8) correctly predict 8 or 9 out of 9 swap arrangements and 48–50 out of 54–56 cases where such arrangements were not made.12

The impacts of financial exposure of US banks, trade exposure to the US, and military alliance are not only econometrically significant but also economically significant. Based on the estimation results of Model 8, if the level of US banks’ exposure rises by one standard deviation, the probability of getting selected for the swap arrangement would go up by 6.2 percentage points. According to Model 7, a one standard deviation increase in the level of trade exposure to the US, would increase the probability of getting selected for the swap arrangement by 7.4 percentage points. US military allies enjoy a higher probability of selection for a swap arrangement than that of non-ally countries, and the difference in the probability is as high as 48.2%.

Our findings appear to be different from Aizenman and Pasricha (2010) with respect to the importance of US trade exposure. Aizenman and Pasricha (2010) found that US bank exposure was a significantly positive contributor to the selection of emerging markets for swap agreements, but that US trade exposure was not. One reason for this difference may be that the financial crisis originated in the US and threatened the solvency of banking institutions. Therefore, the swap lines then were aimed at providing liquidity lines to the economies where US bank exposure was high. During the COVID-19 crisis, despite large-scale falls in the stock markets and widespread dollar shortage, there was no acute systematic financial instability as during the GFC. However, during the COVID-19 crisis, global supply chains got severely disrupted. Several papers have highlighted the increasing importance of global value chains in the past two decades and the associated intermingling of trade and finance (Boissay et al., 2020, Baldwin and Freeman, 2020). The COVID pandemic vividly illustrated the dependence of the US and the EU on the GVC as the source of critical medical supplies, and the sharp appreciation of the dollar in the early stages of the Covid-19 crisis may have had knock-on effects to trade finance from stress in the banking system. These considerations imply that we should take the econometric significance of the ‘trade factors’ in our regressions with a grain of salt: one needs micro data to identify more sharply the role of trade, finance and the interaction between the two.

4. Determinants of the size of liquidity lines via swap agreements and FIMA

While the number of economies that have swap agreements with the US is limited, the newly created FIMA facility is available to economies which do not have any swap agreements as long as they hold stocks of US Treasuries in their foreign exchange reserves which they can swap for dollars through the repo transactions. Therefore, the sum of the maximal amount of swap agreement with the Fed and the amount of US Treasuries in the foreign exchange reserves holding of an economy can be regarded as the potential availability of liquidity lines from the Fed. We estimate the determinants of the total size of liquidity lines available.

The size of liquidity lines available to economy i, $y_i^{},$ is composed of swap lines, ${\mathit{SWAP}}_i^{}$, and the FIMA line, ${\mathit{FIMA}}_i^{}$, namely,

$$y_i^{}={\mathit{SWAP}}_i^{}+{\mathit{FIMA}}_i^{}.$$ (1)

We investigate the determinants of the available amount of liquidity lines using the estimation model below:

$$y_i^{}=\alpha +X_i^{}\beta +{\epsilon }_i$$ (2)

where $y_i$ is the available amount of dollar liquidity and X_i is a vector of candidate explanatory variables.

As for the size of the swap lines, SWAP_i is straightforward because it is stipulated in the agreements. We do not include the economies of the following five central banks: the BOC, BOE, BOJ, ECB, and SNB, in the estimation because there is no limit to the swap lines for these central banks.

Measuring the size of liquidity lines through the FIMA facility is not straightforward because there is no stipulation about the available amount of liquidity lines through the facility. Instead, the FIMA repo facility allows foreign central banks and other international monetary authorities to temporarily exchange US Treasuries held in their accounts at the NY-Fed for US dollars. Hence, the amount of dollar-denominated reserve assets is a good proxy for the dollar liquidity available to economy i.

To measure US dollar-denominated reserve assets, we use a dataset Ito and McCauley (2020) compiled on the shares of key currencies in foreign exchange reserves for about 60 economies in the 1999–2018 period.13 In order to obtain the holdings of dollar-denominated reserve assets, as an approximate of the availability of liquidity lines through FIMA, we multiply the dollar share in foreign exchange reserves as of 2019 from the Ito-McCauley dataset, with total reserves (minus gold) as of 2019.

The candidate factors that may affect the availability of liquidity lines include the variables we tested in the previous exercise, namely, US bank exposure, the share of trade with the US, military alliance, and de jure financial openness. From the perspective of the US, it is rational to build a system that can provide liquidity lines for the economies that have more loans from US banks or that have more trade with the US.

In addition to these variables, we also test the variables for global financial exposure and the share of the economy in world trade. The rationale to test these variables is that the US might be willing to provide liquidity lines for the economies that do not necessarily have stronger financial or trade ties with the US per se. In other words, the US could be “altruistic” and willing to play the role of the guardian of the international economy by making dollar liquidity readily available through swaps or FIMA facility for large players in international finance or trade. We measure the level of global financial exposure (Global_Fin_i) with economy i’s sum of external assets and liabilities as a ratio to the world’s external assets and liabilities over the period of 2005 through 2015, using the database compiled by Lane and Milesi-Ferretti, 2001, Lane and Milesi-Ferretti, 2007, and (Lane and Milesi-Ferretti, 2017). While a significantly positive estimate on BankExp would imply that the US is more self-interested to make dollar liquidity readily available for economies with more consolidated claims of US banks, a significantly positive estimate on Global_Fin_i would mean that the US is willing to make dollar liquidity available for economies playing a larger role in global finance. Parallel to Global_Fin_i, the variable of global trade exposure (Global_Trade_i), measured by the share of economy i’s sum of exports and imports in global trade (i.e., the world’s sum of exports and imports), would show whether or not the US is altruistic to maintain global trade order by providing liquidity to economies that have larger presence in international trade.

We also test the dummy for China since the country’s holding of dollar-denominated assets is by far larger than the rest of the sample economies. As in the previous exercise, all the explanatory variables are sampled as of the end of 2019. We run a cross-sectional OLS regression for 52 economies.

The results presented in Table 3 indicate that higher US bank exposure to an economy led to a greater liquidity from the swap agreement or the FIMA repo facility. Interestingly, the variable for US bank exposure alone explains 58% of the variation of the availability of dollar liquidity. Also, an economy with greater trade exposure to the US tends to have greater access to dollar liquidity, explaining an additional 2% of the variation of dollar liquidity availability.

Table 3.

Determinants of size of available liquidity lines.

	(1)	(2)	(3)	(4)	(5)	(6)
Financial exposure of	10.86	9.72	9.78	5.39	6.04	6.04
US banks	(1.29)***	(1.40)***	(1.44)***	(1.00)***	(1.27)***	(1.28)***
Share in US trade		0.82	0.81	−0.49
		(0.44)*	(0.45)*	(0.31)
D. for military alliance			−0.00	−0.03	−0.03	−0.03
			(0.02)	(0.01)***	(0.01)**	(0.01)**
Share of trade in world				9.38	9.26	9.26
				(1.04)***	(1.10)***	(1.11)***
Share of finance in world					−3.38	−3.36
					(3.07)	(3.16)
De jure financial						−0.00
openness						(0.02)
Constant	0.02	0.02	0.02	0.00	0.00	0.00
	(0.01)*	(0.01)*	(0.01)	(0.01)	(0.01)	(0.01)
N	52	52	52	52	52	52
Adj. R2	0.58	0.60	0.59	0.85	0.84	0.84

Note: * p < 0.1; ** p < 0.05; *** p < 0.01.

Columns (4) through (6) also indicate that the US is altruistic in the sense that it makes dollar liquidity available for global major trading centers regardless of whether those economies are major trading partners with the US. In fact, when both the variable for the share in US trade and the one for the share of trade in world are included (column (4)), only the latter enters the estimation model with a significantly positive estimate. Also, including the variable for the share of trade in world increases the adjusted R² by 26%.

However, the share in global finance does not matter for liquidity access, possibly implying that an economy with a high level of liquidity access does not rely on USD liquidity provided by the US. The extent of financial openness does not matter for the size of dollar-liquidity available.

Although we use the product of total reserves (minus gold) and the USD share in foreign exchange reserves as of 2019 as an approximate of the availability of liquidity lines through FIMA, in fact, 56% of USD reserves are US Treasuries.14 Strictly speaking, the approximate of available liquidity lines should reflect this. Hence, we multiply the USD-denominated assets with 56% and rerun the regression with the modified proxy for the available liquidity lines $(y_i^{\ast }={\mathit{SWAP}}_i^{}+{\mathit{FIMA}}_i^{\ast })$ as a robustness check.

The estimations with the modified proxy for the available liquidity lines yield results that are essentially the same as those in Table 3, indicating the estimates we found in Table 3 are robust. The only difference is that the estimate of the variable for the share of finance in the world turns out to be significantly negative. That suggests that economies with great presence in global finance can find other means to get liquidity than holding dollar-denominated reserve assets, for example, that banks in these jurisdictions may have access to dollar liquidity through their operations in US.

These findings suggest that, as we found in the previous exercise, US bank exposure plays an important role in terms of determining which economies have better access to dollar liquidity. However, while the share in trade with the US per se does not matter, whether the economy of concern is a major player of international trade globally is an important factor. That also suggests the US is altruistic and tries to act as a guardian of international trade.

5. Determinants of the size of US dollar auctions by central banks

The actual use of liquidity lines may differ from the availability of liquidity lines. Both the availability and the use of US dollar liquidity may affect market perceptions and conditions, which we will investigate in Section 6. In this section, we investigate what factors, especially those in the domestic markets, determine the amount of US dollars auctioned (i.e. accepted) by central banks that had access to the Fed liquidity lines.15 The amount of dollars auctioned was largely determined by market conditions.16

We regress the amount of US dollars auctioned on a set of candidate variables. They include: cumulative depreciation (Cuml_fx), which is the rate of change in the local currency value per dollar with respect to the average local currency value during the last two weeks of January 2020; exchange rate volatility (fx_vol), the standard deviations of the rate of depreciation over rolling 14-day windows; stock market volatility (stk_vol), the standard deviations of the rate of stock market total return over rolling 14-day windows; the level of financial instability measured by the Chicago Board Options Exchange Volatility Index (VIX) in log; and US Bank exposure (BankExp) to these economies as of 2019Q4. We also test the variable for the number of new cases of COVID infection (COVID) measured by the moving average coronavirus infection cases per million over 7 days, and the mobility index (MOBILITY) to capture the possible impact of economic activities on daily basis. The estimation model also includes the interactions of US bank exposure with cumulative depreciation and exchange rate volatility. For this estimation, we apply an OLS estimation model to the daily panel data of 51 economies for the period of January 15, 2020 through May 29, 2020.17

In models 1 and 2 reported in Table 4 , the estimate on the cumulative depreciation rate is significantly negative, suggesting that central bankers facing their currency appreciating against the US dollar compared to the beginning of January 2020, auctioned more US dollars in their domestic markets. The estimate on the variable for exchange rate volatility is significantly positive, implying when their currency’s value is unstable, central bankers auction more US dollars. More global financial instability, which we measure with VIX, also increased the amount of US dollars auctioned by the monetary authorities.

Table 4.

Determinants of the size of US dollar auctions by central banks.

	(1)	(2)	(3)
Cumulative Depreciation Rate	−0.04	−0.01	0.00
	(0.01)***	(0.00)**	(0.00)
Exch. rate volatility	0.43	0.01	0.00
	(0.16)***	(0.04)	(0.05)
Stock market Volatility	−0.01	0.00	0.00
	(0.03)	(0.03)	(0.03)
VIX (in log)	0.94	0.50	0.58
	(0.30)***	(0.19)***	(0.33)*
US Bank Exposure, 2019Q4	7.13	−7.89	−3.93
	(1.51)***	(3.96)**	(4.53)
Cumulative Depreciation Rate		−2.10	−2.50
x US Bank Exposure		(0.71)***	(0.78)***
Exch. rate volatility × US Bank Exposure		30.82	29.96
		(9.77)***	(10.05)***
New cases, MA			0.82
			(0.51)
Mobility index, MA			0.00
			(0.00)
N	5457	5457	3693
Adj. R2	0.04	0.11	0.11
# of countries	51	51	49

Notes: * p < 0.1; ** p < 0.05; *** p < 0.01. MA refers to 7-day backward looking moving average. For the sake of presentation, the estimate of “New cases, MA” is multiplied by 1000.

In column (2), we add the interaction term between the cumulative depreciation rate and US bank exposure as well as the interaction term between exchange rate volatility and US bank exposure. The results indicate that the reaction to cumulative depreciation or exchange rate volatility is greater for economies to which US banks has greater exposure.

In model 3, we include the variable for the number of new COVID cases (COVID) and the mobility index (MOBILITY) to capture the possible impact of economic activities on daily basis, both of which are more directly related to the pandemic. Only the variable for new COVID cases is found to be a marginally significant factor (p-value = 10.8%), implying that a worsening of the pandemic situation may lead to a greater amount of US dollar auctions. While the variable for cumulative depreciation becomes insignificant, the estimates of the two interaction terms are robust to the inclusion of the two COVID-related variables.18

5.1. Correlations of the explanatory variables

The sample period of January 15, 2020 through May 29, 2020 is full of economic and non-economic events, which could make it harder to identify the estimates if the explanatory variables can be highly correlated with one another.

Online Appendix III presents the correlations of the variables used in the estimation on the size of US dollar auctions. From the table, we can see that the explanatory variables are not highly correlated with one another, except for the correlation between stock market volatility and VIX, which is to be expected. Overall, the estimates of the explanatory variables do not seem to be influenced by multicollinearity.

5.2. Estimations with the auction amounts with zeros being excluded

In this estimation exercise, the dependent variable takes the values of auctioned funds, otherwise zero when economy i’s central bank does not auction. As a robustness check, we also conduct the estimations only when the auctions take place (i.e., drop the observations of zero auctions). This shrinks the sample size substantially (by about 94%).

The estimation results remain intact (not reported). Naturally, the magnitudes of the estimates change, but the statistical significance of the estimates remain qualitatively unchanged, except for that the interaction term between exchange rate volatility and US Bank Exposure becomes insignificant.

5.3. Heckman two-step estimation

One may suggest that we should use the Heckman (1976) two-step selection model, in which the first step is the probit estimation of the probability of auctioning and the second step is the OLS to estimate the amount of the auctions. If the auctioning were done randomly, the Heckman two-step selection model would be more appropriate than a simple OLS estimation model, because we would only observe the auction amount when the auction takes place with a random probability. In such a case, applying an OLS estimation would lead the estimated coefficients to be inconsistent. However, as Table 1 makes it clear, the timing of auctions is pre-announced, i.e., it is not determined randomly by economic and non-economic conditions. It is rather determined mechanically. Hence, the Heckman estimation is not appropriate.

However, we still conduct the Heckman estimation analysis as a robustness check. In our context, we first estimate the probability of the central bank of country i deciding to auction the USD liquidity it has (i.e. through a swap agreement with the Fed) or it is capable of having access to (i.e., by collateralizing its U.S. Treasuries with the FIMA repo facility). Only when country i’s central bank auctions, do we observe the amount of auctioned funds (i.e., the dependent variable takes the values of auctioned funds, otherwise, the auction observation is regarded as missing. After running the probit estimation, we run the regression of equation (3) on the same set of candidate variables as in Table 4.

Online Appendix IV reports the estimation results. The top portion of the table reports the results of the second estimation (i.e., reaction equation). It shows that the estimation results are qualitatively intact, except for that the interaction term between exchange rate volatility and US bank exposure is no longer statistically significant.

6. The economic impacts of Federal Reserve liquidity lines

In this section, we investigate the impact of Fed liquidity facilities on four financial variables, namely, the exchange rate against the US dollar, 10-year government bond yields, credit default swap (CDS) spreads and the absolute cross currency basis (absolute deviations from covered interest parity), while controlling for an exhaustive list of other domestic and international economic conditions and policies.

We separately assess the impact of the announcements related to Fed facilities and the impact of the implementation of actual auctions. While providing dollar liquidity through a swap agreement or the FIMA facility can have economic impacts through alleviating dollar liquidity shortage, the mere announcement of these liquidity facilities can also have economic impacts because such a policy announcement would affect economic agents’ expectations and consequently their behavior (confidence channel). Further, as far as the impact of auctions is concerned, we separately examine the domestic and international spillover effects of US dollar auctions held by central banks. Finally, we examine whether the announcement or implementation effects differed by country characteristics. Specifically, we examine whether these effects were greater for the economies to which US banks had greater exposure, or for the economies that were more exposed to the US through trade or financial linkages, or economies that had greater balance sheet currency exposure.

6.1. Baseline model

To capture the size and the persistence of the impacts on economic variables, we employ the local projection method (Jordà, 2005) with the Teulings and Zubanov (2014) extension to a sample of 39 economies and run a series of regressions for different horizons, r = 0,1,2,…, p as follows: 19

$${\mathrm{\Delta }\mathit{Y}}_{i,t-1\to t+p}^{}={\sum }_{r=0}^p{\beta }_1^{p,r}{\mathit{S}\mathit{w}\mathit{a}\mathit{p}\mathit{F}\mathit{i}\mathit{m}\mathit{a}\mathit{A}\mathit{n}\mathit{n}}_{i,t+p-r}^{}+{\sum }_{r=0}^p{\beta }_2^{p,r}{\mathit{A}\mathit{u}\mathit{c}\mathit{t}\mathit{i}\mathit{o}{\mathit{n}}_{\mathit{O}\mathit{w}\mathit{n}}}_{i,t+p-r}^{}$$

$$+{\sum }_{r=0}^p{\beta }_3^{p,r}{\mathit{A}\mathit{u}\mathit{c}\mathit{t}\mathit{i}\mathit{o}{\mathit{n}}_{\mathit{M}\mathit{a}\mathit{j}\mathit{o}\mathit{r}}}_{i,t+p-r}^{}+{\sum }_{r=0}^p{\beta }_4^{p,r}{\mathit{O}\mathit{t}\mathit{h}\mathit{e}\mathit{r}\mathit{P}\mathit{o}\mathit{l}\mathit{i}\mathit{c}\mathit{i}\mathit{e}\mathit{s}\mathit{a}\mathit{n}\mathit{d}\mathit{V}\mathit{I}\mathit{X}}_{i,t+p-r}^{}$$

$$+{\sum }_{l=1}^3{\gamma }_1^l{\mathit{S}\mathit{w}\mathit{a}\mathit{p}\mathit{F}\mathit{i}\mathit{m}\mathit{a}\mathit{A}\mathit{n}\mathit{n}}_{i,t-l}^{}+{\sum }_{l=1}^3{\gamma }_2^l{\mathit{A}\mathit{u}\mathit{c}\mathit{t}\mathit{i}\mathit{o}{\mathit{n}}_{\mathit{O}\mathit{w}\mathit{n}}}_{i,t-l}^{}+{\sum }_{l=1}^3{\gamma }_3^l{\mathit{A}\mathit{u}\mathit{c}\mathit{t}\mathit{i}\mathit{o}{\mathit{n}}_{\mathit{M}\mathit{a}\mathit{j}\mathit{o}\mathit{r}}}_{i,t-l}^{}$$

$$+{\sum }_{l=1}^3{\gamma }_4^l{\mathit{O}\mathit{t}\mathit{h}\mathit{e}\mathit{r}\mathit{P}\mathit{o}\mathit{l}\mathit{i}\mathit{c}\mathit{i}\mathit{e}\mathit{s}\mathit{a}\mathit{n}\mathit{d}\mathit{V}\mathit{I}\mathit{X}}_{i,t-l}^{}{+\beta }_4^{}{\mathit{O}\mathit{t}\mathit{h}\mathit{e}\mathit{r}\mathit{C}\mathit{o}\mathit{n}\mathit{t}\mathit{r}\mathit{o}\mathit{l}\mathit{s}}_{i,t-1}^{}+{\mu }_i+w_s+dow+{\epsilon }_{i,t+p}^{}$$ (3)

${\mathrm{\Delta }\mathit{Y}}_{i,t-1\to t+p}^{}$ is the cumulative percentage point change from t-1 to t + p in economy i’s currency per US dollar; economy i’s 10-year sovereign bond yield; the sovereign 5-year credit default swap (CDS) spread for economy i; or absolute cross-currency basis for economy i’s currency against US dollar.

In order to identify the announcement effect, we create a dummy variable that is economy-specific: SwapFimaAnn takes the value of one for swap recipients on the days when the Fed announces swap policies; on the day when economy i announced it had gained access to FIMA; and on the days when the auctions of dollars were announced by the central bank of economy i that has access to Fed facilities. We collected the data on auction announcements from the central banks’ websites. Creating a dummy that takes the value of one only for the swap recipient economies on a particular day rather than for all the economies should allow us to identify whether the announcement affected the relevant economy differently from the broad US dollar movement on that day.

In the baseline specification, Auction_Own is the auction amount accepted on day t by economy i’s central bank with access to Fed facilities. Auction_Major is the sum of the auction amounts accepted on day t by major central banks (ECB, BOE, SNB, and BOJ).20 Both Auction_Own and Auction_Major are in billions of US dollars. We collected data on the details of the auctions - amounts allocated, amounts bid, amounts accepted, maturity, etc. - from central banks’ websites. As the baseline specification does not include the economies of the major central banks, Auction_Major aims to capture the spillover effects of the auctions by these major central banks.21 In robustness checks, we redefine, Auction_Own and Auction_Major to use the weighted auction amounts (i.e., the auction amounts allocated based on the weights of the tenors of the auctions, as was done in Rose and Spiegel, 2012) and also use dummies for auction days instead of auction amounts.

We include an exhaustive list of other policy announcements in the vector Other Policies and VIX to isolate the impact of the variables of interest. The vector includes: (i) domestic economic policy announcements of fiscal policy, macroprudential policy, liquidity provision policy, asset purchase programs, and monetary policy from the Yale Program on Financial Stability22 ; (ii) a percentage point decline in the monetary policy rate; (iii) Rating: a variable that takes the value of +1 on days when any of the three major rating agencies (Standard and Poor’s, Moody’s, or Fitch) announced an upgrade in the country’s sovereign credit rating, −1 for the days with a downgrade, and 0 otherwise; (iv) the number of net tightening of lockdown policies announced (i.e., the number of lockdown tightening minus the number of lockdown easing), using the data from ACAPS23 ; (v) a dummy that takes the value of one on days of major US Fed announcements, including the January and early March FOMC decisions as well as all the major announcements from mid-March to end-May 2020, listed in the data appendix); (vi) VIX in natural log to control for global financial market conditions.

Finally, we include as Other Controls the 7-day moving average of new COVID cases per million, and the domestic stock market total returns (in %).

The estimation also includes economy-fixed effects, weekly fixed effects, and day-of-the-week fixed effects. The estimation is conducted for the business days of the sample period from January 15 through May 29, 2020.24 Economy-fixed effects control for time-invariant, or slowly changing, characteristics of each economy (e.g., the level of development of the medical and insurance systems, economies’ debt levels, the tendency for fiscal interventions). Weekly fixed effects control for common, global waves of the virus infections or related news, and day-of-the-week fixed effects control for seasonality in daily frequency. We employ the local projection method that estimates a series of regressions for each horizon p for each variable. Here, we assign p to be 5, that means we attempt to capture the full dynamics of each of the four dependent variables in the aftermath of policy announcements or actions by central banks over the course of five business days. Standard errors are robust. To focus on the dynamics of the economic variables in response to a policy shock, we focus on reporting the cumulative impulse response functions (IRFs) rather than discussing the estimated coefficients.

Our identification strategy proceeds as follows: (i) drop the economies of major central banks from the estimation; (ii) define the announcement dummies to be economy-specific which allows us to focus on differential impact on the relevant currencies and other variables of interest; (iii) use an exhaustive list of domestic and external controls; and (iv) base the estimation model on Teulings and Zubanov (2014), who argue that unbiasedness requires controlling for lead values of explanatory variables within the forecast horizon (i.e. t + 1 through t + p). Controlling for lead values is important in our specification because auctions by major central banks were held and subscribed on daily basis during the most stressful period in the sample, and also because many other policy announcements crowded that period. Controlling for these leads allows us to identify the individual effect of each announcement or auction.

Using the IRF based on the local projection model is more appropriate when the estimation involves nonlinear estimates. As we discuss later, we test whether the impact of swap and other stimulus policies varies with US trade or financial exposure with the recipient economies. To test this, we include interaction terms between US trade or financial exposure and the policy variable of our interest. While nonlinear VAR models exist, the estimation and interpretation of their estimates can be highly complex. With the local projection model, this kind of exercise can be done with ease.

6.2. Baseline results

The results of the baseline estimation in Fig. 3 show that the announcements of Fed liquidity lines or US dollar auctions had a significant and persistent impact on the exchange rates of the relevant currencies – they led to appreciation of these currencies against the US dollar. The announcements also led to a temporary compression of cross-currency basis, as intended. The estimation did not detect any significant announcement impact on the CDS spreads or 10-year government bond yields. US dollar auctions by central banks with Fed dollars did not have a significant impact on financial prices, but auctions by major central banks led to appreciation of other currencies against the US dollar and a reduction in the size of CIP differentials. They also persistently reduced the 10-year government bond yields. These findings suggest that the auctions by major central banks were an important tool in ameliorating global financial stress.

6.2.1. Announcement effects of Federal Reserve liquidity lines and auctions

The main purpose of Fed swap lines or the FIMA facility was to increase US dollar liquidity globally and stabilize foreign dollar markets. Panel (a) of Fig. 4 shows that on the days of swap-related announcements, the currencies of relevant economies appreciate against the US dollar relative to other economies. The cumulative appreciation is significant after day 2, and reaches 2 percent by day 5, consistent with the policy objective. The announcements also reduce the absolute size of the cross-currency basis, albeit temporarily (between days 1–4) for the relevant currencies (Fig. 4(b)).25 These results suggest that the announcements of the Fed liquidity facilities did mitigate the dollar shortage through the confidence channel.

Fig. 4.

Estimated effect of the announcements of the swap agreements and FIMA on financial variables Notes: The figure illustrates the accumulated impact on (a) the rate of deprecation of the home currency, (b) the absolute 3-month cross-currency swap basis, (c) the CDS spread (d) the 10-year government bond yields of the announcement of Fed swap lines, the approval of FIMA application or the announcement of US dollar auctions using Fed facilities for the relevant economies over others. The announcement takes place at t = 0, and the cumulative impulse response functions are illustrated for the next five business days.

The announcements did not have any significant effect on the sovereign CDS spreads for swap/FIMA recipients, The announcement effects were not found on the long-term interest rates of the recipient economies, or the effect was transient, compared to economies that were not beneficiaries of these announcements (Fig. 4 (c-d)).26

6.2.2. Domestic and spillover effects of US dollar auctions

With this estimation model, we can also examine the impact of central banks’ swap auctions on financial variables. We assess the impact on the economic variables of our interest of a one billion dollar increase in the amount auctioned by economy i’s own central bank/monetary authority with Fed liquidity lines. We also examine the effects of US dollar auctions by the four major central banks (i.e., BOE, BOJ, ECB, or SNB) on other economies’ financial variables.

The effects of US dollar auctions by non-major central banks, using Fed facilities, are not significant (Fig. 5 ). However, a one billion increase in the dollars auctioned by the major central banks led to an immediate appreciation of other economies’ currencies against the US dollar, with its impact persistently becoming larger through the third day. The major central bank auctions also led to a temporary narrowing of the absolute value of the covered interest differentials. Further, these auctions lead to an immediate and persistent decline in 10-year government bond yields. However, the magnitude of the effect on average was small - the cumulative effect on bond yields at day 5 is −0.003 percentage point for an additional 1 billion US dollars auctioned – at the peak of the auctions, the major central banks auctioned 112.1 billion US dollars in a single day. At peak, other things being equal, this would translate to a 0.3 percentage point decline in the average yields of the sample economies. These findings suggest that foreign exchange swap auctions by the major central banks are very effective and have implications beyond the home economies.

Fig. 5.

Estimated effects of US dollar auctions using Fed facilities Notes: The auction takes place at t = 0, and the cumulative impulse response functions are illustrated for the next five business days.

US dollar auctions do not affect the dynamics of the sovereign CDS spreads, except for a small and short-lived ameliorative impact of major central bank auctions on day 2.

6.3. Interactive analysis: Does US- or US-dollar exposure matter for impact of dollar auctions?

In this section, we explore whether the impact of US dollar auctions differ by country characteristics. How the US dollar swap or auction policies affect the dynamics of the economic variables of our interest can depend on the US exposure to these economies in terms of international trade or finance. We found in sections 3 and 4 that US financial exposure was an important determinant of the economies’ access to Fed liquidity. If countries to which the US was more financially exposed benefitted more from auctions, that could suggest that the liquidity policies were successful from a narrow US-oriented perspective. On the other hand, if the policies were in fact broadly useful, then even if it may seem that the Fed was acting from a narrow perspective, its actions were helping everyone. However, economies’ own exposure to the US could also matter. Rose and Spiegel (2012) find that the impact of swap auctions of dollar assets by foreign central banks is larger for economies that have more trade or financial linkages with the US. Finally, it is not just US exposure, but foreign exchange exposure that may be important - greater aggregate foreign exchange rate exposure of an economy, particularly in external debt, could lead to greater dollar shortage once the crisis starts. These countries could benefit more from US dollar auctions.

We therefore examine whether and to what extent the level of trade or financial linkage with the US, or that of US-dollar exposure influences the impact of Fed liquidity policies. The estimation model now includes the following interaction terms:

$${\mathrm{\Delta }\mathit{Y}}_{i,t-1\to t+p}^{}=\sum _{j=1}^3{\sum }_{r=0}^p{\beta }_j^{p,r}{\mathit{X}}_{i,t+p-r}^j+\sum _{j=1}^3{\sum }_{r=0}^p{\delta }_j^{p,r}{\mathit{X}}_{i,t+p-r}^j\bullet {\mathit{E}\mathit{X}\mathit{P}}_i$$

$$+\sum _{j=1}^3{\sum }_{l=0}^3{\gamma }_j^l{\mathit{X}}_{i,t-l}^j+\sum _{j=1}^3{\sum }_{l=0}^3{\theta }_j^l{\mathit{X}}_{i,t-l}^j\bullet {\mathit{E}\mathit{X}\mathit{P}}_i+\rho {\mathit{E}\mathit{X}\mathit{P}}_i+{\sum }_{r=0}^p{\beta }_4^{p,r}{\mathit{O}\mathit{t}\mathit{h}\mathit{e}\mathit{r}\mathit{P}\mathit{o}\mathit{l}\mathit{i}\mathit{c}\mathit{i}\mathit{e}\mathit{s}\mathit{a}\mathit{n}\mathit{d}\mathit{V}\mathit{I}\mathit{X}}_{i,t+p-r}^{}{\sum }_{l=1}^3{\gamma }_4^l{\mathit{O}\mathit{t}\mathit{h}\mathit{e}\mathit{r}\mathit{P}\mathit{o}\mathit{l}\mathit{i}\mathit{c}\mathit{i}\mathit{e}\mathit{s}\mathit{a}\mathit{n}\mathit{d}\mathit{V}\mathit{I}\mathit{X}}_{i,t-l}^{}{+\beta }_5^{}{\mathit{O}\mathit{t}\mathit{h}\mathit{e}\mathit{r}\mathit{C}\mathit{o}\mathit{n}\mathit{t}\mathit{r}\mathit{o}\mathit{l}\mathit{s}}_{i,t-1}^{}+w_s+dow+{\epsilon }_{i,t+p}^{}$$ (4)

where $X^j$ includes each of the key explanatory variables in the baseline model, i.e. SwapFimaAnn, Auction_Own, and Auction_Major. EXP is the exposure variable and takes one of five values: either BankExp (exposure of US banks to individual economies); USShareinTrade (the share of US in economy i’s exports and imports); gross portfolio exposure of country i to US using restated CPIS (Coppola et al., 2021); aggregate FX exposure (from Benetrix et al. (2019) with sign changed so that higher values indicate greater FX exposure in liabilities than assets); or FX exposure in external debt (higher values indicate greater vulnerability). Each of these variables are measured as of the end of 2019, or the latest available value prior to 2019, otherwise. The coefficient of interest is ${\widehat{\mathit{\delta }}}_{\mathit{j}}^{\mathit{p},\mathit{r}}$. If the extent of trade or financial exposure matters, this estimate will be significant with a sign that would enlarge the absolute magnitude of SwapFimaAnn or other policy shock variables (i.e., own, or major central bank auctions).

The responses of financial variables to Fed liquidity facilities/auction announcements or to own or major central bank auctions do not differ by US bank exposure to the relevant countries (Fig. 6 ). Fig. 6 (a) illustrates the interactive effects of swap-related announcements and bank exposure. That is, $\left({\beta }_1^{p,r}+{\delta }_1^{p,r}\bullet {\mathit{EXP}}_{i,t+p-r}\right){\bullet SwapFimaAnn}_{i,t+p-r}^{}$ when EXP is BankExp. To make the interpretation easier, we draw two Cumulative Impulse Response Function (CIRFs), one for the case where BankExp takes the 75th percentile value (blue line) and the other for the 25th percentile value (red line). From the two CIRFs and their confidence intervals, we can see that the extent of bank exposure does not affect the impact of SwapFimaAnn. That is consistent with the finding that ${\widehat{\delta }}_1^{p,r}$ is statistically insignificant. The impacts of swap auctions by either individual economies’ central banks or the major central banks do not depend upon bank exposure, either. The lack of statistical significance is not just the case for the impacts on the exchange rate, but also on the sovereign debt yield, the CDS spreads and cross-currency basis (Fig. 6(b) through 6(d)).Major central banks’ US dollar auctions led to appreciation of other currencies against the US dollar, reduced government bond yields and CIP deviations and had a small ameliorative impact on CDS spreads, regardless of the US banks’ exposures t these countries.

Fig. 6.

Interaction effects: US Bank Exposure.

The baseline results are similarly robust to including other interaction terms as well and do not differ for the different levels of exposures, with few exceptions.27 Economies with high financial exposure to the US (through portfolio investments) did not benefit from a decline in government bond yields in response to their central banks’ own auctions, whereas those with low gross exposure to US did (Fig. 7 (a)). The economies with high gross financial exposure to the US saw an increase in CDS spreads in response to own FX auctions, while those with low exposures did not (although the two cumulative impulse response functions are not significantly different from each other) (Fig. 7 (b)). Another substantive difference between the baseline results and the regressions with interaction terms is that when interactions with FX exposure or FX debt exposure are included, own FX auctions lead to an increase in CDS spreads for economies at both high and low exposures on days 4 and 5 (Fig. 8 (a)), whereas major central bank auctions continue to benefit all economies (Fig. 8(b)).

Fig. 7.

Interactive effects of Financial Exposure to USA.

Fig. 8.

Interactive effects of FX exposure on CDS spreads Notes: The announcement takes place at t = 0, and the cumulative impulse response functions are illustrated for the next five business days.

These estimation results give us an answer to the question of whether the effects of liquidity-providing policies differ depending on country characteristics. The answer is no for the most part. While the effect of own auction using Fed facilities depended in some instances on country characteristic, major central bank auctions had a broadly beneficial impact on financial conditions in other economies. This finding is different from that of Rose and Spiegel (2012). We conclude that US liquidity-providing policies are indiscriminatory. This is consistent with the claim that the US Fed is the lender of the last resort.

7. Robustness checks

The results presented in the previous section are robust to several changes in specification.28 First, when we increase the forecast horizon to 11 days, and then also increase the number of lags to 6. The results for the announcement effects and major central bank auctions remain robust and there are no sign reversals after day 5. The results on own central bank auctions also remain robust except that the CDS spreads on the economy’s sovereign debt rise after day nine.

Second, the estimation results are robust when we use alternative control variables that measure the severity of the pandemic. We have used total number of COVID-19 cases instead of new cases per million, controlled for changes in the level of mobility with respect to the pre-pandemic period. These variables are not significant, and do not affect the results for the variables of interest.

Third, we weighted the US dollar amount auctioned by central banks by maturity, as in Rose and Spiegel (2012) or used a dummy to represent the days when auctions took place rather than auction amounts. In another robustness check, we used the amounts auctioned by all central banks which disclosed auction data, not only the amounts auctioned using Fed liquidity lines, as the Auction_Own variable.29 The results remain unchanged from those described in section 6, with a few exceptions: when using the dummies for auction days rather than auction amounts, the impact of major central bank auctions on exchange rate is significant through day 5 while the impact on the government yields is no longer significant. When defining the Auction_Own variable to include the amounts of auctions by central banks without Fed facilities, the impact of major central bank auctions on absolute cross-currency basis is not significant anymore, while the other results remain intact.

8. Concluding remarks

The outbreak of the COVID-19 pandemic instigated a financial turmoil in March 2020. As in the previous episodes of financial instability, the uncertainty made investors rush to hold US dollar-denominated assets, creating a dollar shortage. To prevent this situation from morphing into a global systematic financial crisis, the Fed made US dollar liquidity readily available through reinforcing or reactivating central bank swap lines and creating the FIMA repo facility.

This paper investigates the motivations for and the effects of US dollar liquidity provision with the following questions: First, what factors led the Fed to select nine economies as swap partners as it did at the time of the GFC? Second, what factors determined the total availability of liquidity lines (swaps and the FIMA facility) from the Fed? Third, what domestic conditions determined the size of dollar auctions by central banks? Fourth, what were the announcement effects of the Fed liquidity arrangements? Fifth, what were the domestic and spillover effects of dollar auctions by central banks? Sixth, did the economic impacts of the US dollar liquidity provision differ depending on country characteristics?

We find that the Fed chose to reactivate swap agreements with nine economics because of these economies’ large trade and financial ties with the US. This result is in contrast with the previous episode of dollar shortage in 2008 when the US signed swap agreements with emerging economies due to their financial ties with the US only. The existence of formal military alliances was also a determinant for the Fed to reactivate the swaps for these economies. Economies with strong financial and trade ties with the US tended to have more access to dollar liquidity lines. Global major trading centers also had greater access to US dollar liquidity via the Fed, regardless of whether they had more financial or trade ties with the US. The dollar amounts auctioned by central banks were larger for currencies that faced greater exchange rate volatility, and when global financial conditions were more unstable, as captured by a higher VIX.

The announcements of swap-related policies had meaningful impacts on economic variables; they led to appreciation of the partner currencies against the US dollar, and a narrowing of the cross-currency basis. Dollar auctions by non-major central banks with Fed facilities did not have significant domestic effects, but dollar auctions by major central banks (BoE, ECB, BoJ and SNB) had spillover effects – they led to short-term appreciation of other, non-major currencies against the US dollar and reduced the size of the cross-currency basis and longer-term government bond yields.

One may conjecture that the announcement or implementation effects may be greater for the economies with greater links with the US or for those with greater foreign exchange exposure. However, our findings suggest that the liquidity facilities had egalitarian impacts.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

The following are the Supplementary data to this article: