BIS Bulletin
No 18
EME bond portfolio flows and long-term
interest rates during the Covid-19 pandemic
Peter Hördahl and Ilhyock Shim
20 May 2020
BIS Bulletins are written by staff members of the Bank for International Settlements, and from time to time by other
economists, and are published by the Bank. The papers are on subjects of topical interest and are technical in
character. The views expressed in them are those of their authors and not necessarily the views of the BIS. The
authors are grateful to Jimmy Shek for excellent analysis and research assistance, and to Louisa Wagner for
administrative support.
This publication is available on the BIS website ().
© Bank for International Settlements 2020. All rights reserved. Brief excerpts may be reproduced or
translated provided the source is stated.
ISSN: 2708-0420 (online)
ISBN: 978-92-9259-386-5 (online)
Key takeaways
Bond portfolio outflows from emerging market economies (EMEs) are typically associated with currency
depreciation and rising domestic long-term interest rates. This relationship asserted itself in a
particularly stark way during the Covid-19 crisis in mid-March 2020.
The relationship between bond portfolio outflows and long-term rates varies across EMEs, depending on
factors such as bond market depth, FX market functioning and sovereign risk. The impact of these factors on the
relationship has been thrown into sharper relief during the Covid-19 pandemic.
Recent policy responses, such as bond purchase programmes, duration swaps and efforts to stabilise
exchange rates, can play an important role in maintaining financial stability in EMEs when they face bond
outflows. Policy measures to develop deep and liquid bond markets and strengthen the resilience of local
currency bond and FX markets are likely to enhance market functioning in the longer term.
Peter Hördahl
@
Ilhyock Shim
@
EME bond portfolio flows and long-term interest rates during the
Covid-19 pandemic
Increasing worries on the part of global investors about the Covid-19 outbreak spreading around the world
sparked a sharp sell-off of assets in March 2020. Local currency bond markets in emerging market economies
(EMEs) experienced outsize portfolio outflows and huge spikes in long-term interest rates (eg Hofmann et al
(2020)). In particular, bond portfolio outflows from EMEs eclipsed those seen in previous episodes of market
turmoil, such as the Great Financial Crisis (GFC) of 2007–09 and the 2013 taper tantrum.
In recent years, financing through bond issuance has grown in importance relative to bank loans in EMEs.
Greater international participation in domestic bond markets has meant that foreign holdings of local currency
bonds have been on the rise. As a result, bond portfolio flows, and their effects on EME local currency government
bond yields, have become increasingly important for domestic monetary conditions.
In what follows, we focus on the relationship between EME bond fund flows and local monetary
conditions as reflected in domestic long-term interest rates, contrasting the dynamics during the Covid-19
pandemic with those during the pre-pandemic period. Given that flows and long-term rates are endogenous
variables, we cannot make claims as to specific directions of causality. It is, nevertheless, important to
understand how the relationship between these variables has changed during the crisis, and how it relates to various
country-specific factors.
In the analysis, we focus on government bond yields. This is motivated by the fact that government bond
prices play an important role in local financial conditions as a benchmark against which other assets, notably
corporate bonds and loans, are priced. We find that in the 10 years prior to the pandemic, bond outflows were
typically associated with rising bond yields in EMEs and inflows with falling yields. During the pandemic, this
relationship has intensified. We also find that the responsiveness of EME yields to bond flows varies across EMEs
depending on factors such as the bid-ask spread, foreign participation, exchange rate volatility and the sovereign
credit default swap (CDS) spread.
Our analysis suggests that policy actions taken during the pandemic are important in stabilising domestic
financial conditions in EMEs. These include central banks purchasing bonds, swapping long-term
mailto:@
mailto:@
securities for short-term securities via auctions and providing liquidity to domestic institutional investors.
Moreover, given the central role that EME exchange rates play as a determinant of financial conditions,
addressing currency fluctuations will help quell turbulence in domestic monetary conditions due to sharp changes in
long-term interest rates. In the longer term, policymakers’ efforts to develop and deepen local capital markets, such
as fostering a strong domestic institutional investor base, would result in better financial resilience and a greater
ability to withstand portfolio outflows from their bond markets.
Bond flows, long-term rates and exchange rates during market turmoil
Graph 1 compares the dynamics of cumulative EME bond outflows during the Covid-19 pandemic with those
observed during the height of the GFC and during the taper tantrum using two different measures: daily non-
resident portfolio bond flows reported by the Institute of International Finance (IIF) and weekly bond fund flows
reported by Emerging Portfolio Fund Research (EPFR).1 In both measures, the pandemic saw substantially larger
bond outflows than the two earlier episodes (first two panels). The bond fund outflows from EMEs over the first
nine weeks, for example, totalled close to $30 billion, ie three to six times those observed during the GFC or the
taper tantrum. However, the difference becomes much smaller if we normalise the dollar amount of bond outflows
by total market size. Normalised this way, the outflows during the pandemic have been around 10–50% larger than
those during the GFC or the taper
EME portfolio bond flows, bond yields and exchange rates during market turmoil Graph 1
Bond flows (IIF)1 Bond flows (EPFR)2 Change in sovereign yield3 Exchange rate change4
USD bn USD bn Percentage point First sample date = 100
1 2 3 4 5 6 7 8 9 10 w 1 2 3 4 5 6 7 8 9 10 w 1 2 3 4 5 6 7 8 9 10 w 1 2 3 4 5 6 7 8 9 10 w
Covid-19
EMEs
Taper tantrum
EMEs
GF
C EMEs
US US US
Covid-19: 4 March to 24 April 2020; taper tantrum: 27 May to 28 August 2013; Great Financial Crisis (GFC): 15 September to 19 December
2008.
1 Cumulative weekly portfolio debt flows to eight EMEs tracked by the IIF: Hungary, India, Indonesia, Mexico, Poland, South Africa, Thailand and
Ukraine 2 Cumulative weekly EME local currency bond fund flows tracked by EPFR. 3 Cumulative weekly change in five-year local currency
sovereign bond yield. Simple average across 19 EMEs used for analysis in the next sections of this Bulletin. 4 Cumulative weekly change in bilateral
dollar exchange rate. Simple average across 19 EMEs used for analysis in the next sections of this Bulletin.
Sources: Bloomberg; EPFR; IIF; authors’ calculations.
1 The IIF flows measure net sales and purchases of bonds by non-residents for a sample of EMEs, based on daily or weekly data from
national central banks, government agencies and securities exchanges. EPFR instead reports investor flows to mutual funds and
exchange-traded funds. Various studies report that variations in EPFR flows closely match those in the much broader (but much less
frequent) balance of payments data, which the IIF data aim to capture.
2 The ratio of the cumulative bond outflows from seven EMEs (Hungary, India, Indonesia, Mexico, Poland, South Africa and Thailand)
tracked by the IIF to the total size of their domestic bond markets from the BIS debt securities statistics was %, % and %
during the GFC, taper tantrum and Covid-19 episodes, respectively. In addition, the ratio of the cumulative bond fund outflows from all
EMEs tracked by EPFR to the total size of all EME domestic bond markets from the same BIS source was %, % and %
during the three aforementioned episodes, respectively.
–40
–30
–20
–10
0
–40
–30
–20
–10
0
–
–
80
85
90
95
100
As these outflows occurred, EME long-term interest rates surged and EME currencies depreciated (Graph 1,
last two panels). In the first half of March 2020, five-year EME government bond yields rose by almost 100 basis
points on average. By contrast, US Treasury yields fell during the same period. The yield increase among EMEs
was similar to that seen after the Lehman collapse and during the taper tantrum, but the peak was reached
substantially quicker than in the two previous episodes. However, EME long- term interest rates also retreated
faster this time around than in the previous crises. This is likely to be due to aggressive policy actions taken by EMEs,
including rate cuts, provision of liquidity to financial institutions and announcements of bond purchase programmes.
Moreover, the expansion or reintroduction of Federal Reserve swap lines to central banks, the Fed’s repurchase
agreement facility for foreign and international monetary authorities and other actions to provide US dollar
liquidity globally helped stabilise EME exchange rates and bond portfolio flows, thereby easing the pressure on
interest rates (Avdjiev et al (2020)).
Bond flows and long-term interest rates before and during Covid-19
To what extent are bond flows systematically associated with monetary conditions as captured by changes in
domestic long-term interest rates? Graph 2 shows scatter plots of weekly changes in five-year domestic bond yields
against contemporaneous standardised bond Although both yield changes and bond flows are noisy, a
pattern emerges of a negative relationship for EMEs in the pre-crisis sample (blue dots and lines): bond outflows
(inflows) are associated with bond yield increases (declines). During the Covid- 19 crisis, the yield-flow
relationship has become more pronounced across all EMEs (red dots and lines).
Long-term interest rates and bond flows in EMEs Graph 2
Emerging Asia1 Latin America2 Other emerging market economies3
y = –– 2
where R2=
1
0
y = ––
1
where R2=
–2
y = – 2
where R2=
1
0
y = ––
1
where R2=
–2
y = –– 2
where R2 =
1
0
y = ––
1
where R2 =
–2
–20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10
Bond flow (standardised) Bond flow (standardised)
Crisis Pre-crisis
Bond flow (standardised)
“Pre-crisis” refers to the period from January 2010 to February 2020, while “Crisis” refers to March 2020. Country bond flows are standardised by their
standard deviation.
1 China, Chinese Taipei, India, Indonesia, Korea, Malaysia, the Philippines and Thailand. 2 Brazil, Chile, Colombia, Mexico and Peru. 3 The Czech
Republic, Hungary, Israel, Poland, Russia and South Africa.
Sources: EPFR; national data; authors’ calculations.
It is evident from Graph 2 that there is some heterogeneity in the relationship between long-term interest
rates and portfolio flows across EME regions. What might explain these different sensitivities? Local market
conditions may be important for the relationship between bond flows and yield changes. EME bond markets, in
particular, differ in their depth and liquidity. When a country’s bond market liquidity
3 Here, and for the remainder of the analysis, we rely on EPFR data due to the wider country coverage than IIF data. The EPFR flows are
reported in millions of US dollars. However, in order to make the results comparable across countries and avoid having the largest
economies dominate the results, we standardise the amount of flows by dividing each country’s flows by its standard deviation. Although
EPFR provides data on EME bond flows starting from January 2004, we limit the sample to January 2010 onwards (and even later for some
countries) due to low coverage and other data issues in the early parts of the sample.
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
–
–
–
–
–
–
–
–
–
–
is lower than that of another country, the same amount of bond outflows and the corresponding amount of bond
sales are likely to have a bigger impact on bond prices. Sorting EMEs by their average local currency bond
market bid-ask spread, we find that, before the Covid-19 crisis, bond prices in markets with higher bid-ask spreads
were more sensitive to bond flows (Graph 3, left-hand panel). This pattern has been less clear during the Covid-19
crisis, although for the group with the lowest bid-ask spreads the yield sensitivity to bond flows has been 60–
70% lower than that in the two groups with higher average spreads.
Sensitivity of long-term interest rates to bond flows in EMEs Graph 3
Sorted by bid-ask spread Sorted by foreign participation
1 2 3
Lowest --> highest
1 2 3
Lowest --> highest
Pre-crisis Crisis
“Pre-crisis” refers to the period from January 2010 to February 2020, while “Crisis” refers to March 2020. The bars represent slope coefficients from
regressions of changes in five-year bond yields on bond portfolio flows, in three different country groups where countries have been sorted by average
bid-ask spread (left-hand panel) or average foreign participation (right-hand panel).
Sources: Arslanalp and Tsuda (2014); AsianBondsOnline; Bloomberg; EPFR; authors’ calculations.
Another possibility is that a higher degree of foreign participation in local currency bond markets is
associated with a greater sensitivity of yields to bond flows (see Cerutti et al (2019) and Hofmann et al (2020)
for a discussion). The right-hand panel of Graph 3 shows that in the past there was a mild tendency for markets with
higher foreign participation to exhibit higher yield sensitivity to flows. During the Covid-19 crisis, this
relationship has strengthened considerably. Hence, EMEs with high foreign bond market participation have on
average suffered much larger yield increases for any given bond portfolio outflow than those with low foreign
participation. A possible explanation for this phenomenon is that, in markets with very high foreign participation,
domestic investors would need to step in and purchase relatively larger amounts of bonds as foreign investors
rush for the exits in a crisis. As a result, in such markets, the price impact of a given bond outflow may be
amplified, as yields would need to rise substantially to entice domestic bond
The relationship between yields and bond flows may also vary depending on how volatile the exchange
rate is. It is possible that, for countries with systematically higher exchange rate variability, this relationship may
be stronger. One reason could be that international investors react more aggressively in markets where they face
greater risk due to higher exchange rate volatility. Graph 4 displays the yield-flow relationship for four groups of
EMEs sorted by the average US dollar exchange rate volatility before and during the Covid-19 crisis. While there
is no clear pattern across the groups before the crisis, we find that the higher the exchange rate volatility during the
crisis, the greater the yield increases in response to bond outflows. It therefore seems that the crisis-time volatility of
the exchange rate is more important than the “typical” volatility for the yield-flow relationship, suggesting that
international investors are more concerned about the former than the latter.
4 Another explanation of this phenomenon is a more pronounced relationship between bond portfolio outflows and exchange rate
depreciation during a crisis. We find that EME local currency depreciations against the US dollar tended to be followed by bond fund
outflows before the Covid-19 crisis, and that this relationship has strengthened in EMEs during the crisis.
Change in long-term interest rates and bond flows in EMEs, sorted by FX volatility Graph 4
Lowest FX volatility Low FX volatility High FX volatility Highest FX volatility
y = –– 2
where R2 =
1
0
–1
y = ––
where R2 =
–2
y = – 2
where R2=
1
0
–1
y = ––
where R2=
–2
y = –– 2
where R2=
1
0
–1
y = ––
where R2=
–2
y = ––
2
where R2 =
1
0
–1
y = ––
where R2=
–2
–20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10
Bond flow (standardised) Bond flow (standardised)
Pre-crisis
Bond flow (standardised)
Crisis
Bond flow (standardised)
“Pre-crisis” refers to the period from January 2010 to February 2020, while “Crisis” refers to March 2020. Country bond flows are standardised by their
standard deviation. Sorting is based on the standard deviation of weekly exchange rate returns against the US dollar during each of the two periods.
Sources: EPFR; national data; authors’ calculations.
Finally, we consider the possibility that financial stress or market-based indicators of the
creditworthiness of EMEs as receiving countries may affect the relationship between local long-term interest
rates and bond flows. It is entirely possible that investors, in particular foreign ones, will react more strongly during a
crisis in markets that are characterised by low creditworthiness or relatively higher financial stress than
elsewhere. We use pre-crisis sovereign CDS spreads to proxy for investors’ perceptions of such market
Before the Covid-19 crisis, there was some tendency for this relationship to strengthen with higher
CDS spreads (Graph 5, blue lines). During the crisis, this effect has become much more pronounced: in the group
with the highest CDS spreads, yield changes were almost eight times higher on average for a given bond flow
than in the group with the lowest CDS spreads (Graph 5, red lines). The perceived fragility, as measured by the
CDS spread, matters greatly in times of crisis.
Change in long-term interest rates and bond flows in EMEs, sorted by CDS spreads Graph 5
Lowest CDS spread Low CDS spread High CDS spread Highest CDS spread
y = –– 2
where R2=
1
0
–1
y = ––
where R2=
–2
y = – 2
where R2=
1
0
–1
y = ––
where R2=
–2
y = –– 2
where R2=
1
0
–1
y = ––
where R2=
–2
y = – 2
where R2=
1
0
–1
y = ––
where R2=
–2
–20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10 –20 –15 –10 –5 0 5 10
Bond flow (standardised) Bond flow (standardised)
Pre-crisis
Bond flow (standardised)
Crisis
Bond flow (standardised)
“Pre-crisis” refers to the period from January 2010 to February 2020, while “Crisis” refers to March 2020. Country bond flows are standardised by their
standard deviation. Sorting is based on the average daily five-year sovereign CDS spread during 2019.
Sources: EPFR; national data; authors’ calculations.
5 Specifically, we use the average daily five-year sovereign CDS spread during 2019 as a proxy.
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Fi
ve
-y
ea
r y
ie
ld
c
ha
ng
e
(%
)
Policy considerations
Our analysis suggests that when EMEs face strong outflows from their local currency bond markets, various policy
initiatives may have an important role to play in directly stabilising those markets. During the Covid-19 crisis,
such policy measures have included government bond purchase programmes, as launched by more than 10 EME
central banks in March–April 2020, and the establishment of bond market stabilisation funds by some EME
financial authorities. Another useful tool utilised during the crisis has been duration swaps, which allow central
banks to swap long-term securities for short-term ones via auctions, thereby stabilising conditions in local
currency bond markets in situations where long-term securities come under great selling pressure and there is
market demand for short-term
Given that exchange rate fluctuations play a central role in these dynamics, addressing such
fluctuations will also help quell turbulence in domestic financial conditions through sharp changes in long- term
interest rates. Moreover, by providing liquidity and temporarily relaxing regulatory requirements, authorities can
increase the room for domestic institutional investors to step in and purchase local currency bonds, thereby further
contributing to the stabilisation of bond markets.
Our results show that markets associated with lower liquidity, higher foreign participation, higher FX
volatility or higher credit spreads have exhibited a sharper relationship between bond yields and bond portfolio
flows during the Covid-19 pandemic. These results point to the importance of strengthening the resilience of
domestic markets/economies in the longer term against a sharp deterioration in global financial conditions. To
that end, local bond markets could become less sensitive to portfolio flow pressures if the efficient functioning
and liquidity/depth of these markets are strengthened. Policymakers’ efforts to develop and deepen local capital
markets, such as fostering a strong domestic institutional investor base, would therefore result in better financial
resilience and an enhanced ability to withstand global funding shocks. Finally, efforts to improve EMEs’
creditworthiness are also likely to be important for maintaining the resilience of EME bond markets during market
turmoil.
References
Arslanalp, S and T Tsuda (2014): “Tracking global demand for emerging market sovereign debt”, IMF
Working Papers, no WP/14/39.
Avdjiev, S, E Eren and P McGuire (2020): “Dollar funding costs during the Covid-19 crisis through the lens of the
FX swap market”, BIS Bulletin, no 1, April.
Cerutti, E, S Claessens and D Puy (2019): “Push factors and capital flows to emerging markets: why knowing your
lender matters more than fundamentals”, Journal of International Economics, vol 119, pp 133–49.
Hofmann, B, I Shim and H S Shin (2020): “Impact of the Covid-19 pandemic on exchange rates and local
currency bond markets”, BIS Bulletin, no 5, April.
6 For example, in April 2020 the Bank of Mexico announced that it would receive securities with maturities of 10 years or longer and deliver
other securities with maturities of up to three years. This is similar to the duration swap operation conducted by the same central bank
during the GFC.
