1 s2.0 S0261560622000791 Main

Journal of International Money and Finance 127 (2022) 102676
Contents lists available at ScienceDirect
Journal of International Money and Finance

journal homepage: www.elsevier.com/locate/jimf
Review
The effects of U.S. monetary policy on international mutual fund

investment q
Gabriele Ciminelli a,b,⇑, John Rogers c, Wenbin Wu c
a
Asia School of Business in collaboration with MIT Sloan, Kuala Lumpur, Malaysia
b
Organisation for Economic Co-operation and Development, Paris, France
c
Fanhai International School of Finance, Fudan University, Shanghai, China
a r t i c l e i n f o a b s t r a c t
Article history: We study the effects of U.S. monetary policy on international mutual fund investment. We
Available online 30 May 2022 apply a novel variant of the shock identification procedure in Bu et al. (2021) to decompose
observed U.S. monetary policy surprises into pure monetary policy shock and information
JEL Codes:: news shock components. An increase in interest rates driven by a pure monetary policy
E44 shock leads to large and persistent outflows from emerging markets (EMs) and to a lesser
E52 extent global funds. On the other hand, increases in interest rates driven by positive infor-
F30
mation news shocks (i) do not cause outflows from EM funds, and (ii) lead investors to real-
G15
locate capital out of safe U.S. bond funds and into growth-sensitive U.S. and global equity
Keywords: funds. We attribute these differences to the risk-taking channel of monetary policy. Pure
Federal Reserve monetary policy shocks heighten risk aversion, while information news shocks lower
Emerging Markets uncertainty.
Monetary Policy Shocks Ó 2022 Elsevier Ltd. All rights reserved.
Information Effects
Investment Funds
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Dataset and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1. U.S. monetary policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. Investment funds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3. Other variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4. Econometric framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3. Main results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1. The effects of monetary policy shocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2. The effects of information news shocks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3. Results using alternative identification methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
q
This paper should not be reported as representing the official views of the OECD or of its member countries. The opinions expressed and arguments
employed are those of the author(s). The first draft of this paper was written when John Rogers was Senior Adviser in the International Finance Division of
the Board of Governors of the Federal Reserve System and Gabriele Ciminelli was Assistant Professor of Finance and Economics at the Asia School of
Business. Please address correspondence to gabriele.ciminelli@asb.edu.my. For regular updates of our shock series, see https://sites.google.com/view/
wenbinwu-ucsd/research.
⇑ Corresponding author.
E-mail addresses: gabriele.ciminelli@asb.edu.my (G. Ciminelli), johnrogers@fudan.edu.cn (J. Rogers), wenbinwu@fudan.edu.cn (W. Wu).
https://doi.org/10.1016/j.jimonfin.2022.102676
0261-5606/Ó 2022 Elsevier Ltd. All rights reserved.
G. Ciminelli, J. Rogers and W. Wu Journal of International Money and Finance 127 (2022) 102676
3.3.1. Results without separating monetary policy and information news shocks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3.2. Alternative approaches to derive pure monetary policy and information news shocks . . . . . . . . . . . . . . . . . . . . 11
3.4. Robustness checks and alternative specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4. Transmission channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5. Extensions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.1. Asymmetric behavior of U.S. and EM investors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.2. Zooming in on EM bond funds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.3. Other extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Declaration of Competing Interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Appendix A. Appendix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
A.1. Additional tables and figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Appendix B. Supplementary material. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
China should be mindful of a potential reversal of capital flows as the US Federal Open Market Committee gets ready to kick off
its monetary tapering by dialing back US$120 billion worth of monthly bond purchases, a former Chinese central bank official
has warned.There are rising concerns among investors that the tapering exercise will result in a more unstable yuan, as large
amounts of capital – i.e. hot money – will rush into the Chinese market as investors bet on the yuan’s one-way appreciation
before the tapering is completed, but then rush out again to take advantage of a stronger US dollar after the tapering is
finalised.”
[— South China Morning Post, 03 November 2021]
1. Introduction
In the aftermath of the global financial crisis, massive unconventional measures employed by the Federal Reserve trig-
gered debates about the consequences of flooding the world with liquidity. A significant adverse consequence of this
unprecedented monetary expansion was the Taper Tantrum in 2013 when the Fed expressed a future plan to gradually taper
its balance sheet, which was followed by turmoil in emerging markets (EMs). Similar concerns have arisen again with the
Fed’s restoring many unconventional policies after the outbreak of the COVID pandemic and subsequently with monetary
policy tightening at the doorstep in early 2022.
In this paper, we take a novel tack in revisiting the role of U.S. monetary policy in influencing capital flows.1 We do so by
examining the effects and transmission channels of U.S. monetary policy on investors’ allocations into international mutual
funds, applying a variant of the monetary policy shock identification procedure of Bu et al. (2021) (BRW). Our new procedure
allows us to decompose observed U.S. monetary policy surprises into pure monetary policy shock and information news shock
components. New to the capital flows literature, we document that allocations into mutual funds exhibit very different
responses to tightening movements in U.S. interest rates caused by a pure monetary policy shock versus an information news
shock.
The pure monetary policy shock captures a sudden shift in monetary policy that is orthogonal to changes in the economic
outlook. The information news shock instead captures a change in the policy indicator that depends on changes in the
FOMC’s economic outlook (even if unexpected by the public). This was emphasized in Romer and Romer (2000), who showed
that the Fed’s forecast of inflation was superior to commercial forecasts. The information effect was further analyzed by
Nakamura and Steinsson (2018), Lunsford (2020), Jarociński and Karadi (2020), and recently by Acosta (2021) and Bauer
and Swanson (2021), who have an alternative view of the information channel.2 Our paper is the first to make use of this
decomposition to analyze capital flows.
We find that the effects of a pure U.S. monetary policy shock on investors’ allocations into mutual funds are conventional
– a tightening leads to capital outflows from global and EM funds. But information news shocks have distinctly different
1
For examples of this sizable literature, see the recent papers by Chari et al. (2020) and Dahlhaus and Vasishtha (2020) and the references there-in, as well as
our more fleshed out discussion below.
2
Bauer and Swanson (2021) note that if the Fed and Blue Chip forecasters are responding to the same economic news and the Fed responds by more than
financial markets expected them to, the consequent positive correlation between monetary policy surprises and forecast revisions would be consistent with a
Fed information effect. For example, if monetary policy tightens in response to positive economic news by an unexpectedly large amount, the resulting positive
monetary policy surprises occur in times of positive economic news. Their alternative explanation to the information effect, labeled the ‘‘Fed response to news”
channel, implies that standard information effect regressions, such as those we estimate below, suffer from an omitted variables problem. Because economic
news released prior to FOMC announcements predicts subsequent monetary policy surprises, the authors find, including pre-FOMC economic news in these
regressions removes the evidence of a Fed information effect. Thus, news causes both the Fed’s change in monetary policy and the private sector forecasts
revision.
2
effects. A positive shock, indicating higher interest rates against the backdrop of a better economic outlook, does not lead
investors to sell EM funds shares; it even leads them to buy shares of U.S. and global equity funds. We reconcile these dif-
ferent effects through the presence of a Fed information effect and the operation of the risk-taking channel of U.S. monetary
policy. This channel sharply differs following pure monetary policy shocks and information news shocks.
As postulated in the existing literature, there are three main transmission channels of U.S. monetary policy for
international capital flows – the portfolio rebalancing channel, the risk-taking channel, and the exchange rate channel
(Chari et al., 2020; Bruno and Shin, 2015). According to the portfolio rebalancing channel, investors shift their portfo-
lios towards high-yield EM assets when global interest rates are low and hence investing in long-term advanced
economies bonds generates meager returns. The opposite holds when the Fed tightens policy and interest rates rise.3
The risk-taking channel operates through the effect of U.S. monetary policy on risk aversion (Bruno and Shin, 2015). A
tightening in interest rates that increases risk aversion makes investors more reluctant to take on risky positions, leading
to capital outflows from EM and other risky funds. Finally, the exchange rate channel depends on the direct effect of the
Fed’s policy on the USD exchange rate, with an appreciation of the dollar leading to capital outflows from other countries,
especially EMs.
We find evidence of an important role for the risk-taking channel. Pure monetary and information news shocks
both increase nominal and real interest rates, leading us to downplay portfolio rebalancing as the main driver of
our results. However, the effects of these two shocks on risk-taking and uncertainty, jointly proxied by the VIX index,
are sharply different. A tightening pure monetary policy shock substantially increases risk aversion, and therefore the
VIX, while positive information news shocks decrease uncertainty, thereby lowering the VIX. As we document, and
consistent with the capital flows literature (Ahmed and Zlate, 2014, and others), risk aversion has strong negative
effects on flows to EM mutual funds, which explains why tightening pure monetary policy shocks have particularly
large negative effects. On the other hand, the decrease in uncertainty following positive information news shocks,
which is consistent with the information effect, neutralizes the negative effects of higher interest rates on higher-
yielding assets. This explains why investors shed riskier assets following pure monetary policy shocks but not informa-
tion news shocks.
The important role of the risk-taking channel in driving our results is confirmed when we zoom in on the vast universe of
U.S. bond funds. We find that investors respond to an increase in interest rates driven by a pure monetary policy shock by
selling shares of funds investing in U.S. corporate bonds and buying shares of funds investing in safe-haven long-term Trea-
suries. Instead, when interest rates increase following a positive information news shock, investors sell long-term Treasury
funds and increase allocations into high-yield bond funds.
The evidence on the exchange rate channel is mixed. The price of the USD shoots up (dollar appreciates) following pure
monetary policy shock tightenings, while it exhibits a qualitatively similar but weaker and more sluggish response after pos-
itive information news shocks. Hence, the exchange rate channel seems to work in the same direction as the risk-taking
channel, but it cannot explain why investors also marginally reduce their positions in U.S. corporate bond funds after pure
monetary policy shocks. This instead can be rationalized with an increase in risk aversion.
We also perform some extensions to these core results. We first exploit additional information on where the different
funds are domiciled to explore heterogeneity in the responses of U.S. and EM investors. U.S. investors sharply decrease
exposure to EM assets following a pure monetary policy tightening. This decrease in foreign capital in EMs is partly mit-
igated by EM investors themselves, who increase flows into EM funds. These results are consistent with home-bias behav-
ior following heightened risk aversion and more generally suggest that tightening monetary policy shocks in the U.S. cause
a decrease in foreign capital. We confirm this intuition by using country-level data on cross-border flows taking place
through investment funds. Tightening Fed monetary policy shocks lead to a general decrease in foreign inflows, particu-
larly debt. Although larger in EMs, this decline in foreign capital also affects the U.S. and other developed markets. On the
other hand, positive information news shocks cause inflows of foreign capital into U.S. equity assets, in line with the news
of a stronger economy.
Our results naturally raise the question of what could be done by EM debt issuers to mitigate the decrease in foreign cap-
ital following tightening Fed monetary policy shocks. To investigate, we leverage information on bond funds’ investment
mandate. Bond funds differ along several dimensions, including credit quality, currency denomination, and issuer of the
bonds in which they can invest. These may affect the response of investors following U.S. shocks. We find that credit quality
and currency denomination are important. Other things being equal, outflows from funds investing exclusively in invest-
ment grade bonds are about half as a large as outflows from other funds, following a tightening pure monetary policy shock.
At the same time, funds which can invest only in hard currency (CHF, EUR, GBP, USD, YEN) bonds suffer much larger outflows
than local and mixed currency funds. This result has important policy implications as it suggests that, by issuing debt in local
currency, EM issuers may decrease exposure not only to exchange rate fluctuations but also to changes in risk aversion fol-
lowing shocks to U.S. monetary policy.
Related Literature: Our paper is closely related to two strands of literature. First, to the literature on international portfolio
flows, especially EM portfolio flows. A vast literature investigates the effects of Fed monetary policy on EM capital flows (see
3
The portfolio rebalancing channel should be more pronounced during the zero lower bound when the short-term interest rate is close to zero and the Fed
may employ large scale asset purchases (LSAPs) and forward guidance to lower long term interest rates further (Neely, 2010; Krishnamurthy and Vissing-
Jorgensen, 2011; Fratzscher, 2012; Fratzscher et al., 2018; Chari et al., 2020).
3
Calvo et al., 1993, among many others; Burger et al., 2018; Bruno and Shin, 2015; Fratzscher, 2012; Fratzscher et al., 2018;
Iacoviello and Navarro, 2019; Dahlhaus and Vasishtha, 2020).4 In addition, Miranda-Agrippino and Nenova (2021) argue that
monetary policy tightenings in both the U.S. and Euro area are followed by a global retrenchment in capital flows, and thus that
ECB and Fed monetary policies propagate internationally through equivalent transmission channels. We differ from these stud-
ies to consider the role of Fed information effects, a topic that has not yet been investigated in the capital flows literature.5
Second, our paper is related to literature on Fed information effects and predictability of monetary policy shocks. As noted
above, the Fed information effect refers to the view that FOMC announcements not only reveal pure monetary policy news
but also that announcements inform the public about either private information the Fed possesses or its outlook for the
economy based on common information. This leads to predictability of monetary policy shocks (see Romer and Romer,
2000; Campbell et al., 2012; Nakamura and Steinsson, 2018; Miranda-Agrippino, 2016; Jarociński and Karadi, 2020;
Hansen et al., 2019; Cieslak and Schrimpf, 2019; Paul, 2019; Bauer and Swanson, 2021; Acosta, 2021). Our paper is thus
related to Hoek et al. (forthcoming), who find that a hike in U.S. interest rates is not always bad news for EMs. They show
that an increase in U.S. interest rates stemming from stronger U.S. growth generates moderate or even opposite spillover
effects on EM asset prices. These authors do not focus on capital flows and use a different and more traditional identification
of U.S. monetary policy shocks.6 In this paper, we show that Fed information news shocks do not cause international capital
outflows from global and EM markets—but pure monetary policy shocks do—and we analyze the transmission channels. Our
main results echo closed-economy macro papers that document differently-signed transmission effects to output and inflation
from pure monetary policy news shocks versus information news shocks (e.g., Jarociński and Karadi, 2020).
2. Dataset and methodology
The dataset covers 2002–2020, spanning over 150 FOMC meetings, and includes more than 22,000 investment funds. In
the rest of this section, we first illustrate how we measure Fed monetary policy. Second, we describe the investment fund
database. Third, we discuss the other variables in the dataset. Fourth, we explain the empirical framework that we employ
to run the analysis.
2.1. U.S. monetary policy
The Federal Open Market Committee (FOMC) sets U.S. monetary policy through eight scheduled meetings, and occasional
unscheduled meetings typically in times of stress, every year.7 The committee discusses current and expected future economic
conditions, decides monetary policy accordingly, and reveals its thinking and policies to the public in various ways, including
through the FOMC Statement. It is well known that a major part of changes of monetary policy is due to systematic reactions to
economic conditions, and that to measure ‘‘pure” monetary policy shocks, it is important to take into consideration the infor-
mation effect component of FOMC announcements. Traditional SVAR and high-frequency approaches might both fail to appro-
priately identify monetary policy shocks because information effects are simultaneously revealed with exogenous monetary
policy changes by the Fed.
Following Bu et al. (2021), we use a heteroskedasticity-based, partial least squares (PLS) approach, exploiting the sensi-
tivity of different outcome variables to FOMC announcements, to identify shocks to U.S. monetary policy. Through this PLS
approach, which is essentially the same as the Fama and MacBeth (1973) method, we are able to separate pure monetary
policy shocks from the information content of FOMC announcements. Our PLS approach consists of three steps. In the first
step we run time-series regressions to estimate the sensitivity of interest rates at different maturities to FOMC announce-
ments. In the second step, for each time t, we regress the cross-section of outcome variables onto their corresponding esti-
mated sensitivity from step one. In a third step, which is new to this paper, we back out the information content of the
monetary policy news from the pure monetary policy shock. This PLS three-step approach has very mild data requirements
and is simple to implement. Bu et al. (2021) show that the monetary policy shock series, derived from steps 1 and 2, stably
bridges periods of conventional and unconventional policymaking, is largely unpredictable, and contains no significant cen-
tral bank information effect. These shocks have particularly large effects on maturities in the middle of the term structure
and produce conventionally-signed impulse responses of output and inflation.
Next, we lay out a simple framework to illustrate the PLS approach. For more details, please refer to Bu et al. (2021). The
monetary policy shock et is assumed to be unobservable. We further assume that the (observable) changes in Treasury yields
4
Many authors have documented how important capital flows are for other key financial and macroeconomic outcomes. Warnock and Warnock (2009)
document the influence of international capital flows on long-term U.S. interest rates and Liu et al. (2020) analyze the effects of capital flow ‘‘surges” on income
inequality.
5
After the first draft of this paper was completed, we became aware of Pinchetti and Szczepaniak (2021), who also examine spillover effects of Fed
information news shocks to capital flows and many other variables. Our paper differs from theirs in a few dimensions, including (1) we use a different
methodology to identify the information news shocks – namely that based on (Bu et al., 2021) rather than that in Jarociński and Karadi (2020) (see the former
for a comparison of these identification strategies, as well as our robustness analysis); (2) we use microdata on allocations into mutual funds while they mainly
rely on macro data.
6
Though they do use the Bu et al. (2021) shock measure in a robustness exercise.
7
We focus on scheduled meetings and check robustness to including unscheduled meetings. The point estimates tend to be very similar and the confidence
bands are slightly tighter (results available upon request).
4
around FOMC announcement days are driven by the monetary policy shock et and a non-monetary policy shock t , which can
include the information content gt of FOMC announcements. Bu et al. (2021) show that monetary policy shocks can be iden-
tified as long as the information effects in the long-run yields (gL;t ) are very weak, or the information signal driving the short
and long-run yields (gS;t and gL;t ) are different, an implication that is supported by Hansen et al. (2019).
In the first step, we estimate the sensitivity of each outcome variable (zero-coupon yields with maturities of 1 to 30 years)
to monetary policy via time-series regressions, according to:
DRi;t ¼ ai þ ci et þ i;t ð1Þ
where DRi;t is the change in the zero-coupon yield with i-year maturity and i;t is the error term, which can include the infor-
mation content driving the short and long-run yields (gS;t and gL;t ). We assume the error term i;t and the unobserved mon-
etary shock et are uncorrelated. As in Bu et al. (2021) and Hansen et al. (2019), we further assume that the identification
assumptions are satisfied. We then normalize the unobserved monetary policy shock to have a one-to-one relationship with
the 5-year Treasury yield.8 Due to our normalization, we can rewrite Eq. (1) as:
DRi;t ¼ hi þ bi DR5;t þ ni;t ð2Þ
The errors-in-variables problem caused by background noises can be eliminated using identification through heteroskedas-
ticity. As shown in Bu et al. (2021), the estimated bi coefficients are proportional to the underlying true ci coefficients.
The second step is to extract the pure monetary policy shock from monetary policy news through cross-sectional regres-
^i for each time t:
sions of DRi;t on the estimated sensitivity index b
DRi;t ¼ di þ epure ^i þ v i;t ;

b for t ¼ 1; 2; ; T ð3Þ
t
where epure
tis the coefficient of interest. The pure monetary policy shock series is then obtained as the series of these T esti-
mated coefficients.
In a third step, new to this paper, we back out the information component as the residuals from projecting the benchmark
T
5-year interest rate onto the estimated series fepure
t gt¼1 .
Fig. 1 plots the pure monetary policy shock and information effect shock over time. Table 1 displays descriptive statistics.
Both shocks have been important in driving interest rates following FOMC meetings in the last two decades. Both series dis-
play higher variability during the GFC and subsequent zero lower bound (ZLB) periods.
Importantly, our information news shock series manages to capture nuances in FOMC communication fairly well. As an
example, we take the 17–18 March 2009’s FOMC meeting, for which our information news shock series takes its most neg-
ative value. In that meeting, the FOMC announced a new round of large scale asset purchases (LSAPs), which led to a sharp
drop in long-term interest rates and a rise in the stock market. Looking at the FOMC meeting minutes, it appears that the new
round of LSAPs was motivated by a sharp fall in economic activity in the preceding months. Our information news shock
series, which registers a value of almost 20 b.p., implies that the FOMC had a pessimistic view of the economic outlook,
which seems to be supported by the information contained in minutes. Alternative methodologies relying on the stock mar-
ket’s response to derive the information news shock imply instead that the FOMC had an upbeat view of the outlook (the U.S.
stock market rose that day). In Table A1 in the Appendix we formally check that our information shock series is associated
with upward revisions to U.S. GDP forecasts.
2.2. Investment funds
To measure allocations into investment funds we rely on Emerging Portfolio Funds Research (EPFR), a commercial data
provider popular among financial professionals to track trends in the investment fund industry and also used in academic
research (see Converse et al., 2020, among others; Fratzscher, 2012). EPFR receives data from individual funds and then dis-
seminates these data to its subscribers. Allocations into investment funds (fund flows) (F i;t ) is our main variable of interest. It
measures total net purchases of fund i’s shares during period t made by individual investors. To standardize flows among
different funds, we also source data on assets under management by fund i at the beginning of period t (Ai;t ). Finally, since
fund flows partly depend on past returns, we collect data on period-on-period changes in net asset value (NAV), excluding
changes due to new investors’ allocations ni;t . Both assets and flows are expressed in USD ($), while NAV returns are in
percent.
EPFR also provides information on a fund’s investment mandate. Hence, for all funds, we collect information on its geo-
graphical focus, and for bond funds, we additionally collect information on whether they invest in (i) sovereign, corporate or
both types of funds, (ii) investment-grade, high-yield, or both types of funds, and (iii) hard currency, local currency or both
types of funds. EPFR data are available at the daily, weekly and monthly frequencies. We opt for weekly data, which cover
more funds than daily data but still allow for a proper identification of the effects of Fed monetary policy on fund flows.9 The
data are available beginning in 2002 and 2004 for equity and bond funds, respectively.
8
Results normalizing to the 2-year or 10-year rate, reported in the Online Appendix, are effectively identical.
9
The week is defined to start on Thursday at the beginning of the U.S. trading day.
5
Fig. 1. Fed’s shocks over time. Notes: The figure plots the pure monetary policy shock and information news shock components identified from Fed
monetary policy surprises over time.
Table 1
Descriptive statistics on Fed’s shocks.
Pure monetary policy shocks information news shocks

obs mean abs. mean s.d. mean abs. mean s.d.
2002–2007 47 0.05 3.09 3.86 0.70 3.56 4.29
GFC 13 2.37 6.90 9.82 0.34 8.83 11.36
ZLB 52 0.95 5.14 4.72 1.13 5.58 6.90
2015–2019 32 0.38 3.25 3.94 1.85 2.83 3.12
COVID 7 1.47 2.04 2.27 0.09 3.44 4.07
All sample 151 0.50 3.97 5.55 0.24 4.54 6.05
Initially, relatively few funds reported to EPFR, but the sample of reporting funds has steadily grown over time. In 2017,
the funds covered by EPFR cumulatively held over $24 trillion of assets, representing roughly half of the entire industry. We
focus on funds with mandate to invest (i) in the U.S. only, (ii) in EMs, or (iii) in any country in the world, therefore including
the U.S., EMs and advanced economies other than the U.S. We refer to this latter type of funds as Global funds.
We screen the data by excluding funds that are in the sample for less than one year, as well as small funds, defined
as those with less than $10 million of assets on average, and we censor abnormal jumps, defined as observations having
flows larger than one-third of assets in absolute value.10 Table 2 reports basic descriptive statistics. Our sample
comprises more than 15,000 equity and 7,000 bond funds, with a bit less than half of them being domiciled in the U.S. There
is large dispersion in fund size, with the standard deviation of the variable measuring average assets of each fund being about
three to four times the mean.
2.3. Other variables
We also source data on additional variables to use as controls in our regressions. The first set of controls is U.S. macroe-
conomic releases, which may affect investment behavior as they provide direct information about the state of the economy
and noisy signals about likely future monetary policy. We consider the four releases that have the largest impact on financial
markets (Faust et al., 2007; Beber and Brandt, 2009; Swanson and Williams, 2014; Gilbert et al., 2017) – non-farm payroll
10
This is equivalent to winsorizing less than the top and bottom percentiles of observations.
6
Table 2
Descriptive statistics of funds covered.
Mandate # funds % domiciled Assets

in the U.S. mean s.d.
U.S. - equity 6025 75.1 830.4 3532.3
Global - equity 5183 33.3 534.1 1673.4
EMs - equity 3883 15.8 341.5 1226.4
Total - equity 15091 45.5 602.8 2523.6
U.S. - bond 3178 83.0 883.0 2785.0
Global - bond 2327 13.8 567.3 1361.7
EMs - bond 1535 10.7 354.8 689.4
Total - bond 7040 44.3 663.5 2064.6
employment net creation, month-on-month core CPI inflation, retail sales percent change, and the ISM manufacturing PMI
index. Since market participants form expectations about upcoming macroeconomic releases, we follow standard practice in
the literature and identify their unexpected — ‘surprise’ — component.11 To do so, we collect data on the median response to a
Bloomberg survey asking economic analysts about their expectation and subtract the median expectation from the actual
release. We also collect data on the TED spread, a measure of tightness in the U.S. interbank market, obtained as the difference
between the 3-month U.S. Treasury bill and LIBOR rates. Relevant data are sourced from Bloomberg and Datastream.
To compare our results to those that would be obtained using alternative approaches to identify monetary policy shocks
we source additional variables. First, we collect data on the current federal funds rate, the federal funds rate expected fol-
lowing the subsequent FOMC meeting, as well as eurodollar futures rates expected 2, 3, and 4 quarters ahead; with these,
we construct an alternative measure of monetary policy ‘‘news”. Second, we source the information news shock and pure
monetary policy shock variables of Jarociński and Karadi (2020).12
To explore potential transmission channels, we source data on the VIX index, a measure of expected financial market
volatility, the USD index (the DXY) as well as 5-year nominal and real (TIPS) U.S. government bond yields from Datastream.
The dataset is completed with measures of the price and the level of risk sourced from Bekaert et al. (2021).
2.4. Econometric framework
To trace out the dynamic response of allocations into mutual funds to pure monetary policy and information news shocks,
we estimate local projections. This approach was pioneered by Jordà (2005) and has been widely used as a flexible alterna-
tive to autoregressive distributed lag specifications (Auerbach and Gorodnichenko, 2012; Jordà and Taylor, 2016; Romer and
Romer, 2017; Ramey and Zubairy, 2018). We consider a 5-week window, including the week of the FOMC meeting and the
following four.
To apply the local projections method in the context of fund flows, we first construct the dependent variable as the ratio
of cumulative allocations to fund i over horizon t þ k to assets under management of fund i at time t. Then, for each
k ¼ 0; ::; 4, we regress our dependent variable onto both the monetary policy and information news shock variables at time
t. In practice, we estimate:
X
k
F i;tþj
24
X
j¼0 F i;tl
¼ bk et þ ck it þ rkl þ /kl ni;tl þ di þ ei;t ð4Þ
Ai;t l¼1
Ai;tl
where subscripts i and t denote fund and time respectively; the superscript k denotes the horizon considered; F i;t denotes
fund flows (investors’ total net purchases of fund i’s shares during period t, in USD); Ai;t is the volume of assets under man-
agement by fund i at the beginning of period t, in USD; et and it are our pure monetary policy and information news shock
variables, which take values equal to the respective shock in FOMC meeting weeks and 0 otherwise; ni;t denotes fund i’s net
returns relative to other funds with the same investment mandate over period t; di are investment fund fixed effects that take
into account time-invariant fund characteristics; and ei;t is the error term, assumed to be uncorrelated with the regressors.
Since a large literature has shown that investors purchases of shares depend on the fund’s past performance, we include
P
24 lags of one-period relative fund returns (captured by the term 24 k
l¼1 /l ni;tl ). Lastly, we account for autocorrelation in fund
P
flows by including 24 lags of one-period flows in the regression (captured by the term 24 l¼1 rl ðF i;tl =Ai;tl Þ).
k 13
The bk s and ck s
are the coefficients of interest. They measure the mean cumulative response over the t þ k horizon of fund flows to a 10 b.p.
monetary policy and information news shock respectively.
11
See Gürkaynak et al. (2005);Boyd et al. (2005); Beber and Brandt (2009),Swanson and Williams (2014) among others.
12
The information and monetary policy shock variables in Jarociński and Karadi (2020) paper are up to end-2016. We source updated data (to March 2019)
from Marek Jarocinski’s website (https://marekjarocinski.github.io/).
13
As the analysis is at the weekly frequency, including 24 lags is equivalent to account for past fund flows and returns over the previous six months. In a
sensitivity analysis, we show that the results do not depend on the chosen lag structure nor on the inclusion of returns as a control.
7
Fig. 2. Investors’ allocations into mutual funds after a 10 bp pure MP shock. Notes: Dashed black lines report point estimates. Deep and shallow gray areas
are 90% and 68% confidence bands. X-axes denote the response horizon (in weeks), with 0 being the week of the FOMC meeting. Y-axes denote the
magnitude of the response (in percentage points). Responses are obtained estimating the bk coefficients from Eq. 4. Estimates are normalized to show
responses to a 10 b.p. shock.
We perform the estimation through OLS with two-way clustered standard errors (fund i and time t). This clustering
approach accounts for both autocorrelation in flows within the same fund and for cross-fund correlation in flows within
^f s and c
the same period. We plot impulse response functions (IRFs) constructed using the b ^fk s for the point estimate and their
k
respective standard errors to derive 68 and 90 percent confidence bands.14
3. Main results
3.1. The effects of monetary policy shocks
Fig. 2 shows average responses of investors’ allocations into mutual funds to a 10 basis point pure Fed monetary policy
shock, over a 5-week horizon. Responses are measured in percentage points of funds’ assets. Panels A & D, B & E and C & F
present impulse responses for funds investing, respectively, in the U.S., in any country of the world (Global) and EMs only.
Responses are estimated over the full sample (January/2002 to December/2020 for equity funds and January/2004 to Decem-
ber/2020 for bond funds). Black dotted lines report point estimates, while 68% and 90% standard error confidence intervals
are displayed as deep and shallow gray areas respectively.15
A Fed tightening shock has negative and persistent effects on flows to most funds. These effects are largest and more pre-
cisely estimated for EM funds. Four weeks after a 10 b.p. shock, flows to EM equity and bond funds are estimated to be about
0.3 and 0.6 p.p. lower, respectively. The same shock has smaller but still statistically significant negative effects on flows to
global equity and global bond funds (the response is about 0.2 p.p. four weeks after the shock). The response of flows to U.S.
14
We examined whether the statistical significance of our results changes if we instead used (Driscoll and Kraay, 1998) standard errors, which are robust to
very general forms of cross-sectional as well as temporal dependence. Those confidence bands are very close to those obtained through double clustering (refer
to discussion in Section 3.4).
15
Fig. 2 and the others presented in this paper show the effects of a 10 b.p. positive shock. The effects of a negative shock are merely the opposite of those
represented in the figures.
8
Fig. 3. Investors’ allocations into mutual funds after a 10 bp INFO news shock. Notes: Dashed black lines report point estimates. Deep and shallow gray areas
are 90% and 68% confidence bands. X-axes denote the response horizon (in weeks), with 0 being the week of the FOMC meeting. Y-axes denote the
magnitude of the response (in percentage points). Responses are obtained estimating the ck coefficients from Eq. 4. Estimates are normalized to show
responses to .a 10 b.p. shock.
funds is similar but less precisely estimated. Overall, these results confirm earlier findings that flows to EMs are particularly
sensitive to U.S. monetary policy (Chari et al., 2020).
3.2. The effects of information news shocks
The responses to Fed information shocks indicate that the effects differ from those of pure monetary policy shocks in sev-
eral important ways (Fig. 3). First, positive information news shocks lead investors to modestly rebalance their portfolios into
equity funds. We estimate flows to U.S. and Global equity funds to be respectively about 0.2 and 0.1 p.p. higher four weeks
after a 10 b.p. shock, while the effect on flows to EMs equity funds is flat. Second, positive information news shocks lead
investors to redeem shares from U.S. bond funds. The effect is exactly the opposite of that on U.S. equity flows, 0.2 p.p. after
four weeks. The effect on global bond fund flows is positive although small and only marginally significant, while flows to
EMs bond funds display a negative but insignificant response. Overall, these results suggest that investors revise upwards
their belief about the state of the economy and thus increase equity investment, particularly in the U.S., after positive infor-
mation news shocks.
Taken together, our estimates indicate that it is crucial to disentangle pure monetary policy shocks from information
news shocks in order to understand the response of mutual funds flows to Fed monetary policy. While an increase in interest
rates due to a pure monetary policy shock generally decreases allocations into investment funds, and particularly to EM
funds, an increase in interest rates due to expectation of higher growth (information news shocks) does not lead to outflows
from EM funds and may even increase allocations into some types of funds. As we show next, relying on identification strate-
gies that do not distinguish pure monetary policy from information news shocks may lead to the conclusion that Fed mon-
etary policy only has small effects on fund flows.
9
Fig. 4. Investors’ allocations into mutual funds after a 10 bp MP ‘‘news” (without separating pure monetary policy and information news shocks). Notes:
Dashed black lines report point estimates. Deep and shallow gray areas are 90% and 68% confidence bands. X-axes denote the response horizon (in weeks),
with 0 being the week of the FOMC meeting. Y-axes denote the magnitude of the response (in percentage points). Responses are obtained estimating the bk
coefficients from an alternative specification of Eq. 4 in which the et and it variables are replaced by a composite high frequency measure of monetary policy
news. Estimates are normalized to show responses to a one standard deviation shock.
3.3. Results using alternative identification methodologies
Our identification of monetary policy shocks and information effects differs from previous literature, which typically
relies on a structural vector autoregression (VAR) approach (among others Christiano et al., 1999; Romer and Romer,
2004) or on the change in different sets of interest rates within a tight window around the release of the FOMC monetary
policy decision (see for instance Nakamura and Steinsson, 2018) or stock prices and interest rates (see for instance
Jarociński and Karadi, 2020) to identify monetary policy news shocks. In this section, we investigate how our results would
change if we followed these alternative identification methodologies.
3.3.1. Results without separating monetary policy and information news shocks
Following Nakamura and Steinsson (2018), we first use a composite measure of monetary policy news obtained as the
first principal component of the unanticipated change over the 30-min window around FOMC monetary policy decisions
of (i) the current federal funds rate, (ii) the federal funds rate expected following the subsequent FOMC meeting, and (iii)
the eurodollar future rates expected 2, 3, and 4 quarters ahead. Relative to a measure of monetary policy identified relying
only on the high-frequency change in the current federal funds rate, this approach has the advantage of accounting for for-
ward guidance shocks, which have become important in the post-GFC period. We then re-estimate Eq. (4) replacing the et and
it variables with this alternative high-frequency measure of monetary policy news, which does not distinguish between
monetary policy shocks and information effects. Although the new estimates look qualitatively similar to those obtained
for our pure monetary policy shock variable, they are much milder and less statistically significant (Fig. 4). This is because
these new estimates mix together the often opposite effects of pure monetary policy and information news shocks. Hence,
using this alternative approach may lead to the incorrect inference that Fed monetary policy has only mild effects on inter-
national mutual fund investment.
10
3.3.2. Alternative approaches to derive pure monetary policy and information news shocks
How do our results compare to those that would be obtained using the approach of Jarociński and Karadi (2020) to
decompose monetary policy news between pure monetary policy shock and information news shock components? A key
difference between our approach and theirs lies in the fact that we exploit movements in the entire yield curve to derive
our shock variables while these authors only use changes in the policy rate expected over a short horizon. This results in
their series displaying quite low variability during ZLB periods. Another key difference is that Jarociński and Karadi
(2020) rely on unusual responses of the stock market to derive the information news shock series. Comparing the two sets
of shock series—ours and theirs—we find that the two pure monetary policy shock variables have a fairly high correlation,
while the two information news shock variables do not.
It follows that estimates of the effect of pure monetary policy shocks on mutual fund flows obtained using Jarociński and
Karadi (2020) series are similar to ours while those using the information news shocks are not (Figs. A1 and A2 in the Appen-
dix). In particular, when we use their information news shock series, we obtain responses that tend to be negative, albeit not
statistically significant, for all types of funds, including U.S. and global equity funds. This runs against results from closed-
economy models and against the presence of a Fed information effect.
3.4. Robustness checks and alternative specifications
Before proceeding further, we check the robustness of our baseline results to a battery of different specifications. Par-
ticularly, we check robustness to (i) using different lag structures and omitting forward shock variables and NAV returns
from the set of controls, (ii) taking the 2-year or the 10-year rate, instead of the 5-year rate, as the reference rate in the
construction of the shock series, (iii) the inclusion of several control variables, including U.S. macroeconomic surprises (iv)
excluding particular sets of funds, (v) estimating the results over different time samples, and (vi) using Driscoll-Kraay
rather than two-way clustered standard errors. Albeit quantitatively different in some cases, the new estimates are
broadly in line with, and not statistically different from, our baseline. These robustness checks are explained in detail
in the Online Appendix.
4. Transmission channels
In this section, we explore portfolio rebalancing, the response of the USD exchange rate, and variation in risk appetite and
uncertainty as potential transmission channels for our baseline reduced form results. According to the portfolio rebalancing
channel, investors rebalance towards higher-yielding assets following monetary policy decisions that lower the yields on
safe government bonds. Fratzscher et al. (2018, 2020) find this channel to have led investors to rebalance their portfolios
towards EM assets after the adoption of unconventional monetary policies in the U.S. and other advanced markets in the
wake of the GFC. Hence, one possible explanation for our results is that the Fed’s pure monetary policy shocks increase (real)
interest rates in the U.S., while information news shocks do not. This would explain why we find that investors react to pure
monetary policy shocks by rebalancing their portfolios away from EMs assets, while the same does not happen following
information news shocks.
Another possibility is that pure monetary policy and information news shocks affect investors’ appetite for risk differ-
ently. While an increase in interest rates that is due to a change in, say, the monetary policy reaction function may well
heighten risk aversion, a ‘‘benign” increase in interest rates due to higher growth expectations should not increase it.
Bruno and Shin (2015) formalize the risk-taking channel of monetary policy in the context of cross-border bank lending
flows through changes in the interest rate, which affect banks’ leverage. Forbes and Warnock (2012) find that global fac-
tors, especially global risk, are significantly associated with extreme capital flow episodes. Ahmed and Zlate (2014) find an
important role of global risk appetite, proxied through the VIX index, in driving portfolio flows to EMs. Here, we investi-
gate how Fed shocks affect the VIX and whether that could explain our results. Moreover, since relying on the VIX alone
does not allow distinguishing between changes in the price of risk (risk aversion) from changes in the quantity of risk
(uncertainty), we also explore the effect of Fed shocks on the risk aversion and uncertainty indexes constructed by
Bekaert et al. (2021). This also allows us to investigate the importance of the Fed’s information effect, which should affect
the level of risk.
We furthermore consider the response of the USD exchange rate as another possible channel. To the extent that Fed
policy affects the international price of the USD, this may have implications for investment decisions as it would
induce a mechanical change in the dollar price of foreign assets. For instance, Bruno and Shin (2015) find that a con-
tractionary U.S. monetary policy shock leads to a drop in cross-border banking capital flows through an appreciation of
the USD.
To empirically investigate these potential transmission channels, we opt for a vector autoregression (VAR) approach. The
advantage of the VAR is that we can estimate the responses of several financial variables to our shock series, as well as the
feedback effects that such responses have on fund flows, within a unified framework. We consider the nominal and real
(TIPS) interest rate on U.S. 5-year government bonds, the VIX index, and the USD exchange rate (measured through the
DXY index). As for fund flows, we only include those to EM equity and bond funds in order to keep the VAR parsimonious.
11
Fig. 5. Financial variables and allocations into EM funds after a pure MP shock. Notes: Dashed black lines report point estimates. Deep and shallow gray
areas are 90% and 68% confidence bands. X-axes denote the response horizon (in weeks), with 0 being the week of the FOMC meeting. Y-axes denote the
magnitude of the response (in percentage points). Estimates are obtained estimating a VAR specification including the pure monetary policy and
information news shock variables, ordered first and second respectively, the 5-year nominal and real rates and the VIX, all in first differences, the dollar
index (DXY), in log first-differences and EMs equity and bond flows, in percent of assets. The VAR includes 5 lags of the endogenous variables. Confidence
bands are obtained through bootstrapping (500 replications). Estimates are normalized so as to show responses to a 10 b.p. shock.
Exploiting the fact that our monetary policy variables are exogenous shocks, we identify the VAR through a Cholesky decom-
position, ordering the pure monetary policy and information news shocks first and second, respectively.16
Figs. 5 and 6 show the VAR IRFs to the pure monetary policy and information news shocks, respectively. The results con-
firm the finding that pure (contractionary) monetary policy shocks lead to negative flows to EM bond and equity funds, while
information news shocks do not. Looking at the response of the other variables, we find that pure monetary policy and infor-
mation news shocks alike drive up both nominal and real interest rates. This suggests that the main driver of our baseline
results is unlikely to be the portfolio rebalancing channel. The importance of the exchange rate channel can instead be exam-
ined by looking at the response of the DXY. Both monetary and information news shocks lead to an increase in the index
(appreciation of the USD), although the response to information news shocks is sluggish and less precisely estimated. This
may suggest that the exchange rate channel can explain some of our baseline results.17
What is really striking, however, is the differential response of the VIX index, which rises after a pure monetary policy
shock and falls following an information news shock. The magnitude of the impact responses are comparable, but the
decrease following information news shocks is short-lived, while the increase following monetary policy shocks is more per-
sistent. We check whether these effects are driven by changes in the price or the quantity of risk by estimating another VAR
specification where the VIX index is replaced by the risk aversion and the uncertainty indexes of Bekaert et al. (2021). The
IRFs derived for EMs fund flows, bond yields and the DXY are very close to those estimated above, hence we only report IRFs
for the risk aversion and uncertainty indexes (Fig. 7). In line with the theory of the risk-taking channel of monetary policy,
16
The identifying assumption is that the shock variables can have contemporaneous effects on the other variables in the VAR but that the opposite does not
hold. As shown by Christiano et al. (1999), the ordering of the variables after the shock of interest does not affect results. Changing the ordering of the shock
variables among themselves (information news shock first and pure monetary policy shock second) does not affect our conclusions. The VAR is estimated at the
weekly frequency, including 5 lags of the endogenous variables. Confidence bands are obtained through bootstrapping (500 repetitions).
17
On the other hand, however, the differential response of the exchange rate may be the flip side of the different effects of pure monetary policy and
information news shocks on fund flows, thus caution is warranted.
12
Fig. 6. Financial variables and allocations into EM funds after an INFO shock. Notes: See notes to Fig. 5.
the price of risk (risk aversion) jumps by 3–5% after a 10 b.p. pure monetary policy shock, while the uncertainty index (the
quantity of risk) is flat. On the other hand, after a 10 b.p. information news shock, the price of risk is broadly unchanged but
uncertainty sharply decreases, by about 2% on impact. This latter response suggests that the presence of the Fed information
effect, with positive news growth decreasing uncertainty, is also important. In the Appendix, we substantiate the link
between changes in the VIX and EM fund flows by showing the feedback effects of an increase in the VIX index (and of
the DXY index) on flows to EM funds (Fig. A3). The results echo the literature in that we find an increase in the VIX sharply
decreases flows to bond and to a lesser extent equity funds ([see Bruno and Shin, 2015, for example).
As a further look, we estimate the response to pure monetary policy and information news shocks of allocations into dif-
ferent types of U.S. bond funds. The new estimates (Figs. A4 and A5 in the Appendix) suggest that funds investing in long-
term Treasuries are particularly sensitive to changes in Fed monetary policy. Investors sharply increase allocations into this
type of fund after a pure tightening shock, likely as a response to the increase in risk aversion. On the other hand, investors
decrease allocations into these funds after a positive information shock. This may be because positive information shocks
signal higher growth or inflation, which is typically associated with a higher interest rates environment. We also find that
high-yield bond funds receive inflows following positive information shocks, further supporting the role of the risk-taking
channel in driving our results.
We conclude that an increase in interest rates that is unpredictable from changes in economic fundamentals (pure mon-
etary policy shock) increases risk aversion and leads investors to shed high-risk assets. On the other hand, an increase in
interest rates that reflects the Fed’s perception of improved U.S. fundamentals tends to decrease uncertainty, neutralizing
the detrimental effect of higher interest rates for EMs fund flows and leading investors to increase exposure to assets the
performance of which is tightly linked to developments in U.S. growth.
13
Fig. 7. Effects of Fed’s shocks on risk aversion and uncertainty. Notes: Dashed black lines report point estimates. Deep and shallow gray areas are 90% and
68% confidence bands. X-axes denote the response horizon (in weeks), with 0 being the week of the FOMC meeting. Y-axes denote the magnitude of the
response (in percent). Estimates are obtained estimating a VAR specification including the monetary policy and information news shock variables, ordered
first and second respectively, the 5-year nominal and real rates, all in first differences, the dollar index (DXY), the risk aversion and the uncertainty indexes
of Bekaert et al. (2021), in log first differences, and EMs equity and bond flows, in percent of assets. The VAR includes 5 lags of the endogenous variables.
Confidence bands are obtained through bootstrapping (500 replications). Estimates are normalized to show responses to .a 10 b.p. shock.
5. Extensions
5.1. Asymmetric behavior of U.S. and EM investors
Previous literature has shown that when investors become more risk averse they tend to rebalance their portfolios away
from foreign and into domestic assets.18 Here, we test whether domestic and foreign investors respond differently to Fed
shocks exploiting information on funds’ domiciles, which we see as a proxy for the country of origin of the underlying investors.
We focus on funds investing in EMs and compare the response of flows to those domiciled in the U.S. versus those domiciled in
EMs themselves. For the estimation, we go back to local projections, which, absent feedback effects to be explored, is preferred
on the grounds of greater flexibility relative to the VAR approach (see discussion in Section 3.4).
Results are shown in Fig. 8. While U.S. investors dis-invest from EM equity funds after a pure monetary policy shock
(Panel A1), EM investors sharply increase flows to the same type of funds (Panel A2), therefore mitigating the reduction
in foreign capital. We find a similar tendency for bond funds, although the response of EM investors is much weaker and
less persistent (Panels A3 and A4). In contrast, after an information news shock, U.S. investors increase flows to EM bond
funds, while EM investors decrease them (Panels B3 and B4). Flows to EM equity funds do not exhibit significant responses
to information news shocks.
We then perform a similar exercise on flows to U.S. funds (Fig. A6 in the Appendix). We find that flows to funds investing
in U.S. equities that are domiciled abroad increase sharply after a positive information news shock. This increase is much
larger than the increase in U.S. equity funds domiciled domestically (likely to be used by domestic investors). In the next
section, we explore what these results imply for cross-border flows.
18
See Coeurdacier and Rey (2013) for a comprehensive review of the early literature.
14
Fig. 8. Effect of Fed’s shocks on allocations into EM funds accounting for domicile. Notes: See notes to Figs. 2 and 3. Estimates are obtained on the subsample
of funds domiciled in the U.S. (Panels A1, A3, B1 and B3) and EMs (Panels A2,. A4, B2 and B4).
5.2. Zooming in on EM bond funds
Here, we zoom in on flows to EM bond funds and investigate heterogeneities in responses to Fed shocks across different
types of funds, exploiting the richness of the EPFR fund-level dataset. Besides the geographical focus, bond funds differ in
several dimensions. Some can invest in any type of bond, but others are constrained to invest only in certain bonds. Here,
we consider three key dimensions that may impact how bond investors respond to U.S. monetary policy news: the issuer
of the bond, its credit quality, and the currency of denomination – whether ‘‘hard” (USD, GBP, EUR, CHF or YEN) or local
(any EM currency).
If changes in risk aversion are the main factor driving the effect of pure monetary policy shocks on EM bond fund flows,
responses may be stronger for funds investing in bonds considered to be riskier. Credit quality is a direct indication of risk,
but other factors may also matter. For instance, corporate bonds may be considered to be riskier than sovereign bonds, even
when they have the same credit rating. Similarly, although foreign investors holding a bond denominated in hard currency
are shielded from exchange rate fluctuations, hard currency bonds may be perceived to be riskier than local currency ones
given the currency mismatch between the income and liability of the bond issuer. Hence, they may be more exposed to
changes in risk aversion.
To investigate whether bond characteristics matter, we define funds that are restricted to invest in bonds that (i) are
issued by the government, (ii) have a credit rating of BBB- or above (investment-grade bonds), and (iii) are denominated
in hard currency to be our baseline group and compare flows to this group to those to other groups having different char-
acteristics. Empirically, we estimate the following equation:
X
k
F i;tþj
X3 X24
j¼0 F
¼ bk et þ ck it þ ðfc;k xci et þ pc;k xci it Þ þ rkl i;tl þ /kl ni;tl þ di þ ei;t ð5Þ
Ai;t c¼1 l¼1
Ai;tl
15
Fig. 9. Effect of Fed shocks on allocations into EM funds accounting for fund characteristics. Notes: Panels A1 and B1 report responses of flows into funds
with mandate to invest only in (i) sovereign, (ii) investment-grade, and (iii) hard currency bonds (obtained estimating the bk and ck coefficients from Eq. 5).
Panels A2-A4 (B2-B4) report responses when each of the investment mandates (i)-(iii) is relaxed one at a time (obtained as linear combinations of the bk and
fi;k (ck and p1;k ) coefficients). Dashed black lines report point estimates. Deep and shallow gray areas are 90% and 68% confidence bands. X-axes denote the
response horizon (in weeks), with 0 being the week of the FOMC meeting. Y-axes denote the magnitude of the response (in percentage points). Estimates
are normalized to show responses to .a 10 b.p. shock.
where the xci s are three dummy variables for funds investing in, respectively: (i) either both sovereign and corporate bonds
or corporate bonds only (c ¼ 1), (ii) either both investment grade and high-yield bonds or high-yield bonds only (c ¼ 2), (iii)
either both hard and local currency bonds or hard currency bonds only (c ¼ 3), and the rest is as in Eq. 4.
Fig. 9 shows the new estimates. Panels A1 and B1 show responses of flows into our baseline group of funds. Such
responses are simply given by the bk and ck coefficients. Panels A2 and B2 show flows to funds that do not necessarily have
to invest in sovereign bonds, but are still constrained to invest in investment-grade and hard currency bonds (estimates
obtained as bk þ f1;k and ck þ p1;k respectively). Panels A3 and B3 focus on funds that must invest in bonds that are issued
by the sovereign and are denominated in hard currency but either do not have a credit quality restriction or can only invest
in high-yield bonds (those with credit rating below BBB-, estimates given by bk þ f2;k and ck þ p2;k ). Finally, Panels A4 and B4
show responses when the currency denomination constraint is relaxed (bk þ f3;k and ck þ p3;k ).
This analysis points to credit quality and currency denomination as having an important role in determining the
response of investors to pure monetary policy shocks. Outflows from funds investing in our baseline group of bonds (sover-
eign, investment-grade and hard currency) are half as large as those from funds having the same characteristics but that are
not constrained to invest in investment grade bonds. By contrast, outflows from funds that are not constrained to invest in
hard currency bonds are not significant, suggesting that local currency bonds may be less exposed to monetary policy shocks
in the U.S. This has important policy implications since local currency bonds also have the advantage that they shield the
issuer from currency mismatches and therefore may be a good instrument to attract foreign capital while at the same time
reducing risks stemming from exchange rate fluctuations and changes in risk aversion. Looking at responses to information
news shocks, we do not find significant differences across funds based on investment mandate.
16
5.3. Other extensions
We also perform some further extensions. First, we explore heterogeneities in responses across U.S. equity funds. We find
that responses of flows into those that invest in the broad stock market are similar to our baseline results. On the other hand,
we find that funds investing in (i) consumer goods, (ii) financial, and (iii) utility stocks receive large inflows following an
increase in interest rates that is driven by a positive information news.
Second, we explore heterogeneities in responses across EMs funds with different geographical focus and find that out-
flows after a pure monetary policy shock are fairly generalized. Third, we use country-level data on allocations by investment
funds into individual countries to investigate the implications of our results for cross-border portfolio capital flows. This
analysis confirms that tightening pure monetary policy shocks lead to a generalized decrease in foreign capital. This is shar-
per for EMs but it is also present for the U.S. and other advanced economies, consistent with an increase in risk aversion. On
the other hand, positive information shocks lead to sharp inflows of foreign capital into U.S. equity assets. These extensions
are explained in greater detail in the Online Appendix.
6. Conclusions
We contribute to the vast literature on the effects of U.S. monetary policy on international capital flows, employing a
novel identification procedure to decompose observed U.S. monetary policy surprises into pure monetary policy shock
and information news shock components. We find that an increase in interest rates driven by a pure monetary policy shock
leads to persistent outflows from EMs and to a lesser extent global mutual funds. On the other hand, when rates increase
following a positive information news shock EMs funds do not suffer outflows. Positive information shocks instead lead
investors to reallocate capital out of safe U.S. government bonds funds and into U.S. and global equity funds as well as
high-yield bond funds. We also show that failing to decompose monetary policy surprises into a pure monetary policy
and an information news shock component would lead to the incorrect conclusion that U.S. monetary policy has small effects
on international mutual fund flows. We attribute our main results to the presence of the Fed information effect and to the
operation of the risk-taking channel of monetary policy. Pure monetary policy shocks tend to tighten financial conditions and
increase risk aversion, while positive information news shocks lower uncertainty.
Data availability
Data will be made available on request.
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. Appendix
A.1. Additional tables and figures
Table A1
Monetary policy shock and information news shock test.
Blue Chip GDP forecasts next 3Q

2002–2017 2009–2017
Pure monetary policy shock (BRW) 0.0712 0.151
(0.194) (0.206)
Information news shock 0.306* 0.590***
(0.168) (0.187)
Observations 106 62
R-squared 0.033 0.145
17
Fig. A1. Jarociński and Karadi (2020) identification - pure MP shock. Notes: Solid black lines report full sample point estimates. Deep and shallow gray areas
are their 90% and 68% confidence bands. Dot-dashed lines report point estimates obtained excluding the GFC period. X-axes denote the response horizon (in
weeks), with 0 being the week of the FOMC meeting. Y-axes denote the magnitude of the response (in percentage points). Responses are obtained
estimating the bk coefficients from an alternative specification of Eq. 4, in which et and it are replaced with Jarociński and Karadi (2020) pure monetary
policy and information news shock series. Estimates are normalized to show responses to a 10 b.p. shock.
Fig. A2. Jarociński and Karadi (2020) identification - INFO news shock. Notes: See notes to Fig. A1. Point estimates and confidence bands refer to the ck
coefficients.
18
Fig. A3. Effects of VIX and USD exchange rate on allocations into EMs funds. Notes: Black solid lines and deep and shallow gray areas respectively denote
point estimates and their 68% and 90% confidence intervals. X-axes denote the response horizon (in weeks), with 0 being the week of the FOMC meeting. Y-
axes denote the magnitude of the response (in percentage points). Responses are obtained estimating a VAR (see Section 4 for details).
Fig. A4. Effects of pure MP shocks on allocations into U.S. bond funds. Notes: See notes to Figs. 2. Estimates are obtained on different subsamples of .U.S.
bond funds.
19
Fig. A5. Effects of INFO news shocks on allocations into U.S. bond funds. Notes: See notes to Fig. 3. Estimates are obtained on different subsamples of .U.S.
bond funds.
Fig. A6. Effects of Fed’s shocks on allocations into U.S. funds accounting for domicile. Notes: See notes to Figs. 2 and 3.
20
Appendix B. Supplementary material
Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.jimonfin.
2022.102676.
References
Acosta, M., 2021. The perceived causes of monetary policy surprises. Working Paper, Columbia University.
Ahmed, S., Zlate, A., 2014. Capital flows to emerging market economies: A brave new world? J. Int. Money Finance 48, 221–248.
Auerbach, A.J., Gorodnichenko, Y., 2012. Measuring the output responses to fiscal policy. Am. Econ. J.: Econ. Policy 4, 1–27.
Bauer, M., Swanson, E.T., 2021. An alternative explanation of the ‘‘fed information effect. http://www.socsci.uci.edu/ swanson2/index.html.
Beber, A., Brandt, M.W., 2009. When it cannot get better or worse: The asymmetric impact of good and bad news on bond returns in expansions and
recessions. Review of Finance 14, 119–155.
Bekaert, G., Engstrom, E.C., Xu, N.R., 2021. The time variation in risk appetite and uncertainty. Manage. Sci.
Boyd, J.H., Hu, J., Jagannathan, R., 2005. The stock market’s reaction to unemployment news: Why bad news is usually good for stocks. The Journal of Finance
60, 649–672.
Bruno, V., Shin, H.S., 2015. Capital flows and the risk-taking channel of monetary policy. Journal of Monetary Economics 71, 119–132.
Bu, C., Rogers, J.H., Wu, W., 2021. A unified measure of fed monetary policy shocks. Journal of Monetary Economics 118.
Burger, J., Warnock, F., Warnock, V., 2018. The effects of u.s. monetary policy on emerging market economies’ sovereign and corporate bond markets. In:
Monetary Policy and Global Spillovers: Mechanisms, Effects, and Policy Measures. Central Bank of Chile volume 25, pp. 49–96.
Calvo, G.A., Leiderman, L., Reinhart, C.M., 1993. Capital inflows and real exchange rate appreciation in latin america: the role of external factors. I.M.F. Staff
Papers 40, 108–151.
Campbell, J.R., Evans, C.L., Fisher, J.D., Justiniano, A., Calomiris, C.W., Woodford, M., 2012. Macroeconomic effects of federal reserve forward guidance [with
comments and discussion]. Brookings Papers on Economic Activity, pp. 1–80.
Chari, A., Stedman, K.D., Lundblad, C., 2020. Taper tantrums: Quantitative easing, its aftermath, and emerging market capital flows. The Review of Financial
Studies.
Christiano, L.J., Eichenbaum, M., Evans, C.L., 1999. Monetary policy shocks: What have we learned and to what end? Handbook of macroeconomics 1, 65–
148.
Cieslak, A., Schrimpf, A., 2019. Non-monetary news in central bank communication. Journal of International Economics 118, 293–315.
Coeurdacier, N., Rey, H., 2013. Home bias in open economy financial macroeconomics. Journal of Economic Literature 51, 63–115.
Converse, N., Levy-Yeyati, E., Williams, T., 2020. How etfs amplify the global financial cycle in emerging markets. International Finance Discussion Papers.
Dahlhaus, T., Vasishtha, G., 2020. Monetary policy news in the us: Effects on emerging market capital flows. J. Int. Money Finance 109, 102251.
Driscoll, J., Kraay, A., 1998. Consistent covariance matrix estimation with spatially dependent panel data. Rev. Econ. Stat. 80, 549–560.
Fama, E.F., MacBeth, J.D., 1973. Risk, return, and equilibrium: Empirical tests. Journal of Political Economy 81, 607–636.
Faust, J., Rogers, J.H., Wang, S.-Y.B., Wright, J.H., 2007. The high-frequency response of exchange rates and interest rates to macroeconomic announcements.
Journal of Monetary Economics 54, 1051–1068.
Forbes, K., Warnock, F.E., 2012. Capital flow waves: Surges, stops, flight, and retrenchment. Journal of International Economics 88, 235–251.
Fratzscher, M., 2012. Capital flows, push versus pull factors and the global financial crisis. Journal of International Economics 88, 341–356.
Fratzscher, M., Lo Duca, M., Straub, R., 2018. On the international spillovers of us quantitative easing. Econ. J. 128, 330–377.
Gilbert, T., Scotti, C., Strasser, G., Vega, C., 2017. Is the intrinsic value of macroeconomic news announcements related to their asset price impact? Journal of
Monetary Economics 92, 78–95.
Gürkaynak, R.S., Sack, B., Swanson, E., 2005. The sensitivity of long-term interest rates to economic news: Evidence and implications for macroeconomic
models. American economic review 95, 425–436.
Hansen, S., McMahon, M., Tong, M., 2019. The long-run information effect of central bank communication. Journal of Monetary Economics 108, 185–202.
Hoek, J., Kamin, S.B., Yoldas, E., forthcoming. When is bad news good news? u.s. monetary policy, macroeconomic news, and financial conditions in
emerging markets. J. Int. Econ..
Iacoviello, M., Navarro, G., 2019. Foreign effects of higher us interest rates. J. Int. Money Finance 95, 232–250.
Jarociński, M., Karadi, P., 2020. Deconstructing monetary policy surprises—the role of information shocks. Am. Econ. J.: Macroecon. 12, 1–43.
Jordà, Ò., 2005. Estimation and inference of impulse responses by local projections. Am. Econ. Rev. 95, 161–182.
Jordà, Ò., Taylor, A.M., 2016. The time for austerity: estimating the average treatment effect of fiscal policy. Econ. J. 126, 219–255.
Krishnamurthy, A., Vissing-Jorgensen, A., 2011. The effects of quantitative easing on interest rates: channels and implications for policy. Technical Report
National Bureau of Economic Research.
Liu, Z., Spiegel, M., & Zhang, J. (2020). Capital controls and income inequality. Federal Reserve Bank of San Francisco Working Paper 2020–14.
Lunsford, K., 2020. Policy language and information effects in the early days of federal reserve forward guidance. Am. Econ. Rev., 2899–2934
Miranda-Agrippino, S., 2016. Unsurprising shocks: information, premia, and the monetary transmission. Bank of England Working Paper 626.
Miranda-Agrippino, S., Nenova, T., 2021. A tale of two global monetary policies. CEPR Working Paper DP16485.
Nakamura, E., Steinsson, J., 2018. High-frequency identification of monetary non-neutrality: the information effect. Q. J. Econ. 133, 1283–1330.
Neely, C.J. et al. (2010). The large scale asset purchases had large international effects. Federal Reserve Bank of St. Louis, Research Division St Louis.
Paul, P., 2019. The time-varying effect of monetary policy on asset prices. Rev. Econ. Stat., 1–44
Pinchetti, M., & Szczepaniak, A. (2021). Global spillovers of the fed information effect. Bank of England Staff working paper 952.
Ramey, V.A., Zubairy, S., 2018. Government spending multipliers in good times and in bad: evidence from us historical data. J. Polit. Econ. 126, 850–901.
Romer, C.D., Romer, D.H., 2000. Federal reserve information and the behavior of interest rates. Am. Econ. Rev. 90, 429–457.
Romer, C.D., Romer, D.H., 2004. A new measure of monetary shocks: Derivation and implications. Am. Econ. Rev. 94, 1055–1084.
Romer, C.D., Romer, D.H., 2017. New evidence on the aftermath of financial crises in advanced countries. Am. Econ. Rev. 107, 3072–3118.
Swanson, E.T., Williams, J.C., 2014. Measuring the effect of the zero lower bound on medium-and longer-term interest rates. Am. Econ. Rev. 104, 3154–3185.
Warnock, F., Warnock, V.C., 2009. International capital flows and u.s. interest rates. J. Int. Money Finance 28, 903–919.
21

1 s2.0 S0261560622000791 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0261560622000791 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0261560622000791 Main

Uploaded by

Copyright:

Available Formats

Journal of International Money and Finance 127 (2022) 102676

Contents lists available at ScienceDirect

Journal of International Money and Finance

The effects of U.S. monetary policy on international mutual fund

2. Dataset and methodology

2.1. U.S. monetary policy

DRi;t ¼ ai þ ci et þ i;t ð1Þ

DRi;t ¼ hi þ bi DR5;t þ ni;t ð2Þ

DRi;t ¼ di þ epure ^i þ v i;t ;

2.2. Investment funds

Pure monetary policy shocks information news shocks

2.3. Other variables

Mandate # funds % domiciled Assets

2.4. Econometric framework

3.1. The effects of monetary policy shocks

3.2. The effects of information news shocks

3.3. Results using alternative identification methodologies

3.4. Robustness checks and alternative specifications

5.1. Asymmetric behavior of U.S. and EM investors

5.2. Zooming in on EM bond funds

5.3. Other extensions

Data will be made available on request.

Declaration of Competing Interest

A.1. Additional tables and figures

Blue Chip GDP forecasts next 3Q

Appendix B. Supplementary material

You might also like