Finance Research Letters 55 (2023) 103931

Contents lists available at ScienceDirect

Finance Research Letters

journal homepage: www.elsevier.com/locate/frl

The impact of EPU spillovers on the bond market volatility:

Global evidence
Yuting Gong a, Xiao Li b, Wenjun Xue c, *
a
Department of Economics and Finance, SHU-UTS SILC Business School, Shanghai University, Shanghai 201800, China
b
Department of Finance, Massry Center for Business, University at Albany, SUNY, Albany, NY 12222, United States
c
Department of Economics and Finance, SHU-UTS SILC Business School, Shanghai University, Shanghai, 201800, China

A R T I C L E I N F O A B S T R A C T

JEL Classifications: This paper studies the effect of economic policy uncertainty (EPU) spillovers from other countries
G12 on local bond market volatility. Using multivariate quantile model (White et al., 2015), we
G15 develop a country-specific EPU spillover measure for 23 economies from 2003 to 2019. We find
C10
that EPU spillovers have a significantly positive effect on local bond market volatility. This effect
F30
becomes stronger if the spillovers are from developed markets and when the spillovers are
Keywords:
measured during financial crises. Recognizing the relation between EPU spillovers and bond
EPU spillovers
volatility can motivate policy makers to closely monitor foreign EPU and take actions to alleviate
Country-level bond market volatility
Multivariate quantile model the detrimental influence when foreign EPU rises.
International asset pricing

1. Introduction

Researchers have explored the impact of economic policy uncertainty (EPU) on many aspects of the economy.1 In the niche of
financial market volatility, the literature has documented EPU’s impact on stocks (e.g., Ma et al., 2020), commodity (e.g., Fang et al.,
2018), foreign exchange (e.g., Chen et al., 2020), and cryptocurrency (e.g., Yen and Cheng, 2021). Surprisingly, bond market volatility
receives very limited attention. However, the year 2022 has strongly announced the importance of bond volatility, as almost all major
economies witnessed turbulence in the bond market when international uncertainty looms large, which necessitates further explo­
ration along the direction of bond market volatility.2
Another angle that also receives little exploration is how foreign EPU affects local financial markets (EPU spillovers). One challenge
this question poses is that, as policy making is internationally connected (e.g., Balli et al., 2017), foreign EPUs inevitably intervolve
with local EPU, especially when the local economy is influential (e.g., the US) or when the countries are closely connected (e.g., the
European Union countries). This challenge renders directly regressing local variables on foreign EPUs questionable (He et al., 2020)
since the effect, if any, cannot be clearly attributed to either domestic or foreign EPUs.
Therefore, this paper attempts to study the effect of EPU spillovers on bond volatility, while tackling the challenge mentioned
above. We attempt to make three contributions. First, recognizing the interconnectedness of EPUs among countries, we apply the

* Corresponding author.
E-mail address: [email protected] (W. Xue).
1
Please see Al-Thaqeb and Algharabali (2019) for a comprehensive review.
2
For example, the “mini-budget” of the UK government not only triggered a crash in the local bond market, but also caused the US bond market’s
volatility to rise. (see Adinolfi, 2022)

https://doi.org/10.1016/j.frl.2023.103931
Received 2 February 2023; Received in revised form 3 April 2023; Accepted 20 April 2023
Available online 23 April 2023
1544-6123/© 2023 Elsevier Inc. All rights reserved.
Y. Gong et al. Finance Research Letters 55 (2023) 103931

multivariate quantile model (White et al., 2005) to explicitly model the relation between local and foreign EPUs and estimate country-
specific EPU spillovers. This method allows a clean attribution of effects to EPU spillovers, if any. Second, our empirical evidence shows
that EPU spillovers are a novel contributor that has a significant and positive impact on local bond market volatility, in addition to local
EPU and other economic, financial, and political factors documented before. We also find that, this effect becomes stronger when the
spillovers are from developed countries, and when the spillovers are measured during crisis period. Finally, our findings suggest that
policy makers must recognize the contagion of EPU spillovers and take effective actions to alleviate the negative influence on bond
markets when foreign EPU rises.

2. Methodology and data

2.1. The measure for EPU spillovers

We use the multivariate quantile model (MVMQ) of White et al. (2015) to estimate the EPU spillovers. This approach has the
following desirable features. First, this method recognizes the connectedness of EPUs among countries and distinguishes the mutual
effect between local EPU and foreign EPUs. Second, the quantile regression estimates are known to be robust to outliers, which is
particularly desirable in the applications of financial data. Third, the quantile regression is a semi-parametric technique imposing
minimal distributional assumptions on the underlying data generating process. Fourth, the multivariate quantile framework allows a
direct measure of quantile dependence among the random variables in a country-specific fashion.
Particularly, our model takes that form of:
⃒ ⃒ ⃒ ⃒
q1,t (τ) = α1 + β1,1 ⃒⃒Y1,t− 1 ⃒⃒ + β1,2 ⃒⃒Y2,t− 1 ⃒⃒ + γ1,1 q1,t− 1 (τ) + γ1,2 q2,t− 1 (τ)
(1)
q2,t (τ) = α2 + β2,1 ⃒Y1,t− 1 ⃒ + β2,2 ⃒Y2,t− 1 ⃒ + γ2,1 q1,t− 1 (τ) + γ2,2 q2,t− 1 (τ)

where qi,t(τ), i = 1, 2 is the τth conditional quantile of Yi,t − 1 given the information up to t − 1. In our model, Y1,t − 1is the first difference
of EPU for country i at time t-1 and Y2,t − 1 is the GDP weighted first difference of EPU of the other N-1 countries at time t-1. α1and α2 are
the corresponding intercepts in Eq. (1) and βi,j, i = 1, 2j = 1, 2 are the coefficients of lags |Yi,t − 1|. γi,j, i = 1, 2j = 1, 2 represent the
measure of tail codependence between the two random variables.3
Our model allows us to directly measure EPU spillovers at a certain quantile for a specific country. For instance, given qi,t − 1(τ), i =
1, 2 at time t-1, and α1,β1,1, β1,2, γ1,1, γ1,2 estimated from Eq. (1) as α
̂1 , ̂ β 1,2 ,̂γ 1,1 ,̂γ 1,2 , the corresponding qi,t(τ) of country i at time t is
β 1,1 , ̂
given by:
⃒ ⃒ ⃒ ⃒
α 1 + ̂β 1,1 ⃒Y1,t− 1 ⃒ + ̂β 1,2 ⃒Y2,t− 1 ⃒ + ̂γ 1,1 q1,t− 1 (τ) + ̂γ 1,2 q2,t− 1 (τ)
q 1,t (τ) = ̂
̂ (2)

We use a 36-month-rolling-window to account for potential time-series variation in EPU spillovers.4 Specifically, to estimate the
EPU spillovers for January 2006, we estimate Eq. (1) using the monthly data from 2003:01 to 2005:12, and use the coefficient esti­
mates to calculate ̂q i,t (τ). Next, we move the 3-year window one month forward and estimate the EPU spillovers for February 2006.
Finally, we repeat the above steps and generate a time series (Jan. 2006 – Sep. 2019) of EPU spillovers estimates for each country in the
sample.

2.2. Regression model and data

We run the following regression to gauge the effect of EPU spillovers on bond market volatility:
VOLi,t = b0 + b1 SPILLOVERi,t− 1 + b2 ΔEPUi,t− 1 + b3 GROWTHi,t− 1 + b4 βPRICESTABi,t− 1
(3)
+ b5 BUDBALi,t− 1 + b6 FXSTABi,t− 1 + b7 INTLIQi,t− 1 + b8 BUREAi,t− 1 + θi + εi,t

In the equation above, VOL is the bond market volatility calculated as the standard deviation of the daily bond returns for country i
in month t. The variable SPILLOVER is the EPU spillovers estimated at month t-1. To explore whether SPILLOVER estimated at different
quantiles exhibit different effects, we estimate SPILLOVER at the 80th, the 50th, and the 20th percentiles, respectively. We expect
stronger effect at higher quantiles, as turmoil in financial markets usually follows abnormally high uncertainty periods. The variable
ΔEPU is the first difference of domestic EPU. Following prior research (Jones et al., 1998; David and Veronesi, 2013; Duyvesteyn et al.,
2016), we also add economic, financial, and political risk factors as controls. Specifically, we use monthly changes in the estimated
GDP (GROWTH), monthly changes in consumer price index (PRICESTAB), and monthly budget as a percentage of GDP (BUDBAL) to
account for economic risk factors. We use monthly changes of the exchange rate against the US dollar (FXSTAB) and the ratio between
official reserves merchandise import (INTLIQ) to account for financial risk factors. We use BUREAU, the bureaucracy score from In­
ternational Country Risk Guide dataset to account for political risk factor. Lastly, we also include country fixed effect θi to account for

3
We use the first difference of EPU in our regression models since the original EPU is not stationary and therefore fails the requirement of the
multivariate quantile model (see White et al., 2015). In addition, the distribution of the change in EPU across 23 countries shows that EPU changes
have heavy tails and are not normally distributed. These findings necessitate the use of the multivariate quantile model.
4
We have also use 48-month and 60-month rolling windows and obtained similar results which are available upon request.

2
Y. Gong et al. Finance Research Letters 55 (2023) 103931

Fig. 1. MVMQ (1,1) model at the 50th percentile: The measure of global EPU spillovers to each country
Notes: This figure reports the global EPU spillovers for each country from 2006:01 to 2019:09.

any country-invariant effects.5

Our sample includes 23 countries for the period between 2003 and 2019.6 We collect developed market bond indexes from
Government Bond Index (GBI) of J.P. Morgan and emerging market bond indexes from the Emerging Market Bond Index (EMBI) of J.P.
Morgan. Both indexes track total US dollar-denominated returns. The developed markets include Australia, Canada, France, Germany,
Greece, Hong Kong, Ireland, Italy, Japan, Netherlands, Singapore, Spain, Sweden, United Kingdom and United States. The emerging
markets include Brazil, Chile, China, Colombia, India, Mexico, Russia and South Korea. We collect EPU index from the website of
Baker et al. (2016). Economic, financial, and political risks are collected from the International Country Risk Guide dataset, including
GROWTH, PRICESTAB, BUDBAL, FXSTAB, INTLIQ and BUREAU. Detailed definitions are provided in Appendix A.
Fig. 1 plots the time series of EPU spillovers for each country in our sample. It is clear that both domestic EPUs and EPU spillovers
fluctuate to a large extent, with peaks being observed during the global financial crisis (2007–2009) and the European debt crisis
(2010–2013).
Fig. 2 reports the 80th percentile impulse-response function of an aggregate EPU shock generated from N-1 countries on domestic
EPU. It is clear that one standard deviation of a shock in foreign EPUs (generated from N-1 countries) positively and significantly
affects the domestic EPU. The foreign-induced EPU shocks have the largest impact at the beginning and start to diminish steadily for
most countries after the fourth month.

3. Empirical results

3.1. Benchmark

Model (1) in Table 1 suggests that higher domestic EPU change leads to greater future bond market volatility. Including EUP
spillovers at the 80th percentile, Model (2) shows that high-quantile EPU spillovers have a positive and significant effect on future
bond market volatility. Model (3) and Model (4) suggest that EPU spillovers obtain a weaker effect when being estimated at lower
quantiles. We use the F test to explore whether the coefficient of EPU spillovers at the 80th percentile, the 50th percentile, and the 20th
percentile is equal. We find that the null hypothesis that all three coefficients are equal is rejected (p-value = 0.006). Therefore, EPU

5
In each county, global EPU spillovers, domestic EPU changes and other macro variables are long time-series data that are strongly correlated
over time. The effects of global EPU spillovers, domestic EPU changes and other macro variables would have largely been absorbed by the time
effect if added in panel data models. Therefore, the time-fixed effect is excluded from the regression.
6
Due to data availability of the control variables, our sample ends in 2019. However, we did run the benchmark regression without the controls
for the period including the post 2019 years, and the findings are unchanged. The table is available at request.

3
Y. Gong et al. Finance Research Letters 55 (2023) 103931

Fig. 2. MVMQ (1,1) model at the 80th percentile: Impulse response function
Notes: This figure plots the quantile impulse-response function where EPU spillovers are estimated at the 80th percentile. The 95% confidence
intervals are drawn in the dotted line.

Table 1
Benchmark.
Model (1) Model (2) Model (3) Model (4)
Variable EPU only 80th Percentile 50th Percentile 20th Percentile

SPILLOVER 0.044** 0.034* − 0.008

(2.120) (1.850) (− 0.560)
ΔEPU 0.022*** 0.018** 0.022*** 0.022***
(2.770) (2.470) (2.760) (2.650)
GROWTH -0.010 -0.010 -0.010 -0.010
(-1.250) (-1.210) (-1.270) (-1.260)
PRICESTAB -0.061*** -0.061*** -0.061*** -0.060***
(-3.080) (-3.310) (-3.100) (-2.990)
FXSTAB -0.030** -0.029** -0.029** -0.030**
(-2.160) (-2.070) (-2.150) (-2.170)
BUDBAL -0.020 -0.020 -0.020 -0.020
(-1.590) (-1.600) (-1.600) (-1.590)
INTLIQ -0.043*** -0.043*** -0.044*** -0.043***
(-2.750) (-2.670) (-2.760) (-2.740)
BUREAU -0.062** -0.073*** -0.069*** -0.060**
(-2.430) (-2.600) (-2.690) (-2.270)
Observations 3376 3376 3376 3376
Model Specification Test
Adj R-square 0.4844 0.4856 0.4848 0.4843
AIC -375.845 -382.817 -377.230 -374.162
LR Test 8.97*** 3.39* 0.32

Note: t-statistics clustered at country level are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

spillovers at different quantiles do exhibit different effects.7 The improvement of economic, financial, and political risk factors would
decrease bond market volatility.
Given that Model (1) is nested in Model (2) to Model (4), a likelihood-ratio (LR) test sheds more light on the necessity for the

7
We also use the F test for the coefficients of spillovers in the 80th and the 20th percentile to confirm that the increase of effect from the 20th to
80th percentile is indeed significant (F statistics = 14.40 with a p-value = 0.0011). Therefore, the evidence indeed suggests that the 80th percentile
does seem to have a stronger effect than the 20th percentile.

4
Y. Gong et al. Finance Research Letters 55 (2023) 103931

Table 2
Developed Markets vs. Emerging Markets.
Panel A: Developed Markets Panel B: Emerging Markets
Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) Model (7)

Variable EPU only 80th Percentile 50th Percentile 20th Percentile 80th Percentile 50th Percentile 20th Percentile
SPILLOVER 0.075** 0.072** 0.010 0.050** 0.025*** 0.007
(2.500) (2.370) (0.650) (1.990) (2.780) (0.480)
ΔEPU 0.022*** 0.020*** 0.022*** 0.023*** 0.020*** 0.023*** 0.023***
(2.770) (2.640) (2.710) (2.620) (2.700) (2.820) (2.750)
GROWTH -0.010 -0.010 -0.010 -0.010 -0.010 -0.010 -0.010
(-1.250) (-1.160) (-1.240) (-1.240) (-1.220) (-1.260) (-1.250)
PRICESTAB -0.061*** -0.060*** -0.059*** -0.061*** -0.063*** -0.061*** -0.061***
(-3.080) (-3.470) (-3.120) (-3.150) (-3.440) (-3.100) (-3.120)
FXSTAB -0.030** -0.028** -0.029** -0.030** -0.029** -0.030** -0.030**
(-2.160) (-2.060) (-2.140) (-2.150) (-2.100) (-2.170) (-2.160)
BUDBAL -0.020 -0.020 -0.020 -0.020 -0.020 -0.020 -0.020
(-1.590) (-1.620) (-1.590) (-1.580) (-1.600) (-1.610) (-1.570)
INTLIQ -0.043*** -0.043*** -0.044*** -0.044*** -0.042** -0.044*** -0.044***
(-2.750) (-2.680) (-2.800) (-2.760) (-2.560) (-2.770) (-2.790)
BUREAU -0.062** -0.088*** -0.082*** -0.069** -0.072*** -0.073*** -0.065**
(-2.430) (-3.520) (-3.130) (-2.360) (-2.590) (-2.700) (-2.440)
Observations 3376 3376 3376 3376 3376 3376 3376
Model Specification Test
Adj R-square 0.4844 0.4883 0.4864 0.4844 0.4864 0.4851 0.4843
AIC -375.845 -400.373 -387.717 -374.479 -387.710 -379.442 -374.209
LR Test 26.53*** 13.87*** 0.63 13.87*** 5.60** 0.36

Note: t-statistics clustered at country level are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Table 3
Global Financial Crisis (2007–2009).
Model (1) Model (2) Model (3) Model (4)
Variable EPU only 80th Percentile 50th Percentile 20th Percentile

SPILLOVER 0.033* 0.031* -0.006

(1.780) (1.750) (-0.470)
ΔEPU 0.024*** 0.023*** 0.025*** 0.024***
(2.870) (2.830) (3.000) (2.770)
SPILLOVER*CRISIS 0.111** 0.109** 0.045
(2.460) (2.040) (0.860)
ΔEPU* CRISIS 0.072*** 0.033** 0.050*** 0.062***
(3.630) (2.040) (2.710) (3.020)
GROWTH -0.010 -0.008 -0.009 -0.009
(-1.240) (-0.890) (-1.100) (-1.050)
PRICESTAB -0.061*** -0.055*** -0.058*** -0.058***
(-3.060) (-2.860) (-2.910) (-2.830)
FXSTAB -0.030** -0.027** -0.029** -0.029**
(-2.200) (-2.070) (-2.220) (-2.210)
BUDBAL -0.020 -0.026** -0.023* -0.022*
(-1.610) (-2.140) (-1.910) (-1.840)
INTLIQ -0.043*** -0.029* -0.039** -0.039**
(-2.760) (-1.810) (-2.470) (-2.260)
BUREAU -0.061** -0.084*** -0.072*** -0.064**
(-2.400) (-2.730) (-2.900) (-2.370)
Observations 3358 3358 3358 3358
Model Specification Test
Adj R-square 0.4858 0.4908 0.4876 0.4858
AIC -370.617 -401.787 -380.4196 -368.8429
LR Test 55.05*** 33.68*** 13.23***

Note: t-statistics clustered at country level are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

inclusion of EPU spillovers. The LR test shows that including EPU spillovers, especially measured at higher quantile, significantly
improves the original model.

3.2. Developed markets vs. emerging markets

Balli et al. (2017) document that EPU spillovers generated from developed markets are much larger than those generated from
emerging markets. Therefore, it is curious to explore whether EPU spillovers from developed market also cause greater impact on local
bond market volatility than those from emerging markets.

5
Y. Gong et al. Finance Research Letters 55 (2023) 103931

Table B1
. Volatility Persistence with Dynamic Panel Regression.
Model (1) Model (2) Model (3) Model (4)

Variable EPU only 80th Percentile 50th Percentile 20th Percentile

VOL 0.355*** 0.353*** 0.354*** 0.356***
(6.950) (6.880) (6.930) (6.950)
SPILLOVER 0.029** 0.028* -0.013
(2.010) (1.820) (-1.530)
ΔEPU 0.014* 0.012* 0.014* 0.014*
(1.920) (1.650) (1.920) (1.820)
GROWTH -0.004 -0.004 -0.004 -0.004
(-0.750) (-0.730) (-0.780) (-0.760)
PRICESTAB -0.042*** -0.042*** -0.042*** -0.041***
(-3.260) (-3.480) (-3.280) (-3.090)
FXSTAB -0.019** -0.019** -0.019** -0.019**
(-2.250) (-2.160) (-2.240) (-2.280)
BUDBAL -0.017** -0.017** -0.017** -0.017**
(-2.050) (-2.070) (-2.070) (-2.050)
INTLIQ -0.034*** -0.034*** -0.035*** -0.034***
(-2.710) (-2.630) (-2.720) (-2.680)
BUREAU -0.042** -0.050*** -0.047*** -0.038**
(-2.500) (-2.600) (-2.660) (-2.210)
Observations 3372 3372 3372 3372
Model Specification Test
Arellano–Bond tests for AR(1) 0.014** 0.014** 0.013** 0.014**
Arellano–Bond tests for AR(2) 0.798 0.783 0.771 0.803
Hansen test of over identifying 10.03 11.76 10.57 9.48
restrictions

Note: t-statistics clustered at country level are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Panel A (Panel B) of Table 2 repeats the benchmark regression with the EPU spillovers being estimated from developed (emerging)
markets only. We find that the effect of EPU spillovers from developed markets is about 50% greater (0.075 vs. 0.05) than that from the
emerging markets at the 80th percentile, and almost 100% greater (0.072 vs. 0.025) at the 50th percentile. Similar to the benchmark
results, EPU spillovers at the 20th percentile obtain no effect. By using the F statistics to test the equality for the coefficients of EPU
spillovers at 80th percentile, the 50th percentile and the 20th percentile, we find their coefficients are significantly different. Domestic
EPU change, together with all other controls, obtains similar effects as before. The LR test also shows the significance to include EPU
spillovers at higher percentiles.
We ascribe the stronger EPU spillover effect of developed markets to the fact that developed countries have larger economies, larger
shares of global trade, and greater amount of capital flows, which quickly transmit shocks to the global financial markets (see Huidrom
et al., 2020).

3.3. Global financial crisis

The global financial crisis provides an ideal test bed for comparing the effect of the domestic EPU change and global EPU spillovers.
Particularly, it is reasonable to expect that EPU spillovers to be much greater during the crisis period because for the majority of the
countries, the global financial crisis is not domestic. We create an interaction term for SPILLOVER and ΔEPU with a CRISIS dummy
(equals to one for years from 2007 to 2009 and 0 otherwise), respectively, to account for crisis periods. In terms of the control var­
iables, the significance of government budget increases compared with previous results since the global financial crisis tends to in­
crease government budgets (see Blinder and Zandi, 2015).
In Table 3, we find that the effect of EPU spillovers becomes particularly larger during the crisis period. For example, at the 80th
percentile, the effect of EPU spillovers during the crisis period is greater than during the normal period (0.111 vs. 0.033). A similar
pattern is observed at the 50th percentile and no effect at the 20th percentile. The LR test is significant, showing the necessity to
include EPU spillovers in the model.

3.4. Robustness

We conduct two robustness checks. First, to consider the persistence of bond market volatility and the potential endogeneity issue
of our model, we run the regressions using the dynamic panel models with GMM estimation. While confirming a positive and sig­
nificant effect of the lagged bond market volatility, our results show that EPU spillovers still maintain a positive and significant effect.
Domestic EPU change and other controls obtain similar effects as before. The detailed results are reported in Table B1 of Appendix B.
The second one is that we use tail risk as another bond market risk measure. We estimate the country-level bond market tail risk by
using the 5% Value at Risk (Bali and Cakici, 2004). The result also shows that global EPU spillovers and domestic EPU change
maintains their significant effect. In the model specification test, LR tests also indicate the importance of global EPU on the bond
market volatility. The detailed results are shown in Table B2 of Appendix B.

6
Y. Gong et al. Finance Research Letters 55 (2023) 103931

Table B2
Tail risk as the dependent variable.
Model (1) Model (2) Model (3) Model (4)

Variable EPU only 80th Percentile 50th Percentile 20th Percentile

SPILLOVER 0.100*** 0.072* 0.013
(2.910) (1.710) (0.490)
ΔEPU 0.026** 0.018* 0.025** 0.027**
(2.140) (1.790) (2.090) (2.160)
GROWTH -0.015 -0.014 -0.015 -0.015
(-1.170) (-1.110) (-1.190) (-1.160)
PRICESTAB -0.081*** -0.080*** -0.080*** -0.081***
(-2.750) (-3.030) (-2.770) (-2.800)
FXSTAB -0.041* -0.039* -0.040* -0.040*
(-1.950) (-1.830) (-1.940) (-1.940)
BUDBAL -0.021 -0.020 -0.021 -0.021
(-1.190) (-1.200) (-1.200) (-1.180)
INTLIQ -0.061** -0.060** -0.062** -0.062**
(-2.420) (-2.330) (-2.450) (-2.420)
BUREAU -0.150*** -0.175*** -0.164*** -0.153***
(-3.610) (-3.850) (-3.930) (-3.480)
Observations 3376 3376 3376 3376
Model Specification Test
Adj R-square 0.3892 0.3916 0.3899 0.3890
AIC -27,576.34 -27,588.99 -27,579.120 -27,574.650
LR Test 14.65*** 4.77* 0.31

Note: t-statistics clustered at country level are in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

4. Conclusion

This paper applies the multivariate quantile model to develop an EPU spillovers measure for each country and investigates the
effect of country-specific EPU spillovers on the local bond market volatility. The results show that EPU spillovers, especially at higher
quantiles, have a positive and significant effect on the bond market volatility. This effect becomes stronger when the spillovers are from
developed market and during the crisis period. After considering volatility persistence and potential endogeneity issue, this effect also
significantly exists.
The findings of our study carry implications for both policy makers and practitioners. Recognition of such a positive effect of EPU
spillovers on bond return volatility may help policy makers discern the external economic shocks resulting from contagion of global
EPU. The central bank should establish a monitoring mechanism for early warning of the international transmission of EPUs. Market
participants are advised to pay attention to the fact that EPU spillovers may serve as an important factor that affects the bond market.

Declaration of Competing Interest

The authors whose names are listed immediately below certify that they have NO affiliations with or involvement in any orga­
nization or entity with any financial interest (such as honoraria; educational grants; participation in speakers’ bureaus; membership,
employment, consultancies, stock ownership, or other equity interest; and expert testimony or patent-licensing arrangements), or non-
financial interest (such as personal or professional relationships, affiliations, knowledge or beliefs) in the subject matter or materials
discussed in this manuscript.

Data availability

Data will be made available on request.

Acknowledgements

We gratefully acknowledge the financial support from the National Natural Science Foundation Of China (Grant No. 71971133),
the National Social Science Foundation of China (Grant No. 21BGL270), the Shanghai Sailing Program (Grant No. 20YF1413100) and
the Shanghai Science and Technology Committee (Grant No. 22692105300).

Appendix A. Definition of Variables and Data Sources

Variable Definition and Measures Data Source

(continued on next page)

7
Y. Gong et al. Finance Research Letters 55 (2023) 103931

(continued )
Variable Definition and Measures Data Source

VOL Volatility of bond returns is the standard deviation of the daily bond returns for country i in month t. The daily bond Datastream
returns is log daily returns of Government Bond Index (GBI) and Emerging Market Bond Indices (EMBI).
EPU Each national EPU index reflects the relative frequency of own-country newspaper articles that contain a trio of terms Baker et al. (2016)
pertaining to the economy (E), policy (P) and uncertainty (U). In other words, each monthly national EPU index
value is proportional to the share of own-country newspaper articles that discuss economic policy uncertainty in that
month.
The specific measure is to obtain a monthly count of articles that contain a trio of terms about the economy (E), policy
(P), and uncertainty (U). The second step is to scale the raw counts, standardize each newspaper’s variation, average
across papers in a country by month, and normalize it. The description of E, P and U are shown below.
GROWTH The annual change in the estimated GDP, at constant 1990 prices, of a given country is expressed as a percentage International Country Risk
increase or decrease. The high percentage annual change of estimated GDP has high points, which means high Guide
economic growth.
PRICESTAB The estimated monthly inflation rate (the unweighted average of the Consumer Price Index) is calculated as a International Country Risk
percentage change. The low percentage change of monthly inflation rate has high points, which means low inflation Guide
risk.
BUDBAL The estimated central government budget balance (including grants) for a given year in the national currency is International Country Risk
expressed as a percentage of the estimated GDP for that year in the national currency. The low budget as a percentage Guide
of GDP has high points, which means low government budget risk.
FXSTAB The appreciation or depreciation of a currency against the US dollar (against the German mark in the case of the USA) International Country Risk
over a month is calculated as a percentage change. The low appreciation or depreciation of a currency against the US Guide
dollar has high points, which means low foreign exchange risk.
INTLIQ The total estimated official reserves for a given year, converted into US dollars at the average exchange rate for that International Country Risk
year, including official holdings of gold, converted into US dollars at the free market price for the period, but Guide
excluding the use of IMF credits and the foreign liabilities of the monetary authorities, is divided by the average
monthly merchandise import cost, converted into US dollars at the average exchange rate for the period. This
provides a comparative liquidity risk ratio that indicates how many months of imports can be financed with reserves.
The high percentage of official reserves over merchandise import cost has high points, which means low
international liquidity risk.
BUREAU The institutional strength and quality of the bureaucracy is another shock absorber that tends to minimize revisions International Country Risk
of policy when governments change. The good bureaucracy quality has the high score. Guide

Appendix B

Table B1, Table B2

References

Adinolfi, J., 2022. U.K. could trigger a global crisis as pound collapses while bond yields soar, Larry Summers says, MarketWatch. https://www.marketwatch.com/
story/u-k-could-trigger-a-global-crisis-as-pound-collapses-while-bond-yields-soar-larry-summers-says-11664292688.
Al-Thaqeb, S., Algharabali, B., 2019. Economic policy uncertainty: a literature. J. Econ. Asymmetries 20(C).
Baker, S., Bloom, N., Davis, S., 2016. Measuring economic policy uncertainty. Q. J. Econ. 131 (4), 1593–1636.
Bali, T.G., Cakici, N., 2004. Value at risk and expected stock returns. Financ. Anal. J. 60 (2), 57–73.
Balli, F., Uddin, G.S., Mudassar, H., Yoon, S., 2017. Cross country determinants of economic policy uncertainty spillovers. Econ. Lett. 156 (C), 179–183.
Blinder, A.S., Zandi, M., 2015. The financial crisis: lessons for the next one, https://www.cbpp.org/research/economy/the-financial-crisis-lessons-for-the-next-one.
Chen, L., Du, Z., Hu, Z., 2020. Impact of economic policy uncertainty on exchange rate volatility of China. Finance Res. Lett. 32(C).
David, A., Veronesi, P., 2013. What ties return volatilities to price valuations and fundamentals? J. Polit. Econ. 121 (4), 682–746.
Duyvesteyn, J., Martens, M., Verwijmeren, P., 2016. Political risk and expected government bond returns. J. Empir. Finance 38 (A), 498–512.
Fang, L., Chen, B., Yu, H., Qian, Y., 2018. The importance of global economic policy uncertainty in predicting gold futures market volatility: a GARCH-MIDAS
approach. J. Futures Markets 38 (3), 413–422.
He, F., Wang, Z., Yin, L., 2020. Asymmetric volatility spillovers between international economic policy uncertainty and the US stock market. N. Am. J. Econ. Finance
51 (C).
Huidrom, R., Kose, M.A., Ohnsorge, F.L., 2020. How important are spillovers from major emerging markets? Int. Finance 23 (1), 47–63.
Jones, C., Lamont, O., Lumsdaine, R., 1998. Macroeconomic news and bond market volatility. J. Financ. Econ. 47 (3), 315–337.
Ma, Y., Wang, Z., He, F., 2020. How does economic policy uncertainty affect stock market volatility? Evidence from G7 countries. Int. J. Health Econ. Manage. 27 (2),
2303–2325.
White, H., Kim, T., Manganelli, S., 2015. VAR for VaR: measuring tail dependence using multivariate regression quantiles. J. Econom. 187 (1), 169–188.
Yen, K., Cheng, H., 2021. Economic policy uncertainty and cryptocurrency volatility. Finance Res. Lett. 38(C).