The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/2044-1398.htm

CFRI
11,4 Geopolitical risk, economic policy
uncertainty and asset returns in
Chinese financial markets
474 Thomas C. Chiang
Finance, Drexel University Bennett S LeBow College of Business,
Received 19 August 2020
Revised 1 December 2020
Philadelphia, Pennsylvania, USA
17 February 2021
Accepted 18 February 2021 Abstract
Purpose – This paper investigates the impact of a change in economic policy uncertainty ðΔEPUt Þ and the
absolute value of a change in geopolitical risk ðjΔGPRt jÞ on the returns of stocks, bonds and gold in the Chinese
market.
Design/methodology/approach – The paper uses Engle’s (2009) dynamic conditional correlation (DCC)
model and Chiang’s (1988) rolling correlation model to generate correlations of asset returns over time and
analyzes their responses to ðΔEPUt Þ and jΔGPRt j:
Findings – Evidence shows that stock-bond return correlations are negatively correlated to ΔEPUt , whereas
stock-gold return correlations are positively related to the jΔGPRt j; but negatively correlated with ΔEPUt :
This study finds evidence that stock returns are adversely related to the risk/uncertainty measured by
downside risk, ΔEPUt and jΔGPRt j, whereas the bond return is positively related to a rise in ΔEPUt ; the gold
return is positively correlated with a heightened jΔGPRt j.
Research limitations/implications – The findings are based entirely on the data for China’s asset markets;
further research may expand this analysis to other emerging markets, depending on the availability of GPR
indices.
Practical implications – Evidence suggests that the performance of the Chinese market differs from
advanced markets. This study shows that gold is a safe haven and can be viewed as an asset to hedge against
policy uncertainty and geopolitical risk in Chinese financial markets.
Social implications – This study identify the special role for the gold prices in response to the economic
policy uncertainty and the geopolitical risk. Evidence shows that stock and bond return correlation is
negatively related to the ΔEPU and support the flight-to-quality hypothesis. However, the stock-gold return
correlation is positively related to jΔGPRj, resulting from the income or wealth effect.
Originality/value – The presence of a dynamic correlations between stock-bond and stock-gold relations in
response to ΔEPUt and jΔGPRt j has not previously been tested in the literature. Moreover, this study finds
evidence that bond-gold correlations are negatively correlated to both ΔEPUt and jΔGPRt j:
Keywords Stock–bond return correlation, Stock-gold return correlation, Downside risk,
Economic policy uncertainty, Geopolitical risk, Safe haven
Paper type Research paper

Abstract
Highlights

(1) This paper investigates the impact of changes in economic policy uncertainty
ðΔEPUt Þ and geopolitical risk ðjΔGPRt jÞ on asset returns in the Chinese market.
(2) The dynamic correlation of stock and bond returns is negatively correlated to the
ΔEPUt , and bonds are a good instrument to avoid ΔEPUt.
(3) Dynamic correlations of stock-gold returns are negatively correlated to ΔEPUt but
positively correlated with the jΔGPRt j.
China Finance Review
International
Vol. 11 No. 4, 2021
pp. 474-501 JEL Classification — G10, G11, G14, G15
© Emerald Publishing Limited The author would like to thank for the support of the Austin Chair fund from LeBow College of
2044-1398
DOI 10.1108/CFRI-08-2020-0115 Business, Drexel University.
 (4) Time-varying bond-gold return relations are negatively correlated with a rise in The impact in
ΔEPUt and jΔGPRt j. economic
(5) Both ΔEPUt and jΔGPRt j create an adverse effect on stock returns. policy
(6) Heightened ΔEPUt correlates with a greater bond return, whereas a rise in jΔGPRt j uncertainty
gives rise to a higher return in gold. Gold is an effective instrument to hedge against
jΔGPRt j.
475
1. Introduction
Dynamic asset pricing models that highlight a risk-return tradeoff constitute the core of
modern finance theory. Rational behavior implies that an economic agent constantly reacts to
news due to external shocks or government policy innovations. This on-going market
adjustment phenomenon is cast in a dynamic approach, often giving rise to time-varying risk-
return relations. In a system which operates under the stock-bond apparatus, investors
attempt to make an efficient portfolio allocation that equalizes assets returns after adjusting
for the features of different asset instruments. Should an external shock disrupt an
equilibrium condition, asset holders would readjust portfolio combinations by moving funds
from a lower return asset to an asset with a higher expected return in response to the return
differentials. This “gross substituting” property (Tobin, 1969) would undoubtedly alter the
existing correlation of stock-bond returns. However, if the shock is triggered by unexpected
asset growth, the resulting wealth effect could drive higher demand for both stocks and bonds,
this “wealth effect” leads to a positive correlation between the returns of these two asset
categories. These market adjustments and an interaction between stocks and bonds have not
incorporated the factors driven by the amount of risk/uncertainty appetite in the market.
The notion of positive correlation between stock-bond returns can be traced to the “Fed
model” (Yardeni, 1997), which states that the P/E ratio derived from a stock investment
should be approximately equal to the reciprocal of the bond’s yield to maturity, which leads to
a positive correlation between the stock market’s E/P ratio and bond yields. Kwan (1996)
observes that both required rates of stock returns and bond yields are considered as part of a
discount factor for calculating the future income flows in a valuation model; it is, therefore,
natural to form a positive correlation between stock and bond returns. Yet, an extreme market
condition may drive investors to move their funds to a safe heaven, generating a fly-to-quality
phenomenon, where increase stock market volatility induces investors to sell off stocks in
favor of bonds. Thus, the stock-bond returns would display a negative correlation. In
examining the US market, Connolly et al. (2005) and Guidolin and Timmermann (2006) find
evidence of negative spikes in stock-bond correlations at the sample period at the end of 1997
and the end of 1998. A similar finding is documented by Chiang et al. (2015) in the G7 markets.
Recent studies by McMillan (2018), Pericoli (2018), Campbell et al. (2020) conclude that the
relative equity and bond yield values and their relation are mainly driven by inflation rate,
which causes negative correlations. However, Baele et al. (2010) report that macroeconomic
fundamentals contribute little to explaining stock and bond return correlations, whereas the
“variance premium” is critical in explaining stock return volatility.
Implicit in the above-mentioned literature is the notion that the portfolio composition is
restricted to the two assets: stocks and bonds (Asness, 2003; Connolly et al., 2005; Baele et al.,
2010; Chiang et al., 2015; McMillan, 2018). It also assumes that bonds are the only alternative
asset to stocks as a way to avoid risk. This approach typically appears in empirical analyses
of advanced markets. This premise may not hold true in Chinese society, since bond
investment is not well developed and the instruments in bond trading are not generally
understood by investors or households in the Chinese market; hence, bonds are not a popular
asset used to hedge against uncertainty. Much more in line with traditional culture is the
propensity of Chinese households to purchase gold in forms of coins, bars or pure gold
CFRI jewelry, which are favorably held as a way of storing value, facilitating future consumption,
11,4 fostering a peaceful mind and providing a hedge against uncertainty while offering
anonymity. Further, in times of national crisis, gold can provide both financial flexibility and
investment options outside the home country. Among other reasons, China’s central bank can
also use gold to maintain its exchange rate stability and to guard against national crisis
during eras of uncertainty without worry about default risk [1]. Evidence, which shows that
since 1900, all national currencies have been depreciated relative to gold (Clark, 2020),
476 supports the claim that physical gold has been the best long-term vehicle for storing value.
Although the international financial arrangement has departed from the Bretton Woods
system, gold is still considered as a tangible asset to be held by the central bank for achieving
portfolio diversification or providing a risk hedge. For these reasons, it is worth investigating
the dynamic correlation between stock and gold returns in addition to the stock-bond return
relation.
In analyzing the stock-bond correlation due to a shock from economic uncertainty, it has
been the custom to use measures of market volatility, such as implied volatility of stock
market (Whaley, 2009; Hakkio and Keeton, 2009; Connolly et al., 2005; Baele et al., 2010),
implied volatility of bond market (Chiang et al., 2015) or inflation volatility (McMillan, 2018;
Pericoli, 2018) to explain the stock-bond variations. No consideration is given to the impacts
from ΔEPUt and jΔGPRt j. Thus, the traditional definition of uncertainty/risk based on
market volatility may not adequately encompass all information of uncertainty that
sufficiently conveys investors’ fears and sentiments in decision making.
The recent literature finds that a change in economic policy uncertainty ðΔEPUt Þ can
cause further policy uncertainty in the future and hence increase stock market volatility
(Caggiano et al., 2014; Bloom, 2014). Moreover, it has been seen that news of a change in
geopolitical risk ðΔGPRt Þ arising from an increase in uncertainty due to military tensions,
political regime changes, terrorist threats, corruptions, etc. (Chan et al., 2019) can interrupt
economic activity and households’ incomes. In fact, ΔGPR uncertainty could lead investors to
take a more prudent approach on investments and postpone their decision-making because of
an emotional dimension shaped by fear, pessimism and other negative reactions that cause
investors to respond intensively to political events (Hillen et al., 2017). To cope with political
uncertainty, investors are more likely to fly-to-gold instead of bonds, since hedging against
uncertainty with gold is perceived as an acceptable means of storing value and could
generally be converted to cash in global markets.
This paper contributes to the literature in the following aspects. First, in addition to the
stock market risk, this study presents evidence by employing ΔEPUt and jΔGPRt j to explain
asset returns and their correlation variations. This treatment of model specification helps to
reduce the bias of only using stock market risk as an argument due to its correlation with
ΔEPUt and jΔGPRt j [2]. Further, the selection of these two uncertainty variables differs from
the traditional approach, which emphasizes an economic measure of stock return volatility.
Both indices of EPUt (Davis et al., 2019) and GPRt (Caldara and Iacoviello (2019) are news-
based variables, which are obtained from the reading of local newspapers. As a result, the
information is current and efficient and fairly delivered to the public without suffering from
time delay and model biases.
Second, evidence shows a time-varying stock-bond correlation in the Chinese market,
which is negatively correlated to the ΔEPUt. The inverse relation also displays in the stock-
gold return relation. The evidence of a negative response to a rise in the ΔEPUt clearly
presents a substitution effect for stock-bond and stock-gold relations and is also consistent
with the decoupling phenomenon that occurs as uncertainty rises.
Third, with respect to the reaction to ΔGPRt uncertainty, there is no clear direction for the
stock-bond return correlation; however, evidence suggests that stock-gold return correlations
are positively correlated to the jΔGPRt j, displaying a dominance of income effect.
 Fourth, testing of the uncertainty effects on individual asset returns produces evidence, The impact in
which is consistent with the finding of Jones and Sackley (2016) in US and Europe markets economic
and suggests that stock and gold returns are negatively correlated with ΔEPUt; however, the
bond return is positively correlated with ΔEPUt, indicating that bonds are a good instrument
policy
for hedging against a rise in ΔEPUt. On the other hand, evidence concludes that both stock uncertainty
and bond returns are negatively correlated to a rise in jΔGPRt j, yet, the gold return displays
a positive sign, implying that gold is a good instrument to hedge against the jΔGPRt j. Clearly,
Chinese investors are able to identify the difference and make a clear choose among different 477
instruments to hedge against uncertainty whether it arises from economic policy or from
geopolitical policy.
Fifth, this study also analyzes the time-varying correlations between bond and gold
returns. Evidence indicates that the bond-gold return relation is negatively correlated to a
heightened ΔEPUt and jΔGPRt j, respectively. This study thus provides a thorough analysis
of the triangular relations of stock-bond, stock-gold and bond-gold return variations.
The remainder of this paper is organized as follows. Section 2 provides a brief literature
review that leads to the specification of the asset-return correlation models. Section 3 presents
an econometric model pertinent to empirical analysis. Section 4 describes the data and
variable construction for the study’s empirical analysis. Section 5 presents empirical evidence
for dynamic correlations in relation to ΔEPUt and jΔGPRt j. Section 6 reports estimated
results of ΔEPUt and jΔGPRt j on returns in different asset markets. Section 7 concludes the
empirical findings.

2. Literature review
Correlations of asset returns are rather complex. In addition to the common time trend and
individual time series correlations, return variables commonly react to economic news,
changes in inflation rate, the business cycle and policy innovations on the one hand. On the
other hand, there are common cross-correlations with each other or with some unobservable
variables (Stock and Watson, 1988), which create spurious correlations. Thus, the effect of
stock returns and bond yields on the discount factor may not be a unique channel that
explains correlations of stock-bond return relations. Although the relative values of stocks
and bonds are normally driven by the return disparities, an upward shift in the wealth effect
and an enthusiastic high-tech boom that induced investors to hold both stocks and bonds
simultaneously in the late 1990s in the US market may also be responsible for this behavior
(d’Addona and Kind, 2006). In fact, economic fundamentals could be the key to boost asset
returns. Andersson et al. (2008) report that inflation and economic growth expectations are
significant in interpreting the time-varying correlation between stock and bond returns in US
and German markets. McMillan (2018) and Campbell et al. (2020) explain that higher inflation
lowers real bond returns and higher output raises stock returns, causing bond and stock
returns to present a negative correlation. Similarly, Yang et al. (2009) investigate markets in
the US and the UK and find that the correlation of stock–bond returns is related to the
business cycle. However, Baele et al. (2010) cannot find that variables such as interest rates,
inflation, the output gap and cash flow growth make a strong contribution in explaining
stock-bond return correlations in the U.S. data.
During a period in which a stock market experiences a severe downturn, it is generally
observed that stock-bond returns reveal a negative correlation. This is essentially due to the
risk adverse behavior of investors who tend to sell off stocks and replace them with bonds as
volatility increases. This shift triggers a “flight-to-quality” phenomenon (Baur and Lucey, 2009;
Lucey, 2009; Chiang et al., 2015). But as economic prospects turn more optimistic, risk adverse
investors opt to re-enter the stock market, leading to a “flight-from-quality” phenomenon. Thus,
the result is a negative correlation for stock-bond returns. Evidence documented by Gulko
(2002), Connolly et al. (2005), Baur and Lucey (2009), Lucey (2009) and Chiang et al. (2015)
CFRI confirms this hypothesis and finds that stock–bond returns are negatively correlated with
11,4 stock market uncertainty as measured by the implied volatility of the S&P 500 index.
Despite the contributions of the above empirical analyses, the premise in previous
empirical analyses is restricted to a two-asset model, which is applicable to advanced markets
and may not be relevant to emerging market where the bond market is relatively thin. Thus,
several studies hypothesize gold, instead of bonds, should be considered as a safe haven for
investors. To test whether gold acts as a safe haven, which is defined as a security that is
478 uncorrelated with stocks and bonds in a market crash, Baur and Lucey (2010) examine the US,
UK and German markets and find that gold is a hedge against stocks on average and a safe
haven in extreme stock market conditions. A portfolio analysis further shows that the safe
haven property is short-lived. O’Connor et al. (2015) apply a Markov-Switching model to assess
whether two distinct states exist between gold’s return with the stock return. The evidence
provided by their study based on the UK and US markets indicates that gold is consistently a
hedge, but no distinct safe haven state exists between gold and stock markets. Tripathy (2016)
investigates the India market and finds that the gold price and stock market price are co-
integrated, indicating a long-run equilibrium relationship between them. However, no causal
relationship exists between the gold price and stock market price in the short run.
In spite of the evidence that gold acts as an instrument for hedging risk, analyses are
limited to time periods when stock market conditions are extreme, or the inflation rate is high.
In practice, investors are continually monitoring their portfolio performance and maintain a
consistent watch on the parametric effect of uncertainty. In fact, the scope of the shocks,
which occur and are revealed in daily newspapers, indicates that the origin of uncertainty is
by no means limited in the times of significant financial shocks. For this reason, this study
examines the impact of news-based changes in economic policy uncertainty ðΔEPUÞ and
geopolitical risk ðΔGPRÞ on asset returns and their dynamic correlations. In facing a rise of
ΔGPR uncertainty, gold is an instrument universally acceptable as a means of storing value,
whereas a perceived limitation of bonds is that their value is often subject to validation of a
political regime and they are often constrained by their marketability in global markets.
However, holding gold versus securities entails opportunity costs that become readily
apparent when the economy recovers. In this climate, the costs continue to grow as the
economy booms; thus, the property of gold as a safe haven is expected to be short-lived. This
study provides an empirical study on the dynamic correlations between stock-bond and
stock-gold relations in Chinese markets [3]. Further, this study extends the analysis to the
time-varying bond-gold return variations. The evidence from this paper, therefore,
undoubtedly provides more insight on investors’ portfolio behavior that goes beyond the
traditional approach taken in analyzing asset correlations, which typically is limited to
financial volatility and economic shocks.

3. A dynamic conditional correlation model

To estimate the dynamic conditional correlation (DCC) path between asset returns, it is
convenient to follow Engle’s (2009) model, [4] which has been widely used to investigate the
time-varying coefficient property in the literature (Chiang et al., 2007; Antonakakis et al., 2013;
Allard et al., 2020, among others). This model is appealing because of its ability to capture
multivariate return correlations and has a capacity for dealing with the volatility clustering
phenomenon that alleviates the heteroscedasticity problem.
Let us consider a bivariate return series, fRt g, which can be expressed as:
Rt ¼ δt þ ut (1)
0
where Rt ¼ ½Ri; t Rj; t  , which can be simplified as a 2 3 1 vector for asset returns i and j; δt is
the mean values of assets Ri;t and Rj;t, which has the conditional expectation of multivariate


time series properties [5]; ut
Ft−1 ¼ ½u1;t u2;t  ∼ N ð0; Ht Þ; Ft−1 is the information set up for
0

The impact in
time t–1. In the multivariate DCC-GARCH apparatus, Engle (2009) sets up the conditional economic
variance-covariance matrix Ht as:
policy
Ht ¼ Dt Pt Dt (2) uncertainty
where Dt ¼ diagfHt−1=2 g is the diagonal matrix of time-varying standard deviations from the
model; Pt is the time-varying conditional correlation matrix, which is obtained from the 479
following transformation:
Pt ¼ diagfQt g−1=2 Qt diagfQt g−1=2 (3)

The expression of Pt matrix in equation (3) is satisfied by ones on the diagonal and off-
diagonal elements that have an absolute value less than one. Since finance theory suggests
that the impact of a negative shock may have a more profound effect on the volatility than
that of an equivalent positive shock, there is an extra impact on the dynamic correlations.
This notion leads to an asymmetric DCC model. In expressing it, we write [6]:
0 0
Qt ¼ Ω þ a εt−1 εt−1 þ γ ηt−1 ηt−1 þ βQt−1 ; (4)

where the Qt is a positive definite and ηt ¼ min½εt ; 0.

The product of ηi; t ηj; t will be non-zero only if these two variables are negative. The
coefficient of γ captures the notion that the correlation increases more in response to negative
shocks than positive shocks. The estimation of Ω is given by:
 
b ¼ ð1  α  βÞQ  γN ;
Ω (5)

where Qt ¼ ðQij;t Þ is assumed to be the 2 3 2 time-varying covariance matrix of εt ,


0 pffiffiffiffiffiffi
Q ¼ E½εt εt  is the 2 3 2 unconditional variance matrix of εt (where εi;t ¼ ui;t = hi;t),
 0
N ¼ E½ηt ηt  is the 2 3 2 unconditional variance matrix of ηt ; a; β and γ are non-negative
scalar parameters satisfying ð1 − α − β − γÞ>0: The substitution of equation (5) for (4) yields:
Q
ρij;t ¼ pffiffiffiffiffiffiffiij;t
pffiffiffiffiffiffiffi : (6)
Qi;t Qj;t

which can be used to calculate the correlations for the two assets. As proposed by Engle
(2009), the ADCC model can be estimated by using a two-stage approach to maximize the log-
likelihood function.

4. Data
This study employs monthly stock data for the sample period of January 2000 through May
2020, including the total stock market index (TTMK), Shanghai A-share index (SHA),
Shanghai B-share index (SHB), Shenzhen A-share index (SZA) and Shenzhen B-share index
(SZB). The stock indices are downloaded from Datastream International.
The indices of EPU and GPR data were downloaded from economic policy uncertainty
website, which is available at: https://www.policyuncertainty.com/all_country_data.html.
The original data of EPU were constructed by Baker et al. (2013) and updated from time to
time. The EPU Index for China (Davis et al., 2019) is obtained from monthly counts of articles
that contain at least one term in each of three words sets: Economics, Policy and Uncertainty.
The newspaper-based uncertainty data for China were obtained from two Chinese
newspapers: the Renmin Daily and the Guangming Daily. The first step was to search for
the words uncertain/uncertainty/not sure/hard to tell/unknowns/economy/business, etc. in
CFRI each set based on Chinese characters and correspondingly, translations in English. The
11,4 second step focuses on scaling the raw monthly EPU counts by the number of total articles for
the same newspaper and month. Davis et al. (2019) show that the Chinese EPU index was
elevated with the Global Financial Crisis of 2008–2009, and especially, rising trade policy
tensions in 2017–2018.
Caldara and Iacoviello (2017) construct a monthly index of geopolitical risk (GPR) Index
by counting the occurrence of words related to geopolitical tensions in 11 leading
480 international newspapers. The search identifies articles containing references to the
following words involving actual adverse geopolitical events: {military-related tensions,
nuclear tensions, war threats, terrorist threats, terrorist acts or the beginning of a war}.
Caldara and Iacoviello’s (2017) data set includes monthly country specific GPR indexes
constructed for 18 emerging economies. This study utilizes the Chinese GPR index, which
was downloaded from Caldara and Iacoviello (2019) and is available at https://www2.bc.edu/
matteo-iacoviello/gpr.htm.
The daily stock index data are used to calculate the downside risk, which is the minimum
value of daily returns of the past 22 trading days (Bali et al., 2009); the original VaRt s are
multiplied by 1 before running regressions. The bond return is calculated based on the
30-year government bond. The stock returns are calculated by taking the first natural
log-difference of price indices times 100.

5. Empirical evidence
5.1 Time series paths of asset indices
Figure 1 depicts the time series plots of various stock indices that measure the total market
(TTMK), SHA, SHB, SZA, SZB, bond and gold price. The stock indices display a high degree
of comovements as exhibited by some major turning points, which reflect they are subject to
some common events or external shocks. Two special points are noteworthy. First, the
TTMK index lies above the positions of SHA, SHB, SZA and SZB at different points of time,
revealing the impact of dividend yields distributed to investors. The difference is also
reflected in a higher value for the mean as reported in Table 1. Second, the time paths of
TTMK show historical peaks at three different times; the first one occurred in October 2007
during a long bull market, which then turned to a bear market in October 2008 as a global
financial crisis gripped markets across the world. The second time Chinese stock prices
reached a new historical high was in May 2015 when the markets were driven by an easing in
monetary policy, which eased margin financing and prompted an enthusiastic expansion by
individual investors in speculative investment [7]. The bubble then burst and markets
corrected. Starting early 2016, the market rebounded and reached a third peak in January
2018. Since then, the markets have reversed due to the impact of a Chinese-U.S. trade war [8].

2,000

1,600

1,200

800

Figure 1. 400
Time series plots of
stock and gold price 0
indices in Chinese 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20
markets TTMK(RI) SHA SHB
SZA SZB GOLD
Notice that during the global financial crisis and its post period, gold prices rose steadily. The impact in
Additionally, the change in attitude of risk adverse investors, which was partly induced by economic
the stimulus plan (Barboza, 2008), has caused the gold index to outperform other indices since
October 2008. This performance reflects to some extent investors’ use of gold as a hedge
policy
against uncertainty after the global financial crisis and recent COVID-19’s spreads. uncertainty

5.2 Summary of statistics of asset returns 481

Presented in Table 1 is the monthly returns for different assets along with aggregate markets.
The mean value in Table 1 indicates that the TTMK stocks have the highest performance,
while the SHA underperforms all other markets and is just slightly higher than the bond
market return. The performance of gold return, which lies behind the TTMK and SZB
markets, is reasonably good. Yet, the gold return variability is much lower than the stock
returns as shown in the SD. With respect to the bond return, the mean value presents the
lowest value and is accompanied by the smallest SD.

5.3 Dynamic conditional correlations of asset returns

Figure 2 depicts time series plots of dynamic conditional correlations for the monthly data
between 30-year Chinese bond returns and stock returns for TTMK, SHA, SHB, SZA and SZB
indices, respectively, based on equation (6) as given by ρt ðRi ; Rj Þ ¼ pffiffiffiffiffii;j;tpffiffiffiffiffi. The plots
h
hi;t hj;t
clearly indicate the correlations are time-varying and mostly lie in the positive regime and
exhibit a stationary pattern, except for the SHB stock-bond correlation, which has spikes in
periods of the late 2008 – early 2009, which may be influenced by the US stock market
disturbances during the period of the financial crisis . However, the correlations between gold
return and stock returns are somewhat different. As shown in Figure 3, the time-varying
patterns for the pairwise correlations between stock and gold returns swing across different
regimes. Nevertheless, the correlations of stock-gold relations for the SHB-share markets are
more stable. The different time paths between Figures 2 and 3 demonstrate the different
preferences that investors have toward gold and bonds, even though both of these two assets
can alternatively be viewed as instruments for hedging against stock risk.
To provide a more concise analysis, we present summary statistics of pairwise correlated
asset returns. Table 2 reports the correlations of stock-bond returns. The results indicate a
positive relation for all correlation coefficients although negative figures are shown in the
minimum values. The positive correlations may be attributable to the income effect that
simultaneously drives stocks and bonds. This result is not consistent with the evidence in
advanced market (Connolly et al., 2005; Chiang, 2020). For a comparison, the estimation is also
provided in the last column, which shows a negative relation in the US market.

RTTMK; t RSHA; t RSHB; t RSZA; t RSZB; t Rb; t RGold; t

Mean 1.1254 0.3527 0.7436 0.6880 0.9699 0.3315 0.7800

Median 1.0564 0.6425 0.3200 1.0066 1.2993 0.3492 0.6332
Maximum 24.1840 24.3787 53.9978 25.6839 87.5985 0.5408 13.2127
Minimum 30.8850 28.2353 40.5917 31.3191 33.3615 0.1000 14.5966
Std. Dev 8.1147 7.5553 10.2115 8.6279 9.9221 0.0962 4.6330 Table 1.
Skewness 0.3956 0.5159 0.4504 0.3716 2.3353 0.1067 0.0571 Summary statistics of
Kurtosis 4.1522 4.9438 7.3307 4.1311 26.0189 2.0497 3.3291 monthly stock market
Jarque-Bera 20.1058 49.8427 20.3700 18.8497 56.7700 9.7622 1.2485 returns: 2000.01–
Observations 245 245 245 245 245 245 245 2020.05
 0.8
CFRI
11,4 0.4

0.0

–0.4

482 –0.8

Figure 2.
–1.2
Time-varying 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20
correlations of stock
and bond returns RHO(Rm,Rb_30Y) RHO(R_SHA,Rb_30Y) RHO(R_SHB,Rb_30y)
RHO(R_SZA,Rb_30Y) RHO(R_SZB,Rb_30Y)

1.0

0.8

0.6

0.4

0.2

0.0

–0.2

–0.4

Figure 3. –0.6
Time-varying –0.8
00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20
correlations of stock
and gold returns RHO(Rm,R_GOLD) RHO(R_SHA,R_GOLD) RHO(R_SHB,R_GOLD)
RHO(R_SZA,R_GOLD) RHO(R_SZA,R_GOLD)

The correlations of stock-gold returns are reported in Table 3. The statistics show that the
correlation coefficients between stocks and gold in the Chinese market are time-varying and
positive as well. The Bai–Perron test indicates the breakup dates, and the Chow test
demonstrates the rejection of absence of structural changes. These results of a positive
correlation between stock and gold returns appear contrary to general observations in the US
market whose coefficient is provided in the last column. Why does Chinese market
performance follow a path contrary to conventional anticipation and instead displays a
positive sign for the stock-gold relation? Obviously, the evidence suggests that gold does not
seem to serve as an asset that can be substituted for stocks. It appears that both stock and
gold work to complement each other due to either an income or wealth effect that arises from
the high saving rate in China. The other possibility may stem from the country’s thin bond
markets.

5.4 Determinants of time-varying asset return correlations

Although portfolio theory claims that risk adverse investors tend to allocate their assets
following the mean-variance approach (Tobin, 1969) by moving funds from assets with low
returns to instruments with higher returns during periods of relatively low volatility and
reversing the course during periods of relatively high volatility (Gulko, 2002), this approach
can be viewed as restrained from financial perspective. In practice, investors often promptly
react to policy uncertainty as well as the fear for geopolitical risk. As noted in the recent
literature, heightened EPU can impede economic activity, causing a potential deterioration in
future cash flows that would depress stock prices (Bloom, 2009, 2014; Davis et al., 2019;
 US
ρt ðRTTMK ; Rb Þ ρt ðRSHA; t ; Rb Þ ρt ðRSHB; t ; Rb Þ ρt ðRSZA ; Rb Þ ρt ðRSZB ; Rb Þ ρt ðRTTMK ; RbUS Þ

Mean 0.1213 0.0832 0.0882 0.0378 0.1329 0.2381

Median 0.1109 0.0666 0.0589 0.0427 0.0943 0.1824
Maximum 0.4420 0.6357 0.9984 0.2958 0.4201 0.1033
Minimum 0.1762 0.1199 0.2990 0.2887 0.1339 0.7708
Std. Dev 0.1278 0.0989 0.1770 0.1280 0.1311 0.1893
Skewness 0.2039 1.3668 1.7922 0.1600 0.6619 1.0276
Kurtosis 2.5429 7.4172 9.0181 2.3013 2.5608 3.5411
Jarque-Bera 3.8304 274.3371 500.8872 6.0041 19.8566 39.3350
Bai–Perron test 2010.06 2009.10 2010.11 2010.05 2009.10 2010.07
Chow test-χ 21 146.84 64.26 21.08 52.23 10.19 154.42
Observations 245 245 245 245 245 245
uncertainty
economic

483
policy
The impact in

Summary statistics of

between stock and

Table 2.

2019.03
bond returns: 2002.06–
monthly correlations
 11,4

484

2019.03
CFRI

Table 3.

returns: 2002.06–
monthly correlations
Summary statistics of

between stock and gold

US
ρt ðRTTMK ; RG Þ ρt ðRSHA; t ; RG Þ ρt ðRSHB; t ; RG Þ ρt ðRSZA ; RG Þ ρt ðRSZB ; RG Þ ρt ðRTTMK ; RGUS Þ

Mean 0.0989 0.2210 0.0843 0.1742 0.1507 0.0524

Median 0.0742 0.2071 0.1223 0.1594 0.1455 0.0685
Maximum 0.6381 0.8335 0.2491 0.7147 0.5336 0.2766
Minimum 0.3337 0.3661 0.6476 0.3110 0.4875 0.3677
Std. Dev 0.1957 0.2426 0.1426 0.2262 0.2299 0.1317
Skewness 0.2440 0.1570 2.8062 0.1272 0.2480 0.4217
Kurtosis 2.5661 2.4874 12.4159 2.1637 2.0632 2.6110
Jarque-Bera 4.3520 3.6731 1226.6120 7.7685 11.4699 8.2672
Bai–Perron test 2013.12 2008.09 2009.01 2016.07 2003.26 2015.02
Chow test-χ 21 15.02 35.78 58.34 42.80 35.36 13.24
Observations 245 245 245 245 245 245
Note(s): Bai–Perron test is used to locate the breakpoint, Chow test with χ 21 distribution rejects the null hypothesis regarding the absence of structural changes ρt ðRi ; Rj Þ
denotes a correlation between asset returns i and j
Chiang, 2019). Likewise, a rise in GPR variability ðjΔGPRt jÞ would lead to market instability The impact in
and cause delays in business investment decisions, which would result in low economic economic
activity and pessimistic economic forecasts that would depress advances in the stock market
(Pastor and Veronesi, 2013; Liu et al., 2017; Caldara and Iacoviello, 2019). Incorporating these
policy
qualitative elements into an analytical framework yields: uncertainty
  *
1  fp Lp bρt ¼ b0 þ b1 ΔEPUt þ b2 jΔGPRt j þ εt (7)
485
where  ρ*t is a 
b Fisher transformation of the asset return correlation coefficient, that is,
* 1þbρt
ρt ¼ 2 ln
b 1
ρt contains negative value (Chiang et al., 2007). The
, since the original b
1 −b
ρt

ð1 − fp Lp Þbρ*t term takes care of the bρ*t series, which may entail the pth orders of
autoregressions. To explain the time-varying nature of the coefficients, the arguments of a
change in economic policy uncertainty, ΔEPUt and uncertainty from change in geopolitical
risk, jΔGPRt j, are included.
The literature has demonstrated that a sudden rise in ΔEPUt or jΔGPRt j would interrupt
the flow of business operations, causing income uncertainty, which would tamp down liquidity
and lead to a decline in demand for assets [9]. This phenomenon may be called the income
uncertainty effect (Tobin, 1969; Bloom, 2009, 2014). Conversely, as ΔEPUt or jΔGPRt j lessens,
investors are likely induced to increase their demand for assets, which would drive a positive
correlation between asset returns (Hong et al., 2014).
On the other hand, the movement in investments could undergo a phenomenon, known as
a substitution effect, which sees risky assets (stocks) and safe haven assets (bonds and gold)
moving in opposite directions as uncertainty rises. Under this scenario, investors sell off their
riskier assets (stocks) and move their funds into safer assets (bonds and gold) as uncertainty
rises [10]. This shift results in a negative relation between stocks and bonds (gold) returns. As
uncertainty declines, however, investors then switch from lower return assets to higher
return assets, resulting in a negative correlation between risky and safe assets (Hong et al.,
2014; Li et al., 2015a, b). The ultimate impact on the correlations of these assets depends on the
relative influence from the income effect and substitution effect as well as the investors’
preference toward assets.
Estimates of equation (7) use a GED-GARCH procedure (Nelson, 1991; Bollerslev et al.,
1992; Li et al., 2005) and control for the autocorrelations. Estimates of stock-bond correlations,
which are reported in Table 4, indicate that coefficients for ΔEPUt display negative signs and
are statistically significant. The evidence suggests that investors governed by a risk adverse
attitude are inclined to move funds from stocks to bonds as uncertainty escalates. This
decoupling behavior is apparent as ΔEPUt rises. This result is consistent with the flight-to-
quality phenomenon in the US market as stock market volatility increases. The evidence is
documented by Connolly et al. (2005), d’Addona and Kind (2006), Li et al. (2015a, b), Fang et al.
(2017) and Arouri et al. (2016).
The coefficients of jΔGPRt j, however, exhibit mixed results. For the SZA and SZB
markets, the coefficients continue to display negative signs, indicating that investors in these
markets do not treat the impacts of ΔEPUt and jΔGPRt j differently in terms of attitude
toward risk/uncertainty. However, the same is not true for the SHA and SHB and TTMA
where investors react differently to an escalation of uncertainty from ΔEPUt vs. jΔGPRt j.
These results appear to demonstrate positive coefficients, which imply that income effect
dominates that of substitution effect (Hong et al., 2014; Li et al., 2015a, b) between stocks and
bonds for SHA and SHB investors as they confront a rise in jΔGPRt j:
The statistics in Table 5 report the parametric effects on stock-gold return correlations in
response to uncertainty. The results are more uniform, and the R-squares are higher than
 11,4

486
CFRI

Table 4.

change in
Time-varying

uncertainty and

geopolitical risk
absolute value of
in economic policy
correlations of stock-
bond returns that are
explained by a change
 2
Correlation C ΔEPUt jΔGPRt j AR(1) AR(2) AR(3) ω ε2t−1 σ 2t−1 R

b
ρ*t ðRTTMK ; Rb Þ 0.1405 0.0003 0.0005 0.5309 0.0968 0.1286 0.0245 0.2959 0.2436 0.45
32.49 6.13 3.92 22.72 5.19 8.02 0.74 0.38 0.17
b
ρ*t ðRSHA;t ; Rb Þ 0.0518 0.00005 0.0002 0.9380 0.0909 0.0000 0.0665 0.9705 0.77
11.31 3.81 7.63 34.68 4.13 0.16 0.81 21.59
b
ρ*t ðRSHB;t ; Rb Þ 0.1393 0.0004 0.0013 0.1856 0.0892 0.3027 0.2916 0.2284 0.08
15.76 5.75 2.30 12.40 5.21 0.20 0.22 0.06
b
ρ*t ðRSZA ; Rb Þ 0.0531 0.0003 0.0005 0.2240 0.1553 0.0227 0.0313 0.7003 0.09
29.52 9.71 5.02 29.88 8.57 0.21 0.03 0.48
b
ρ*t ðRSZB ; Rb Þ 0.1385 0.0001 0.0007 0.0541 0.0896 0.0589 0.5174 0.8587 0.02
44.66 5.95 7.31 17.90 13.06 0.33 0.12 1.89
* *
Note(s): bρt ðRi ; Rj Þ denotes a correlation between asset returns i (stock) and j (bond). bρt is the Fisher transformation of the original correlation coefficient. ΔEPUt is a
change in economic policy uncertainty. jΔGPRt j measures the uncertainty of the geopolitical policy risk. For each model, the first row reports the estimated coefficients,
the second row is the estimated t-statistics. The critical values of t-distribution at the 1%, 5% and 10% levels of significance are 2.63, 1.98 and 1.66, respectively
  2
Correlation C ΔEPUt jΔGPRt j AR(1) AR(2) ω ε2t−1 σ 2t−1 R

b
ρ*t ðRTTMK ; RG Þ 0.2967 0.0001 0.0003 0.9626 0.0016 0.4551 0.6303 0.87
6.64 4.08 6.39 164.80 1.00 0.95 2.27
b
ρ*t ðRSHA;t ; RG Þ 1.6121 0.0001 0.0004 0.8751 0.1212 0.0004 0.1460 0.9018 0.92
0.61 10.71 5.79 32.78 4.26 0.70 0.70 8.32
b
ρ*t ðRSHB;t ; RG Þ 1.3145 0.00002 0.0000 0.8330 0.1723 0.0001 0.0191 0.8720 0.96
1.33 3.97 3.54 69.31 14.59 1.33 0.39 7.55
*
b
ρt ðRSZA ; RG Þ 2.9296 0.00005 0.0011 0.1705 0.8266 0.0078 7.6465 0.0583 0.90
0.50 6.62 19.18 6.41 30.35 1.30 1.27 0.31
b
ρ*t ðRSZB ; RG Þ 2.5096 0.00002 0.0005 0.6087 0.3883 0.0019 2.998 0.070 0.96
1.47 5.97 33.50 22.18 14.09 2.58 1.91 1.69
Note(s): b ρ*t ðRi ; Rj Þ denotes correlation between asset returns i (stock) and j (gold). bρ*t is the Fisher transformation of the original correlation coefficient. ΔEPUt is a change
in economic policy uncertainty. jΔGPRt j measures the geopolitical policy risk uncertainty. For each model, the first row reports the estimated coefficients, the second row
is the estimated t-statistics. The critical values of t-distribution at the 1%, 5% and 10% levels of significance are 2.63, 1.98 and 1.66, respectively
uncertainty
economic

487
policy
The impact in

absolute value of

geopolitical risk
uncertainty and

change in
explained by a change
gold returns that are
Table 5.

in economic policy
correlations of stock-
Time-varying
CFRI those stock-bond return correlations. First, the coefficients of ΔEPUt for various stock-gold
11,4 correlations exhibit negative signs and are highly significant. The negative stock-gold
relation indicates that investors are inclined to switch from stocks to gold as ΔEPUt
increases. This trend could stem from a traditional view toward gold in Chinese society that is
seen as a better and universal instrument to hedge against uncertainty. Second, evidence
reveals that the estimated slope of jΔGPRt j is positive for stock-gold correlations and the
estimated t-statistics are highly significant. The comovements of stock and gold returns in
488 response to jΔGPRt j may be attributable to an increase in volatility for ΔGPR (Balcilar et al.,
2018). An increase in volatility generates fear that would delay investment and consumption
decisions. This results in an income effect, causing a decline in demand for both stocks and
gold. Note that heightened ΔGPR uncertainty could produce a substitution effect, as
investors replace stocks with gold in their portfolio to hedge against geopolitical risk. The
positive sign of the correlation for stock-gold relation implies the income effect outweighs the
substitution effect. In addition, low government bond yields, as observed by Spratt et al.
(2020), would induce investors searching for alternative assets to consider rising stock and
gold prices as an option to the recent ΔGPR. Third, the different signs for the coefficients of
ΔEPUt and jΔGPRt j imply that investors are able to discriminate between the potential
impact arising from economic uncertainty vs. political tension uncertainty. In summary, the
evidence clearly indicates that market participants are more likely to move their funds from
stocks to bonds or gold as EPUt increases; however, when the jΔGPRt j escalates, it would
raise stock market volatility (Balcilar et al., 2018), causing noise trades to sell off; however, the
rational traders would long stocks (Sentana and Wadhwani, 1992), leading to an increase in
demanding for stocks and gold [11]. The demand rise for stocks is due to an anticipation of
higher risk premium, while the demand upward shift for gold is to hedge risk. These two
forces drive a positive correlation between stock-gold returns [12]. (Hong et al., 2014).

5.5 Dynamic correlations based on rolling sample approach

Although the DCC models have been predominantly employed to derive the time-varying
correlations, some researchers prefer to use the rolling sample approach because of its
simplicity. Studies by Chiang (1988), Connelly et al. (2005) and McMillian (2018) are examples.
Thus, employing the rolling sample approach to generate the time-varying correlation is
worthy of examination. In this context, 36 observations are used as a fixed window length,
and the sample period is rolled ahead one month at a time to estimate the dynamic correlation
between stock and bond (gold) return relations. This approach provides a means of testing
market participant reactions to ΔEPUt and jΔGPRt j as news releases. Table 6 reports
the rolling correlations between stock-bond returns and Table 7 reports the rolling
correlations between stock-gold return that are explained by ΔEPUt and jΔGPRt j: In these
tables, the time-varying correlations are denoted by Corð$Þ, which are used to distinct from
the DCC estimators, b ρ*t ð$Þ. As before, the estimations in Tables 6 and 7 controls for
autocorrelations.
Evidence in Tables 6 and 7 suggests that the estimated results are qualitatively
comparable. First, coefficients of ΔEPUt for stock-bond return relations in Table 6
consistently show negative signs and are statistically significant. The outcomes hold true for
the stock-gold relations in Table 7, which displays a decoupling phenomenon and suggests
that a rise in ΔEPUt will cause flight-to-qualities in bonds and gold. Second, the coefficients of
jΔGPRt j are positive and t-statistics are significant at the 5%. The testing results, which
support the positive comovements of stock and gold returns in response to a heightened
jΔGPRt j; may result from a volatility-fear-income channel as noted by Hong et al. (2014) and
Balcilar et al. (2018). The evidence thus concludes that the estimated results are robust across
different methods in generating the time-varying correlation between asset returns [13].
  2
Correlation C ΔEPUt jΔGPRt j AR(1) AR(2) AR(3) AR(12) ω ε2t−1 σ 2t−1 R

CorðRTTMK ; Rb Þ 9.1795 0.0105 0.0090 0.9058 0.1620 0.0953 26.8253 1.8487 0.3898 0.83
36.71 4.04 7.22 152.06 14.21 11.39 0.79 0.83 0.70
CorðRSHA; t ; Rb Þ 41.5881 0.0051 0.0116 0.9866 0.0063 43.1650 0.6695 0.4717 0.92
2.96 2.06 1.29 191.02 6.15 1.45 1.56 2.23
CorðRSHB; t ; Rb Þ 16.7906 0.0076 0.0140 1.2018 0.2721 0.0188 156.2918 1.2258 0.3883 0.89
8.35 2.82 2.36 211.10 20.04 2.17 1.23 1.18 1.23
CorðRSZA ; Rb Þ 41.5920 0.0815 0.0453 0.9770 35.0429 0.5741 0.4244 0.92
1.04 2.68 0.99 64.58 3.01 3.58 4.32
CorðRSZB ; Rb Þ 5.7458 0.0695 0.0415 0.8614 0.1118 0.1751 0.9660 0.3089 0.0269 0.90
3.85 6.69 3.80 45.92 6.47 11.96 1.67 1.52 0.05
Note(s): CorðRi ; Rb Þ denotes correlation coefficient between stock returns i (stock) and b (bond) derived from rolling estimations with 36 months as a fixed window.
ΔEPUt is a change in economic policy uncertainty. jΔGPRt j measures the uncertainty of the geopolitical policy risk. For each model, the first row reports the estimated
coefficients, the second row is the estimated t-statistics. The critical values of t-distribution at the 1%, 5% and 10% levels of significance are 2.63, 1.98 and 1.66,
respectively
uncertainty
economic

489
policy
The impact in

absolute value of

geopolitical risk
that are explained by a

policy uncertainty and

change in
change in economic
of stock-bond returns
Rolling regression
Table 6.

correlation coefficients
 11,4

490
CFRI

Table 7.

change in
geopolitical risk
absolute value of
Rolling regression

changes in economic
of stock-gold returns
that are explained by
correlation coefficients

policy uncertainty and

2
Correlation C ΔEPUt jΔGPRt j AR(1) AR(2) AR(12) ω ε2t−1 σ 2t−1 R

CorðRTTMK ; RG Þ 0.5519 0.0001 0.0001 0.9380 0.0312 0.0011 0.3750 0.8094 0.91
16.42 11.09 2.29 92.53 3.39 0.52 0.97 3.88
CorðRSHA; t ; RG Þ 0.2014 0.0016 0.0011 0.2502 0.5962 0.1905 7.7797 20.9832 0.8411 0.73
27.87 14.96 5.09 13.15 19.44 14.21 1.05 0.18 1.48
CorðRSHB; t ; RG Þ 0.3388 0.0012 0.0005 0.7933 0.2288 0.0601 0.0594 4.8104 0.56 0.93
9.36 26.06 4.46 43.51 14.00 12.22 0.34 0.37 0.58
CorðRSZA ; RG Þ 0.0135 0.0001 0.0006 0.9133 0.0701 0.4315 0.2302 0.87
4.34 2.45 8.90 139.13 0.18 0.07 0.04
CorðRSZB ; RG Þ 0.2506 0.0003 0.0006 0.9423 0.0378 5.3456 0.0091 0.92
15.41 13.10 21.31 246.91 1.68 0.78 0.59
Note(s): CorðRi ; RG Þ denotes correlation coefficient between stock returns i (stock) and G (gold) derived from rolling estimations with 36 months as a fixed window.
ΔEPUt is a change in economic policy uncertainty. jΔGPRt j measures the geopolitical policy risk uncertainty. For each model, the first row reports the estimated
coefficients, the second row is the estimated t-statistics. The critical values of t-distribution at the 1%, 5% and 10% levels of significance are 2.63, 1.98 and 1.66,
respectively
Having demonstrated the correlations of stock-bond and stock-god returns, it is natural to The impact in
extend the analysis to the bond-gold return correlation, especially the latter has received less economic
attention in the literature. Two thoughts factor into establishing the correlations between
returns of bond and gold. The first is the return forgone by holding gold, the opportunity. For
policy
example, a higher interest rate would induce asset holders to switch from gold to bond uncertainty
holding. In contrast, a low or negative interest rate would lead to a demand for gold. The
second is an investor’s risk adverse attitude to uncertainty as EPU and GPR change. It is
conceivable that bond yield is more sensitive to economic policy changes, such as a central 491
bank’s monetary policy or a change in a government’s fiscal policy since interest rate changes
would accompany these policy changes. However, gold tends retain its value not only in
hedging against financial crisis, but also in storing value as geopolitical uncertainty
heightens. This is especially the case as a local currency loses value.
Table 8 reports the dynamic correlations for bond-gold return. This table contains both
DCC and rolling correlation estimates that are explained by ΔEPUt and ΔGPRt uncertainty.
The upper panel is based on ΔGPRt squared; the lower panel is measured by the absolute
ΔGPRt. The estimates clearly conclude that the coefficients of bond-gold return correlations
are negatively related to the ΔEPUt and jΔGPRt j, (or ΔGPR2t ),respectively. The evidence
appears to be consistent with the opportunity cost hypothesis and with different reactions to
the ΔEPUt vis-
a-vis to jΔGPRt j ðΔGPR2t Þ; which becomes much clearer in the following
section, which examines the individual asset returns in response to the ΔEPUt and jΔGPRt j,
respectively.

6. Evidence on individual asset returns

In the previous section, we examined the correlations of two asset returns, which revealed no
information on the reaction of an individual asset return in response to uncertainty or shock
changes. Since uncertainty changes have affected investors’ sentiments, aversion to
uncertainty may cause bonds, gold and stocks to load with different signs. More concrete
empirical evidence on individual asset returns will provide evidence to justify the directions
of correlation.
This section examines the impact of the key factors, ΔEPUt and jΔGPRt j, on asset returns
controlling for ΔVaRt and autocorrelations. The regression model is given below as:
 
1  fp Lp Ri; t ¼ c0 þ c1 ΔVaRt þ c2 ΔEPUt þ c3 jΔGPRt j þ εt (8)

where Ri; t: fRTTMK; t ; RSHA; t ; RSHB; t ; RSZA; t ; RSZB; t ; Rb; t ; RG; t g denotes returns from total
market stock, Shanghai A-shares, Shanghai B-shares, Shenzhen A-shares, Shenzhen B-
shares, Chinese 30-year bonds and gold market returns. The expression ð1 − fp Lp Þ represents
the current and lagged parameter operator applied to asset return i at time t, which is included
for controlling the autocorrelations up to order p. ΔVaRt is the change in the Value-at-Risk to
be used to measure the downside risk, which is obtained from the minimum value of the last
22 daily stock returns (Bali et al., 2009). The ΔVaRt series is multiplied by (1) before running
the regression [14]. Definitions of the other variables were defined earlier in the paper.
The rationale to include the downside risk is due to its significant information content,
which captures the financial market volatility and the higher moments of stock returns
implied in the Cornish–Fisher expansion (1937) as noted by Bali et al. (2009) and Chen et al.
(2018). Inclusion of ΔVaRt in the test equation helps to control for the influence that arises
from spurious correlation with ΔEPUt and jΔGPRt j in the test equation.
As argued by Bloom (2009, 2014), a rise in ΔEPUt will create uncertainty among economic
agents who will likely delay corporate investment and households’ consumption, which in
turn would impede economic activity and future cash flow. For this reason, the stock return
 11,4

492
CFRI

Table 8.

uncertainty
economic policy
Rolling regression

of bond-gold return

movements that are

correlation by using

uncertainty and GPR

both DDC and rolling
methods and test their
correlation coefficients

explained by change in
2
Correlation C ΔEPUt ΔGPR2t AR(1) ω ε2t−1 σ 2t−1 R

b
ρ*t ðRb ; RG Þ 0.0397 0.00004 0.00001 0.5211 0.0041 0.2750 0.3190 0.34
38.86 2.68 12.06 32.30 0.19 0.22 0.10
CorðRb RG Þ 9.7145 0.0029 0.0007 0.9565 50.8001 0.9402 0.3375 0.85
10.04 4.27 8.34 483.25 1.21 1.14 1.01
2
Correlation C ΔEPUt jΔGPRt j AR(1) ω ε2t−1 σ 2t−1 R

b
ρ*t ðRb ; RG Þ 0.0512 0.0001 0.0002 0.5655 0.0053 0.1533 0.0341 0.33
30.17 6.76 7.01 58.96 0.21 0.13 0.01
CorðRb RG Þ 11.903 0.0039 0.0479 0.9447 90.1315 0.9754 0.0745 0.85
7.00 4.24 15.13 253.86 1.61 1.06 0.22
Note(s): bρ*t ðRb ; RG Þ denotes correlation between bond return and gold return using DCC method, CorðRb ; RG Þ denotes correlation coefficient between bond return and
gold return using rolling correlation method. ΔEPUt is a change in economic policy uncertainty ΔGPR2t and jΔGPRt j measure the geopolitical policy risk uncertainty,
respectively. For each model, the first row reports the estimated coefficients, the second row is the estimated t-statistics. The critical values of t-distribution at the 1%, 5%
and 10% levels of significance are 2.63, 1.98 and 1.66, respectively
and ΔEPUt are expected to be negatively correlated. Evidence provided by Antonakakis The impact in
et al. (2013), Li et al. (2015b), Chen et al. (2017), Li (2017), Chan et al. (2019), Chiang (2019) and economic
Chen and Chiang (2020) confirms this expectation.
The effect of ΔEPUt on the bond yield is mainly conveyed through the term structure
policy
relation. Leippold and Matthys (2015) find that increased government policy uncertainty uncertainty
leads to a decline in bond yields. Bordo et al. (2015) find that policy uncertainty has a
significant negative effect on bank credit growth. When facing the dampening effect of
ΔEPUt on the economy, a monetary authority usually would lower the short rate, which 493
through the term structure relation further affects long-term interest rates. Of course, an
expected increase in money supply could generate a higher inflation expectation and induce a
heightened long-term interest rate.
When the economy is under extreme stress or experiencing political turbulence, gold is
perceived as a favorable instrument to hedge against uncertainty and becomes a safe haven
(Jones and Sackley, 2016; Baur and Lucey, 2010; Qin et al., 2020). In addition, gold has been
held as part of a central bank’s international reserve as an instrument to safeguard the value
of its national currencies (Baur and McDermott, 2010) and to avoid default risk or political
risk. From this perspective, the price of gold should respond positively to an increase in
ΔEPUt : Bilgin et al. (2018) report that a rise in EPU leads to increases in the price of gold.
However, they find that gold prices are less likely to decline when economic policy conditions
improve. Jones and Sackley (2016) find that increases in EPU contribute to increases in the
price of gold. Similarly, by using the nonparametric causality-in-quantiles, Raza et al. (2018)
show that change EPU causes gold prices to increase.
An escalation of ΔGPR uncertainty (ΔGPR2t or jΔGPRt j) has a similar effect on asset
returns (Brogaard et al., 2020). As documented by Caldera and Iacoviello (2019), an
aggravation in GPR tends to weaken economic activity and, in turn, stock returns. This is
essentially due to fact that ΔGPR can have an adverse effect on production and
consumptions, which is triggered by delays in investment as well as the postponement of
household spending on durable goods (Wade, 2019). This weakening in demand would
undoubtedly threaten potential profits and hence depress stock prices. A heightened level
of ΔGPR can also create greater volatility in financial markets as evidenced by Balcilar et al.
(2018), causing a flight-to-quality and an increase in demand for bonds (Fang et al., 2017) or
gold (Baur and Lucey, 2010). Thus, it is anticipated that a rise in ΔGPR or its
volatility would give rise to higher demand for gold, causing a decoupling phenomenon.
However, as mentioned earlier, there is an income/wealth effect, making stock and gold
complementary.
Estimates based on the use of the GED-GARCH procedure (Nelson, 1991) are reported in
Tables 9 and 10, which reveal several interesting findings. Focusing first on the stock returns,
we find that all of the estimated coefficients from uncertainty measures (Table 9), including
ΔVaRt ; ΔEPUt and ΔGPR2t , display negative signs and are significant at the 1% level. The
robustness test in Table 10, which replaces ΔGPR2t with jΔGPRt j; achieves the same
qualitative results. The evidence strongly suggests that risk or uncertainty would produce an
adverse effect on stocks regardless of whether the risk is from a stock market per se ðΔVaRt Þ,
economic policy uncertainty ðΔEPUt Þ, or geopolitical risk ðΔGPR2t or jΔGPRt jÞ.
Second, in response to the risk, investors intend to “fly to quality”, which can be seen in the
movement of the funds from stocks to bonds with a rise in policy uncertainty. This shift is
indicated by negative coefficients of stock returns along with positive coefficients of bond
returns with respect to ΔEPUt as shown in both Tables 9 and 10. The evidence is consistent
with the notion illustrated by Gulko (2002) and evidence documented by Jones and Sackley
(2016) and Lucey (2009).
Third, the evidence also indicates investors would “fly-to-gold”, which is seen in the
coefficients of gold returns that are positive with a heightened level of ΔGPR2t as shown in
 11,4

494
CFRI

Table 9.

stock returns
uncertainty and
geopolitical risk
risk, economic policy

squared on aggregate
Estimates of downside
Market return C ΔVaRt ΔEPUt ΔGPR2t ARð1Þ ARð11Þ ω ε2t−1 σ 2t−1 R2

RTTMK; t 0.992 0.032 0.042 0.0001 41.574 11.584 0.843 0.06

21.96 31.69 30.96 3.17 0.07 0.61 3.11
RSHA; t 0.576 0.025 0.023 0.0003 34.194 0.988 0.645 0.06
7.77 18.42 6.48 5.64 0.63 0.77 1.78
RSHB; t 0.919 0.019 0.043 0.001 48.018 0.651 0.647 0.03
9.37 10.53 5.63 13.77 0.69 0.83 1.77
RSZA; t 0.568 0.032 0.040 0.0003 47.974 0.798 0.629 0.05
4.82 14.54 16.73 4.37 0.59 0.70 1.41
RSZB; t 1.072 0.029 0.047 0.001 45.047 0.390 0.668 0.05
11.45 47.45 37.07 14.74 0.71 0.61 1.77
Rb; t 0.352 0.000003 0.00002 0.000003 0.972 0.001 0.584 0.017 0.97
34.21 0.90 8.45 15.40 334.48 0.80 0.66 0.02
RG; t 0.553 0.002 0.007 0.00004 0.108 5.452 0.994 0.776 0.04
15.62 2.69 12.73 2.56 9.48 0.47 0.70 2.92
Note(s): This table presents evidence of asset returns, Rt on ΔVaRt , ΔEPUt and ΔGPR2t using a GED-GARCH(1,1) procedure. The asset returns are: TTMK (Total stock
market index), SHA (Shanghai A-share index), SHB (Shanghai B-share index), SZA (Shenzhen A-share index), SZB (Shenzhen B-share index), Rb; t (bond return) and RG; t
(gold return). The independent variables are changes in downside risk ðΔVaRt Þ, economic policy uncertainty ðΔEPUt Þ) and geopolitical policy risk uncertainty ðΔGPR2t Þ:
AR(p) denotes autocorrelation with pth orders. The statistics of ω; ε2t and σ 2t in the variance equation are constant parameters. For each model, the first row reports the
estimated coefficients, the second row contains the estimated t-statistics. The critical values of t-distribution at the 1%, 5% and 10% levels of significance are 2.63, 1.98 and
1.66, respectively. R2 is the coefficient of determination
Market return C ΔVaRt ΔEPUt jΔGPRt j ARð1Þ ARð11Þ ω ε2t−1 σ 2t−1 R2

RTTMK; t 1.176 0.026 0.042 0.012 131.082 1.604 0.264 0.06

11.39 19.29 12.04 3.81 0.78 0.75 0.41
RSHA; t 0.459 0.026 0.020 0.009 18.559 1.009 0.673 0.06
13.26 18.61 5.16 2.31 0.57 0.95 2.39
RSHB; t 1.477 0.015 0.037 0.060 20.638 0.371 0.803 0.03
10.08 6.38 4.51 22.71 0.60 0.85 3.71
RSZA; t 0.792 0.028 0.033 0.013 46.415 1.085 0.612 0.05
11.36 11.53 8.69 2.13 0.57 0.72 1.39
RSZB; t 1.727 0.026 0.054 0.076 53.601 0.339 0.662 0.05
11.73 16.03 8.14 13.47 0.66 0.53 1.50
Rb; t 0.322 0.00003 0.00003 0.00010 0.952 0.001 1.016 0.057 0.97
58.17 8.53 6.66 8.37 479.09 0.96 0.88 0.08
RG;t 0.679 0.006 0.009 0.014 0.025 17.266 1.292 0.818 0.03
25.49 5.25 7.21 12.09 2.17 0.36 0.44 2.33
Note(s): This table presents evidence of asset returns, Rt on ΔVaRt , ΔEPUt and jΔGPRt j using a GED-GARCH(1,1) procedure. The asset returns are: TTMK (Total stock
market index), SHA (Shanghai A-share index), SHB (Shanghai B-share index), SZA (Shenzhen A-share index), SZB (Shenzhen B-share index), Rb; t (bond return) and RG; t
(gold return). The independent variables are changes in downside risk ðΔVaRt Þ, economic policy uncertainty ðΔEPUt Þ and geopolitical policy risk uncertainty ðjΔGPRt jÞ:
AR(p) denotes autocorrelation with pth orders. The statistics of ω; ε2t and σ 2t in the variance equation are constant parameters. For each model, the first row reports the
estimated coefficients, the second row contains the estimated t-statistics. The critical values of t-distribution at the 1%, 5% and 10% levels of significance are 2.63, 1.98 and
1.66, respectively. R2 is the coefficient of determination
uncertainty
economic

495
policy
The impact in

uncertainty and

returns
risk on aggregate stock
risk, economic policy
Table 10.

absolute geopolitical
Estimates of downside
CFRI Table 9 and jΔGPRt j as shown in Table 10 while the coefficients with stock returns are
11,4 negative. Thus, the evidence appears to have a clear substitution effect. However, the
statistics in Tables 9 and 10 do not contain the income effect or the wealth effect. Apparently,
the movement of funds depends on the nature and the source of risk/uncertainty along with
the substitution effect and income/wealth effect. For Chinese investors, gold is likely viewed
as a favorable instrument to hedge against ΔGPR uncertainty given the fact that bond
market is thin in the country. In addition, gold is known to be an asset to be portable and is a
496 flexible instrument to be converted into cash for fulfilling future investments or consumption.
Fourth, the evidence is consistent with the negative correlation between bond and gold
returns as evidenced by these asset returns, which display opposite reactions to the ΔEPUt
and jΔGPRt j ð or ΔGPR2t Þ. Evidence suggests that estimated results are robust with
different methods to be used to derive the time-varying correlations of stock-bond returns.

7. Conclusions
A substantial number of empirical studies have been done by focusing on stock market risk
and economic policy uncertainty (EPU) as main arguments to explain equity market
behavior; however, attention to geopolitical risk is relatively limited. This study highlights
the role of jΔGPRt j, along with ΔEPU, in explaining the impact of investors’ behavior on asset
returns. By focusing on Chinese financial markets, this study achieves several important
empirical findings. First, evidence indicates that the stock-bond relations are negatively
related with the ΔEPUt , indicating a flight-to-quality behavior. The evidence of a negative
correlation is consistent with Gulko (2002), Connolly et al. (2005), Baur and Lucey (2009),
Lucey (2009) and Li et al. (2015b) in US market. However, with respect to the uncertainty from
ΔGPR, only SZA and SZB markets display negative signs, this is consistent with the notion
that investors in these markets tend to move funds from stocks to bonds as ΔGPR rises.
Second, the evidence on the stock-gold return correlation in the Chinese market
consistently displays a positive relation, which differs from the finding in US market. In
particular, the stock-gold return correlation is positively correlated with escalating jΔGPRt j
(or ΔGPR2t ). This evidence concerning the Chinese market is consistent with the fact that the
income/wealth effect outweighs the substitution effect or the behavior that the rational
traders tend to purchase stocks and gold in facing a rise of jΔGPRt j in anticipating higher
prices in these assets. While testing the stock-gold return correlation in response to the
ΔEPUt ; the evidence uniformly shows a negative relation, indicating the presence of a flight-
to-quality phenomenon; but in this case the quality asset is gold.
Third, extending the analysis of asset return correlation, this study presents new evidence
on the time-varying correlations for bond-gold return relations. Interestingly, the results
consistently conclude that the coefficients of bond-gold return correlations are negatively
correlated with ΔEPUt and jΔGPRt j, ðor ΔGPR2t Þ, respectively. Evidence suggests that
Chinese market participants tend to use bond instruments to hedge against EPU; however, in
facing the geopolitical uncertainty, the behavior is more consistent with holding gold to hedge
the risk.
Fourth, by testing the impacts of ΔVaRt ; ΔEPUt and jΔGPRt j ðΔGPR2t Þ on each stock
returns, this study finds that all stock returns are negatively correlated with the uncertainty
measures, including changes in stock market downside risk, ΔEPUt ; and jΔGPRt j. This
holds true for the stocks in the aggregate market and markets in A-share and B-share stocks.
Fifth, the evidence consistently shows that the bond return is positively correlated with a
rise in ΔEPUt, rather than ΔVaRt and jΔGPRt jðor ΔGPR2t Þ; while the gold return is
positively correlated with jΔGPRt j ðor ΔGPR2t Þ not the risk from ΔVaRt and ΔEPUt,
indicating that investors are extremely willing to move funds into the gold market with an
increase in uncertainty of ΔGPR as measured byjΔGPRt jor ΔGPR2t . This result is consistent
with the notion of a flight-to-gold phenomenon. The preference toward gold in Chinese society The impact in
has a long tradition. This preference along with a high saving rate would also induce economic
households to increase their demand for gold as political uncertainty hits the market, since
gold is a more generally acceptable and universally recognized instrument to store value,
policy
which can be easily converted to a local currency for facilitating future consumptions. From uncertainty
this perspective, funds do not necessarily flow from the sale of stocks, which limits investors
by the asset constraint as described in the Tobin’s general equilibrium framework. Instead,
funds that could satisfy the demand for gold could come from savings due to a rapid growth 497
in Chinese GDP, promoting the income/wealth effect and moderating the substitution effect.

ORCID iDs
Thomas C. Chiang http://orcid.org/0000-0002-1586-3437

Notes
1. A survey conducted on the demand of global Central Banks for gold reserves indicates an increase
in 2020 due to low interest rates. The responses to the factors that influence the decision of holding
gold are rather informative. It is reported that 79% of respondents believe that gold stores long-term
value; 79% of respondents view gold’s performance during times of crisis as a valuable instrument;
74% of respondents see gold as having no default risk; 64% of respondents think gold has an
advantage in effectively diversifying portfolios, and 63% think gold lacks political risk (Goldhub,
2020). Thus, holding gold has unique merits.
2. The correlation between Chinese stock market volatility and EPU is 0.57 (t 5 10.52) and the
correlation of VaR and EPU is 0.24 (t 5 3.75). However, the correlations of stock return volatility and
VaR with GPR are low and statistically insignificant.
3. This study focuses on the Chinese stock market as it becomes the second largest in the world market
capitalization. As of 2019, the total value in China was: 8515.5 billion U.S. dollars. https://www.
theglobaleconomy.com/rankings/stock_market_capitalization_dollars/. The recent stock connect
programs (Shanghai-Hong Kong and the Shenzhen-Hong Kong Stock Connects) have provided
important financial gateways for both institutional and investors (Borst, 2017).
4. Some research uses a rolling regression method to generate the time-varying correlations (Chiang,
1988; Connolly et al., 2005). The analysis will be discussed in the due time.
5. Some researchers prefer to estimate the asset return on domestic macroeconomic factors, such as the
inflation rate, interest rate and other state variables first and use the resulting residual to examine
the correlations of asset return relation. (Yang et al., 2009; Pericoli, 2018).
6. This section follows closely to Engle (2009, pp. 45–49).
7. A big wave of new accounts occurred from late 2014 through early 2015. A little more than 40
million accounts were opened for the period of June 2014 ∼ May 2015.
8. Hu et al. (2018) provide a good review on the Chinese capital market.
9. Using ΔGPR2t or jΔGPRt j to serve as uncertainty measure of GPR is based on the notion provided
byDing et al. (1993) in their study of stock return properties.
10. A heightened GPR would more likely induce investors to move their funds into the gold market
rather than to the bond market, since the latter is subject to political risk as well. This statement
holds true without having government’s interference in the gold market.
11. Sentana and Wadhwani (1992) use the term of positive feedback trader as the noise trader and
negative feedback trader as the rational trader.
12. In re-estimating equation (7) by replacing jΔGPRt j with ΔGPR2t , the results are comparable to those
reported in Tables 5 and 6 Since ΔGPR2t tends to produce a smaller coefficient, the measure of
jΔGPRt j makes more sense.
CFRI 13. It should be noted that the rolling sample correlations tend to generate the serial correlation,
which could inflate the t-statistics. In addition, the estimated coefficients are subject to the choice
11,4 of fixed window length. However, our attempt sufficiently demonstrates the robustness of
the model.
14. Frank Knight makes a distinction between risk and uncertainty based on the measurability. Knight
(1921) notes that only quantifiable uncertainty is defined as risk. The news-based uncertainty
indices constructed by Baker et al. (2013) and Davis (2016) provide a measurability of EPU.
498
References
Allard, A.F., Lania, L. and Smedts, K. (2020), “Stock-bond return correlations: moving away from ‘one-
frequency-fits-all’ by extending the DCC-MIDAS approach”, International Review of Financial
Analysis, Vol. 71, 101557, doi: 10.1016/j.irfa.2020.101557.
Andersson, M., Krylova, E. and V€ah€amaa, S. (2008), “Why does the correlation between stock and
bond returns vary over time?”, Applied Financial Economics, Vol. 18, pp. 139-151.
Antonakakis, N., Chatziantoniou, I. and Filis, G. (2013), “Dynamic co-movements of stock market
returns, implied volatility and policy uncertainty”, Economics Letters, Vol. 120, pp. 87-92.
Arouri, M., Estay, C., Rault, C. and Roubaud, D. (2016), “Economic policy uncertainty and stock
markets: long run evidence from the US”, Finance Research Letters, Vol. 18, pp. 136-141.
Asness, Cliff (2003), “Fight the Fed model: the relationship between future returns and stock and bond
market yields”, Journal of Portfolio Management, Vol. 30, pp. 11-24.
Baele, L., Bekaert, G. and Koen, I. (2010), “The determinants of stock and bond return co-movements”,
Review of Financial Studies, Vol. 23, pp. 2374-2428.
Baker, S.R., Bloom, N., Davis, S.J. and Wang, X. (2013), Economic Policy Uncertainty in China,
unpublished paper, University of Chicago.
Balcilar, M., Bonato, M., Demirer, R. and Gupta, R. (2018), “Geopolitical risks and stock market
dynamics of the BRICS”, Economic Systems, Vol. 42, pp. 295-306.
Bali, T.G., Demirtas, K.O. and Levy, H. (2009), “Is there an intertemporal relation between downside
risk and expected returns”, Journal of Financial and Quantitative Analysis, Vol. 44, pp. 883-909.
Barboza, D. (2008), “China unveils $586 billion stimulus plan”, New York Times, Vol. 10 Nov, 2008,
available at: https://www.nytimes.com/2008/11/10/world/asia/10iht-10china.17673270.htmlBaur
(D.G. and).
Baur, D.G. and Lucey, B.M. (2009), “Flights and contagion—An empirical analysis of stock–bond
correlations”, Journal of Financial Stability, Vol. 5, pp. 339-352.
Baur, D.G. and Lucey, B.M. (2010), “Is gold a hedge or a safe haven? An analysis of stocks, bonds and
gold”, Financial Review, Vol. 45, pp. 217-229.
Baur, D.G. and McDermott, T.K. (2010), “Is gold a safe haven? International evidence”, Journal of
Banking and Finance, Vol. 34, pp. 1886-1898.
Bilgin, M.H., Gozgor, G., Lau, C.K.M. and Sheng, X. (2018), “The effects of uncertainty measures on the
price of gold”, International Review of Financial Analysis, Vol. 58, pp. 1-7.
Bloom, N. (2009), “The impact of uncertainty shocks”, Econometrica, Vol. 77, pp. 623-685.
Bloom, N. (2014), “Fluctuations in uncertainty”, Journal of Economic Perspectives, Vol. 28, pp. 153-176.
Bollerslev, T., Chou, R.Y. and Kroner, K.F. (1992), “ARCH modeling in finance: a review of the theory
and empirical evidence”, Journal of Econometrics, Vol. 52, pp. 5-59.
Bordo, M.D., Duca, J.V. and Koch, C. (2015), “Economic policy uncertainty and the credit channel:
aggregate and bank level U.S. evidence over several decades”, Federal Reserve Bank of Dallas,
Working Paper 1605, available at: https://www.dallasfed.org/-/media/documents/research/
papers/2016/wp1605.pdf.
Borst, N. (2017), “Connecting China’s stock markets to the world”, Federal Reserve Bank of San The impact in
Francisco, January 4, 2017, Pacific Exchange Blog, available at: http://www.frbsf.org/banking/
asia-program/pacific-exchange-blog/china-stock-market-connect-shenzhen/. economic
Brogaard, J., Dai, L., Ngo, P.T.H. and Zhang, B. (2020), “Global political uncertainty and asset prices”,
policy
Review of Financial Studies, Vol. 33, pp. 1737-1780, doi: 10.1093/rfs/hhz087. uncertainty
Caggiano, G., Castelnuovo, E. and Groshenny, N. (2014), “Uncertainty shocks and unemployment
dynamics in US recessions”, Journal of Monetary Economics, Vol. 67, pp. 78-92.
499
Caldara, D. and Iacoviello, M. (2017), Measuring Geopolitical Risk, working paper, Board of Governors
of the Federal Reserve Board, 2017.
Caldara, D. and Iacoviello, M. (2019), “Measuring geopolitical risk”, available at: https://www2.bc.edu/
matteo-iacoviello/gpr_files/GPR_PAPER.pdf.
Campbell, J.Y., Pflueger, C. and Viceira, L.M. (2020), “Macroeconomic drivers of bond and equity
risks”, Journal of Political Economy, Vol. 128, pp. 3148-3185, available at: https://scholar.
harvard.edu/files/campbell/files/campbellpfluegerviceira_20190510.pdf.
Chan, Y.-C., Saffar, W. and Wei, K.C.J. (2019), “How economic policy uncertainty affects the cost of
raising equity capital: evidence from seasoned equity offerings”, (June 24, 2019), available at
SSRN: https://ssrn.com/abstract53017843.
Chen, X. and Chiang, T.C. (2020), “Empirical investigation of changes in policy uncertainty on stock
returns – evidence from China’s market”, Research in International Business and Finance,
Vol. 48, published online, doi: 10.1016/j.ribaf.2020.101183.
Chen, J., Jiang, F. and Tong, G. (2017), “Economic policy uncertainty in China and stock market
expected returns”, Accounting and Finance, Vol. 57, pp. 1265-1286.
Chen, C.Y.-H., Chiang, T.C. and H€ardle, W.K. (2018), “Downside risk and stock returns in the G7
countries: an empirical analysis of their long-run and short-run dynamics”, Journal of Banking
and Finance, Vol. 93, pp. 21-32.
Chiang, T.C. (1988), “The forward rate as a predictor of the future spot rate - a stochastic coefficient
approach”, Journal of Money, Credit, and Banking, Vol. 20, pp. 210-232.
Chiang, T.C. (2019), “Financial risk, uncertainty and expected returns: evidence from Chinese equity
markets”, China Finance Review International, Vol. 9, pp. 425-454.
Chiang, T.C. (2020), “Risk and policy uncertainty on stock–bond return correlations: evidence from the
US markets”, Risks, Vol. 8, p. 58, published, June 2020, doi: 10.3390/risks8020058.
Chiang, T.C., Jeon, B.N. and Li, H. (2007), “Dynamic correlation analysis of financial contagion:
evidence from Asian markets”, Journal of International Money and Finance, Vol. 26,
pp. 1206-1228.
Chiang, T.C., Li, J. and Yang, S.-Y. (2015), “Dynamic stock-bond return correlations and financial
market uncertainty”, Review of Quantitative Finance and Accounting, Vol. 45, pp. 59-88.
Clark, J. (2020), “Why buy gold? 10 reasons to invest in physical gold bullion”, available at: https://
goldsilver.com/blog/why-own-gold-top-10-reasons/ (accessed 20 November 2020).
Connolly, R.A., Stivers, C.T. and Sun, L. (2005), “Stock market uncertainty and the stock-bond return
relation”, Journal of Financial and Quantitative Analysis, Vol. 40, pp. 161-194.
Cornish, E.A. and Fisher, R.A. (1937), “Moments and cumulants in the specification of distribution”,
Review of the International Statistical Institute, Vol. 5, pp. 307-320.
Davis, S. (2016), “An index of global economic policy uncertainty”, NBER Working Paper 22740,
available at: http://faculty.chicagobooth.edu/steven.davis/pdf/GlobalEconomic.pdf.
Davis, S.J., Liu, D. and Sheng, X. (2019), “Economic policy uncertainty in China since 1949: the view
from mainland newspapers”, Fourth Annual IMF-Atlanta Fed Research on China’s Economy,
available at: https://www.frbatlanta.org/-/media/documents/news/conferences/2019/0919-
workshop-on-chinas-economy/presentations/davis.pdf.
CFRI d’Addona, S. and Kind, A.H. (2006), “International stock–bond correlations in a simple affine asset
pricing model”, Journal of Banking and Finance, Vol. 30, pp. 2747-2765.
11,4
Ding, Z., Granger, C.W. and Engle, R.F. (1993), “A long memory property of stock market returns and
a new model”, Journal of Empirical Finance, Vol. 1, pp. 83-106.
Engle, R.F. (2009), Anticipating Correlations: A New Paradigm for Risk Management, Princeton
University Press, Princeton.
500 Fang, L., Yu, H. and Li, L. (2017), “The effect of economic policy uncertainty on the long-term
correlation between US stock and bond markets”, Economic Modelling, Vol. 66, pp. 139-145.
Goldhub (2020), “2020 central bank gold reserve survey”, World Gold Council, available at: https://
www.gold.org/goldhub/data/2020-central-bank-gold-reserve-survey?.
Guidolin, M. and Timmermann, A. (2006), “An econometric model of nonlinear dynamics in the joint
distribution of stock and bond returns”, Journal of Applied Econometrics, Vol. 21, pp. 1-22.
Gulko, L. (2002), “Decoupling”, Journal of Portfolio Management, Vol. 28, pp. 59-66.
Hakkio, C. and Keeton, W.R. (2009), “Financial stress: what is it, how can it be measured, and why
does it matter?”, Federal Reserve Bank of Kansas City Economic Review, Vol. 2, pp. 5-50.
Hillen, M.A., Gutheil, C.M., Strout, T.D., Smets, E.M.A. and Han, P.K.J. (2017), “Tolerance of
uncertainty: conceptual analysis, integrative model, and implications for healthcare”, Social
Science and Medicine, Vol. 180, pp. 62-75.
Hong, G., Kim, Y. and Lee, B.-S. (2014), “Correlations between stock returns and bond returns: income
and substitution effects”, Quantitative Finance, Vol. 14, pp. 1999-2018.
Hu, G.X., Pan, J. and Wang, J. (2018), “Chinese capital market: an empirical overview”, NBER Working
Papers 24346, National Bureau of Economic Research.
Jones, A.T. and Sackley, W.H. (2016), “An uncertain suggestion for gold-pricing models: the effect of
economic policy uncertainty on gold prices”, Journal of Economics and Finance, Vol. 40,
pp. 367-379.
Knight, F. (1921), Risk, Uncertainty, and Profit, 5th ed., Dover Publications, New York, NY.
Kwan, S.H. (1996), “Firm-specific information and the correlation between individual stocks and
bonds”, Journal of Financial Economics, Vol. 40, pp. 63-80.
Leippold, M. and Matthys, F.H.A. (2015), “Economic policy uncertainty and the yield curve”, available
at SSRN: https://ssrn.com/abstract52669500.
Li, X. (2017), “New evidence on economic policy uncertainty and equity premium”, Pacific-Basin
Finance Journal, Vol. 46, pp. 41-56.
Li, Q., Yang, J., Hsiao, C. and Chang, Y. (2005), “The relationship between stock returns and volatility
in international stock markets”, Journal of Empirical Finance, Vol. 12, pp. 650-665.
Li, X., Balcilar, M., Gupta, R. and Chang, T. (2015a), “The causal relationship between economic policy
uncertainty and stock returns in China and India: evidence from a bootstrap rolling-window
approach”, Emerging Markets Finance and Trade, Vol. 52, pp. 674-689.
Li, X., Zhang, B. and Gao, R. (2015b), “Economic policy uncertainty shocks and stock-bond correlation:
evidence from the US market”, Economic Letters, Vol. 132, pp. 91-96.
Liu, L.X., Shu, H. and Wei, K.C.J. (2017), “The impacts of political uncertainty on asset prices: evidence
from the Bo scandal in China”, Journal of Financial Economics, Vol. 125, pp. 286-310.
Lucey, B.M. (2009), “Flights and contagion—an empirical analysis of stock–bond correlations”,
Journal of Financial Stability, Vol. 5, pp. 339-352.
McMillan, D.G. (2018), “The behaviour of the equity yield and its relation with the bond yield: the role
of inflation”, International Journal of Financial Studies, MDPI, Basel, Vol. 6, pp. 1-18.
Nelson, D. (1991), “Conditional heteroskedasticity in asset returns: a new approach”, Econometrica,
Vol. 59, pp. 347-370.
O’Connor, F.A., Lucey, B.M., Batten, J.A. and Baur, D.G. (2015), “The financial economics of gold - a The impact in
survey”, International Review of Financial Analysis, Vol. 41, pp. 186-205.
economic
Pastor, L. and Veronesi, P. (2013), “Political uncertainty and risk premia”, Journal of Financial
Economics, Vol. 110, pp. 520-545.
policy
Pericoli, M. (2018), “Macroeconomics determinants of the correlation between stocks and bonds”, Bank
uncertainty
of Italy Working Paper, No. 1198, available at: SSRN: https://ssrn.com/abstract53429148http://
dx.doi.org/10.2139/ssrn.3429148.
501
Qin, M., Su, C.-W., Xiao, Y.-D. and Zhang, S. (2020), “Should gold be held under global economic policy
uncertainty?”, Journal of Business Economics and Management, Vol. 21, pp. 725-742, doi: 10.
3846/jbem.2020.12040.
Raza, S.A., Shah, N. and Shahbaz, M. (2018), “Does economic policy uncertainty influence gold prices?
Evidence from a nonparametric causality-in-quantiles approach”, Resources Policy, available at:
https://www.researchgate.net/publication/325285667_Does_economic_policy_uncertainty_
influence_gold_prices_Evidence_from_a_nonparametric_causality-in-quantiles_approach.
Sentana, E. and Wadhwani, S. (1992), “Feedback traders and stock return autocorrelations: evidence
from a century of daily data”, Economic Journal, Vol. 102, pp. 415-425.
Spratt, S., Mogi, C. and Vasquez, J. (2020), “U.S. bond markets are driving force behind the new gold
rush”, July 23, 2020, 3:08 AM EDT, available at: https://www.bloomberg.com/news/articles/
2020-07-22/ever-lower-u-s-bond-yields-are-driving-gold-s-relentless-rally.
Stock, J.F. and Watson, M.W. (1988), “Variable trends in economic time series”, Journal of Economic
Perspectives, Vol. 2, pp. 147-174.
Tobin, J. (1969), “A general equilibrium approach to monetary theory”, Journal of Money, Credit and
Banking, Vol. 1, pp. 15-29.
Tripathy, N. (2016), “A study on dynamic relationship between gold price and stock market price in
India”, European Journal of Economics, Finance and Administrative Sciences, available at:
https://www.researchgate.net/publication/306262190_A_Study_on_Dynamic_Relationship_
between_Gold_Price_and_Stock_Market_Price_in_India.
Wade, K. (2019), “Measuring the market impact of geopolitics”, Foresight, pp. 1-12, available at: https://
www.schroders.com/en/sysglobalassets/digital/insights/2019/pdfs/2019_sept_measuring-the-
market-impact-of-geopolitics_kw_il_cs1696.pdf.
Whaley, R.E. (2009), “Understanding the VIX”, Journal of Portfolio Management, Vol. 35, pp. 98-105.
Yang, J., Zhou, Y. and Wang, Z. (2009), “The stock–bond correlation and macroeconomic conditions:
one and a half century of evidence”, Journal of Banking and Finance, Vol. 33, pp. 670-680.
Yardeni, E. (1997), “Fed’s stock market model finds overvaluation. US equity research”, Deutsche
Morgan Grenfell, available at: http://www.yardeni.com/premiumdata/t_970825.pdf.

Further reading
Baker, S.R., Bloom, N. and Davis, S.J. (2016), “Measuring economic policy uncertainty”, Quarterly
Journal of Economics, Vol. 131, pp. 1593-1636.
Zhou, Y., Han, L. and Yin, L. (2017), “Is the relationship between gold and the US dollar always
negative? The role of macroeconomic uncertainty”, Applied Economics, Vol. 50, pp. 354-370.

Corresponding author
Thomas C. Chiang can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]