Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Economic Uncertainty and Exchange Market Pressure: Evidence From China:

Economic Uncertainty and Exchange Market Pressure: Evidence From China: This paper evaluates the impact of local and external economic uncertainty shocks on China’s exchange market pressure from a time-varying perspective. We first construct a simple behavioral model to provide some economic background for our empirical analysis. The model identifies two channels, namely market sentiment and market demand, through which economic uncertainty has a time-varying impact on exchange market pressure. Notably, we calculate a new exchange market pressure index for China by considering China’s practice. Using the economic policy uncertainty index as a proxy for economic uncertainty and our new estimates of China’s exchange market pressure, we employ a novel TVP-VAR model controlling for over-parameterization to a monthly dataset from January 2001 to October 2020 to analyze the impact of China’s and U.S. economic policy uncertainties on China’s exchange market pressure. Our empirical findings robustly show that an upsurge in U.S. EPU is followed by appreciation pressure of the RMB against the dollar, while a hike in China’s EPU triggers devaluation pressure on the RMB. In addition, the impact of economic policy uncertainty has considerable time variation in magnitude, especially after mid-2011, showing a trade-off mechanism between the effects of domestic and foreign economic uncertainties on China’s exchange market pressure. Moreover, we attribute the time- varying features to the changes in China’s dependence on the U.S. and in the exchange rate flexibility. Finally, China should further improve the exchange rate flexibility, reduce its dependence on the U.S. and develop a more diversified currency basket in the exchange rate formation mechanism. JEL: F31 C50 E70 Keywords economic Uncertainty, EMP, EPU, TVP-VAR, Over-parameterization pressures on the exchange rate to depreciate or appreciate. Introduction Ideally, these pressures will be fully reflected in the There is an anecdotal story about globalization that eco- exchange rate movements for the countries with floating nomic activity in an open economy is primarily driven by exchange rate regimes. However, this would not happen in foreign shocks, while domestic shocks are reflected in its the economies like China and Japan, where the central currency value (Nilavongse et al., 2020). Since economic banks are known for intervening in the foreign exchange uncertainty is a representative shock that conveys informa- market (hereafter FX market). tion about overall economic conditions, various attempts In China, government intervention in the FX market is a have been made to assess the impact of economic uncertainty stylized fact (Das, 2019; Wang et al., 2020). Accordingly, on the exchange rate, see Kido (2016), Simo-Kengne et al. since the exchange rate is managed to change, depreciation (2018), Nasir and Morgan (2018), Nilavongse et al. (2020), or appreciation pressures on the renminbi against other cur- among others. Granted that most of the literature argues that rencies cannot be manifested totally in the exchange rate the exchange rate is affected by local and external uncertain- movements. Thus, uncertainty shocks may have a negligible ties, but Simo-Kengne et al. (2018) and Nasir and Morgan impact on the realized exchange rate in China owing to cen- (2018) highlight that foreign uncertainty plays a small role in tral bank intervention, although it has been argued that driving the exchange rate. Theoretically, economic uncertainty is the primary source of risk of holding a currency (Hu, 1997). Elevated Nantong University, Jiangsu, China economic uncertainty can make it challenging to predict Corresponding Author: the exchange rates (Beckmann & Czudaj, 2017a, 2017b), Lin Liu, Nantong University, Room 515, Teaching building #10, No.9, leading agents to reduce their demand for the currency or Seyuan Road, Chongchuan District, Nantong, Jiangsu 226019, China. raise the risk premium (Taylor, 1989), thereby putting Email: liulintyu@outlook.com Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License (https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage). 2 SAGE Open market expectations are affected by the uncertainty shocks 2013), exchange rate dynamics are often interpreted as non- (Beckmann & Czudaj, 2017a, 2017b; Ter Ellen et al., 2013). linear processes in the literature; see Sengupta and Sfeir Naturally, when analyzing the impact of uncertainty for (1998), Mahajan and Wagner (1999), Imbs et al. (2003), the economies that resort to FX intervention, the realized Nakagawa (2010), among others. Second, the impact of eco- exchange rate needs to be restored to incorporate govern- nomic uncertainty shocks may be nonlinear due to changing ment intervention. A well-known measure of such an dynamics, policy regimes, and economic shocks (Mumtaz & exchange rate recovery is the exchange market pressure Theodoridis, 2018). Third, China has undergone many dra- (hereafter EMP) index. While most previous studies have matic phases over the past decades (Chang et al., 2016) and focused on the realized exchange rates and exchange rate has substantially reformed its exchange rate system, raising expectations, the impact of economic uncertainty on EMP the possibility of potential nonlinearities and multi-equilibria has been less touched upon, except for two path-breaking (Liu, 2021). Therefore, it is reasonable and desirable to apply works by Olanipekun, Güngör et al. (2019) and Olanipekun, a nonlinear model when exploring the relationship between Olasehinde-Williams et al. (2019). They, however, only aim the foreign exchange market and the macroeconomy. at the causal relationships between economic uncertainty and Theoretically, any nonlinear model can be approximated EMP. Two recent studies by Olasehinde-Williams and by a time-varying parameter (TVP) model (Granger, 2008). Olanipekun (2020) and Olanipekun and Olasehinde-Williams Given the appeal of TVP models, a few studies regarding (2021) provide empirical evidence that U.S. uncertainty exchange rate dynamics have utilized TVP-VAR models to affects EMP in some emerging market countries and African examine the effects and causes of exchange rate movements; economies, respectively. Nonetheless, these two studies do see Choi et al. (2018), Nasir and Morgan (2018), Zheng et al. not consider the impact of domestic uncertainty on EMP, (2019), Nasir and Vo (2020), and others. However, assuming which may be jointly motivated by local and external factors, that all model parameters are time-dependent increases as derived from the theory of exchange rate determination model complexity and may suffer from the “curse of dimen- (Mussa, 1984). A similar issue also exists in the study by Liu sionality” of over-parameterization, which would eventually (2020), who explores the impact of domestic uncertainty bias the estimates. Recent studies have shown their concerns shocks on China’s EMP, overlooking the potential effects over this issue; see, for example, Koop and Korobilis (2013), exerted by foreign uncertainty. However, because of China’s Eisenstat et al. (2016), Huber et al. (2019), Chan et al. (2020). high dependence on the United States, China’s financial mar- On this account, using a monthly dataset running from kets may endure the significant impact of uncertainty shocks January 2001 to October 2020, we conduct our empirical from the United States (Gupta et al., 2020; Liu, 2021). study with a TVP-VAR model by combing the approach Against this backdrop, we attempt to fill this gap by inspect- developed by Eisenstat et al. (2016) to control for ing China’s EMP dynamics following unexpected rises in over-parameterization. domestic and foreign economic uncertainties. Nevertheless, an essential issue that setting a proper proxy In this paper, we explicitly assess the role of economic for economic uncertainty arises in our empirical study. uncertainty in governing EMP with China’s monthly data. Currently, there are three optional proxies for economic China has been pursuing the globalization of the renminbi uncertainty in the literature; the news or reports-based eco- and building a responsible national image. Hence, maintain- nomic policy uncertainty index (EPU) (Baker et al., 2016), ing the exchange rate stability is a pressing concern for the real data-based macroeconomic uncertainty index (Jurado China. As the exchange rate stability is critical to the macro- et al., 2015), and the survey-based uncertainty index economy and plays a crucial part in the Chinese govern- (Bachmann et al., 2013). Comparatively, the news-based ment’s objectives (Das, 2019), evaluating the impact of uncertainty index is computed by text-mining newspaper economic uncertainty on EMP would benefit the Chinese reports available equally to every market participant. policymakers to better understand the underlying causes of Therefore, the EPU index would catch more the uncertainty exchange rate fluctuations. caused by specific events (Shin et al., 2018), as these events Specifically, as documented in Pástor and Veronesi (2013) tend to evoke public or political concerns about the economic and Krol (2014), the impact of economic uncertainty on consequences, leading to an increase in the frequency of EMP might be state-dependent, and thus motivated by economic-relevant news and thus a higher EPU. In addition, Bartsch (2019), we construct a simple behavioral model with the news reports used to construct EPU are usually easy to heterogeneous agents to demonstrate that the impact of eco- understand for market participants and may sometimes sig- nomic uncertainty on EMP, mainly through the market senti- nal policy stance, especially for countries where the govern- ment and market demand channels, evolves over time. ment controls public media in some way. Davis et al. (2019) Importantly, as our analytical result indicates, it is neces- construct an EPU index for China by retrieving news reports sary to consider nonlinearity when modeling EMP and eco- from two influential newspapers circulating in mainland nomic uncertainty. First, since the FX market has been China, namely the People’s Daily and the Guangming Daily. surveyed as dominated by heterogeneous agents using differ- Since these two government-run newspapers are the fore- ent trading strategies (de Jong et al., 2010; Ter Ellen et al., most government mouthpieces in China (Qin et al., 2018), Liu 3 market participants may translate an upsurge in the frequency index representing China’s exchange market pressure by of economic-relevant coverage into a rise in economic uncer- considering the occasional expenditure of FX reserves and tainty and policy concerns. Therefore, in this paper, we use the intervention implemented in recent years via the counter- EPU as a proxy for economic uncertainty. cyclical adjustment factor. Doing so distinguishes our paper Additionally, as the largest member holding dollar assets, in a significant way from several previous studies, for exam- China has been witnessed as a dollar-pegged country for a ple, Olanipekun, Güngör et al. (2019), Olanipekun, long time (Tervala, 2019), even though moving toward a Olasehinde-Williams et al. (2019), and Liu (2020), who all more flexible exchange rate regime after the reform initiated use slightly biased EMP estimates for China. in July 2005. Moreover, trading of the renminbi against the Third, in line with the theory of exchange rate determina- U.S. dollar dominates China’s onshore spot FX market, tion, an exchange rate is determined concomitantly by accounting for more than 95% of the overall FX transactions domestic and foreign fundamentals, which have been well- on average. Therefore, in our empirical section, we estimate documented to be affected by the uncertainty shocks (Bloom, a new EMP index of the RMB against the dollar by consider- 2009, 2014). Moreover, the uncertainty has been found to ing the contingent spending of FX reserves and the interven- spill over globally, especially from advanced economies to tion through the counter-cyclical adjustment factor, which less developed countries (Liu, 2021). Hence, in this paper, have been overlooked in most previous related studies, see, unlike Liu (2020), Huynh et al. (2020) and others who only for example, Olanipekun, Güngör et al. (2019), Olanipekun, consider the impact of local (or external) uncertainty, we Olasehinde-Williams et al. (2019), and Liu (2020). To this assess the impact of China’s and U.S. uncertainties on end, we estimate a new EMP index for China using daily China’s EMP in a consistent model. exchange rate data from August 11, 2015, when China Finally, as outlined above, we resort our analysis to a authority announced adopting a new regime for formulating novel TVP-VAR model to incorporate potential nonlineari- the central parity of the RMB/USD exchange rate, to October ties. Liu (2020) also uses a classic TVP-VAR model to evalu- 30, 2020, the last trading day of our sample. ate the impact of macroeconomic and financial uncertainties The empirical findings based on a monthly dataset for on EMP for China. Differently, considering the over-param- China from January 2001 to October 2020 show that a ris- eterization problem arisen from the time-varying specifica- ing shock in U.S. EPU would trigger the appreciation pres- tions, we apply a typical TVP-VAR model with the stochastic sure on the RMB against the dollar, while a hike in China’s model specification (SMSS) framework proposed by EPU is followed by the devaluation pressure of the RMB Eisenstat et al. (2016) to implement our empirical analysis. against the dollar. Moreover, the magnitude of the EPU The rest of our paper is arranged in the following manner. impact on EMP is found to be time-dependent. Specifically, The next section briefly reviews the current literature rele- the effect is relatively stable until mid-2011 but more vola- vant to this paper. Section 3 illustrates the economic intuition tile thereafter. As for the rationale for this time-varying that economic uncertainty affects exchange market pressure impact, we attribute it to changes in China’s dependence on through a simple behavioral exchange rate model. Section 4 the U.S. and increased exchange rate flexibility. Further, describes the empirical methodology and the data we use. the empirical evidence indicates a trade-off mechanism The empirical results and discussions, as well as robustness between the effects of China’s and U.S. economic policy checks, are analyzed in Section 5. The last section concludes uncertainties on China’s EMP. the whole paper and highlights the policy implications. Our marginal contribution to the growing literature is that we focus on the time-varying impact of local and external Related Literature economic uncertainties on China’s exchange market pres- sure. Our work reaches a step closer to an understanding of Overall, our work is related to two strands of the existing how the FX market responds to the shocks originating from literature. One of these strands has been devoted to uncov- domestic and foreign economic uncertainties. First, unlike ering the link between economic uncertainty and exchange Kido (2016), Nasir and Morgan (2018), Huynh et al. (2020), rate movements, whereby the focus, however, has been pri- Nasir (2020), Chen et al. (2020), among others, who center marily on exchange rate returns. Another direction of the on the roles of uncertainty in the exchange rate dynamics, we literature, which is the closest to our work, has been to attempt to explore the impact of uncertainty on China’s EMP. assess the role of economic uncertainty in provoking pres- Since the RMB exchange rate remains under control, its fluc- sures on the exchange rate. tuations are less informative in revealing market pressures. The impact of economic uncertainty on exchange rate Second, as documented in The Economist (2020), China’s movements (both returns and volatilities) is increasingly intervention in the FX market has become less pronounced in well documented. Earlier studies find pricing effects of recent years, keeping FX reserves fairly intact. Therefore, a uncertainty on the currency risk premium; see, for example, better index capturing exchange market pressure should con- Taylor (1989) and Hu (1997). Recently, there has been evi- sider the changes in FX reserves and the veiled measures as dence for the roles of home and foreign uncertainties in well. Consequently, we estimate a new and more accurate explaining exchange rate movements in industrial and 4 SAGE Open emerging economies, albeit analyzing mainly the spillovers particular, a bidirectional relation between global EPU and of U.S. EPU. However, while the impact of home-grown EMP and a one-way causality from EMP to EPU were found uncertainty on exchange rate stability is well understood, for China. Further, Olanipekun, Olasehinde-Williams et al. the findings on the influence of foreign uncertainty are (2019) expand the sample to 20 countries and use four EMP mixed. Compared to developed countries, Krol (2014) finds measures as in Aizenman and Binici (2016) to conclude that that exchange rate volatilities in emerging economies are domestic EPU and EMP are cointegrated in the long run. less affected by U.S. EPU during recessions. The author Most recently, Olasehinde-Williams and Olanipekun (2020) attributes the explanation to the low level of financial open- and Olanipekun and Olasehinde-Williams (2021) report ness of these economies to the United States. However, more evidence about the causality from uncertainty to EMP Huynh et al. (2020) offer empirical findings that exchange for African economies and emerging market countries, rate returns and volatilities of nine international currencies respectively. Moreover, based on China’s data, Liu (2020) against the U.S. dollar are associated with U.S. trade policy finds that macroeconomic uncertainty and financial uncer- uncertainty and global economic policy uncertainty. Chen tainty weaken (strengthen) EMP in a state of RMB apprecia- et al. (2020) show that the EPU shocks resulting from the tion (depreciation). United States, Europe, and Japan have significant and The recent study by Liu (2020) is undoubtedly the closest asymmetric effects on the Chinese onshore RMB/USD to ours. As we will do in our paper, Liu (2020) uses a TVP- exchange rate volatility. Nevertheless, the exchange rate VAR model to investigate the impact of China’s domestic volatilities of developed countries may be less exposed to uncertainty on EMP and the jump risk of the RMB/USD U.S. uncertainty shocks. Nasir and Morgan (2018) and exchange rate. However, there are several differences. First, Nilavongse et al. (2020) highlight the significant devalua- Liu (2020) jointly estimates the effects of macroeconomic tion effect of uncertainty locally from the UK on the ster- uncertainty and financial uncertainty measures, estimated ling exchange rate. Notably, the latter and Simo-Kengne using the methodology developed by Jurado et al. (2015). et al. (2018) reckon that foreign EPU is not a determinant Moreover, Liu (2020) models the impact of domestic uncer- of the exchange rate. In addition, Bartsch (2019) documents tainty only, but it is necessary to incorporate foreign uncer- that UK EPU rather than U.S. EPU impairs the stability of tainty since, in theory, the exchange rate is determined by the USD/GBP exchange rate. domestic economic conditions and foreign counterparts However, a growing literature demonstrates the salient simultaneously. Second, Liu (2020) directly uses the EMP impact of foreign uncertainty on the exchange rate returns. measure estimated by Patnaik et al. (2017). A closer scrutiny Kido (2016) finds that the real effective exchange rate is of Patnaik et al. (2017)’s EMP estimates for China reveals negatively correlated with U.S. EPU for Australia, Brazil, that they have overlooked China’s intervention through Korea, and Mexico except for Japan, which exhibits a posi- adjusting the central parity rate in recent years and the irregu- tive pattern, indicating that the yen plays as a safe-haven cur- lar spending of FX reserves. However, we estimate a new rency in the face of the U.S. uncertainty. Beckmann and Chinese EMP index by considering these practices. Third, Liu Czudaj (2017a) further corroborate the yen’s safe-haven (2020) relies on a four-variable TVP-VAR model of uncer- standing, though they concentrate on the aftermath of uncer- tainty and market stability to achieve an empirical analysis. tainty on the exchange rate expectations. Addressing the Nevertheless, as documented in Aizenman and Binici (2016), spread between the onshore and offshore RMB/USD variables regarding market sentiment and market demand, as exchange rates, a recent paper by Li et al. (2020) documents well as macroeconomic fundamentals, which might change in a finding that a rise in their constructed composite EPU the light of economic uncertainty, could be important factors (extracted from the EPU indices of China and G7) widens affecting EMP. Therefore, we will consider these variables to the spread. In addition, a closely related literature in this line assess the impact of economic uncertainty on EMP. In doing of research centers on the impact of relative uncertainty. so, we can better explore the underlying mechanism of how Focusing on the relative value of difference or ratio of economic uncertainty shapes EMP. Finally, we differ mark- domestic EPU to external EPU, Balcilar et al. (2016), edly from previous studies in that we illustrate some eco- Christou et al. (2018), and Zhou et al. (2020) show that the nomic background for our empirical study through a simple relative EPU has much power in predicting exchange rate behavioral exchange rate model with heterogeneous agents. movements. Another strand of the literature has attempted to dissect A Simple Behavioral Exchange Rate the relation between economic uncertainty and EMP. Model Olanipekun, Güngör et al. (2019) study a case of four BRIC countries but give somewhat puzzling results. In the paper, We extend the FX market model in Dieci and Westerhoff Olanipekun, Güngör et al. (2019) find the one-way causation (2010) to incorporate economic uncertainty. Specifically, we from global EPU to EMP and mutual interplays between consider a world consisting of two-country and a domestic domestic EPU and EMP in all four countries, which is incon- FX market in which two currencies are traded. Two types of sistent with the results of the country-specific analysis. In traders, namely fundamentalists and chartists, invest in the Liu 5 market. Moreover, investors switch between these two trad- 1 W = , ing strategies depending on market conditions. To conve- (5) 1+− fS S () tt niently describe exchange market pressure, we assume no trading restrictions in the market (e.g., capital controls and f > 0 where parameter is a sensitive parameter controlling FX intervention). chartists’ insistence on the chartist trading strategy for a given exchange rate misalignment. A higher f is accompa- nied by a lower , indicating more traders switch to funda- Fundamentalists t mentalists. Furthermore, (4) suggests that more traders turn Following Dieci and Westerhoff (2010), fundamentalists are fundamentalists when the exchange rate deviates more from usually assumed to formalize their demand as, its fundamental price. Recall that the FX market is supposed to be perfect and FF (1) DS =− β S , () free of intervention, so following the EMP definition in t tt Girton and Roper (1977), we can define exchange market where S is the indirectly quoted (log) nominal exchange t pressure in our context as, rate and represents the fundamental exchange rate. S β > 0 reflects the belief in mean reversion. EMPS =− S . (6) tt ++ 11 t It is indicated in (1) that if the exchange rate is overvalued (undervalued), fundamentalists would expect the exchange Using (6), we can obtain EMP from (3), rate to move toward its underlying fundamental price, pro- F C moting them to reduce (increase) their demand for the local EMPW =− α 1 DW + D . () () (7) tt +1 t tt currency against the foreign one. Clearly, (7) provides a theoretical basis for an intuition that EMP is jointly determined by market sentiment and cur- Chartists rency demand. For simplicity, follow Dieci and Westerhoff (2010), we define the chartists’ demand as follows, Role of Economic Uncertainty CC DS =− β S , (2) () t tt Our model suggests that investors’ demands for currency partially depend on the fundamental exchange rate, which where parameter β > 0 governs the confidence in the per- is determined by domestic and foreign macroeconomic sistence of deviations. variables in traditional theory. Since economic uncer- tainty has detrimental effects on the macroeconomy (Bloom, 2009, 2014; Jurado et al., 2015), investors’ Evolution of the Exchange Rate F C demands D , D and market sentiment would be t t As in Dieci and Westerhoff (2010, 2013), the exchange rate influenced by economic uncertainty through the funda- h a at time t + 1 is determined by the excess demand formed by mental exchange rate S . Therefore, let and denote δ δ t t t heterogeneous investors in period t. Thus, the exchange rate unexpected rising shocks in domestic and foreign eco- is developed by, nomic uncertainties, respectively, we can model S as a binary function of δ and δ , t t F C SS =+ α 1−WD +WD , () (3) () tt +1 tt tt h a Sg = δδ ,, () tt t (8) where parameter α> 0 controls the price adjustment. The gg <> 00 , proportion of chartists in period t is denoted by , while where the assumptions on the first-order partial deriva- fundamentalists occupy the portion of because we 1−W g > 0 g > 0 tive that and imply that an increase in domes- 1 2 assume only these two types of investors in the market. tic uncertainty leads to a decline in the exchange rate, while Lengnick and Wohltmann (2013) and De Grauwe and Ji a rise in foreign uncertainty causes the foreign currency to (2020) refer to these portions as market sentiment, reflecting depreciate. which trading rule prevails in the market. Motivated by De FC ββ = For simplicity, let , substituting (4) into (7) Grauwe and Ji (2020), we define market sentiment as a mea- yields, sure of the dominance of fundamentalists, EMPE =− αβ SS , EW =− 11 −= WW − 2. (4) () (9) () t+1 tt t tt tt In addition, following Dieci and Westerhoff (2010), we We hence can derive the impact of economic uncertainty define the proportion as, on EMP by taking partial derivatives based on equation (9), t 6 SAGE Open the variabilities in the exchange rate misalignments and the     share of chartists, as well as market sentiment. Moreover, as gE +− 4gW fS S () ∂EMP 11 t tt t   t+1 F analyzed by Chiarella et al. (2017), traders’ beliefs and sen- = αβ 12 4443 444 4  ∂D  ∂δ ∂E t sitivity to price misalignments, which we have assumed to be t h  ∂δ h  (10) ∂δ   constant, may also change over time, bringing additional   sources to the time variability of the impact. F 2 =+ αβ gE 4. Wf SS − () 1 tt tt () Empirical Methodology and Data Similarly, Empirical Model We work with a VAR framework to allow for potential   endogeneity generated by the self-filling mechanism and the   gE +− 4gW fS S ∂EMP ()  22 t tt t  t+1 F autocorrelations. Since a large body of evidence indicates = αβ 12 4443 444 4  ∂D  ∂δ ∂E various nonlinearities in the macroeconomic time series, it t a  ∂δ a  ∂δ (11) is quite desirable to take these nonlinearities into account     when studying the behavior of the macroeconomy. While F 2 many studies have contributed to nonlinear modeling, the =+ αβ gE 4. Wf SS − () 2 tt tt () results obtained from nonlinear models may conflict with economic theory. However, Granger (2008) demonstrates that From (10)–(11), the impact is straightforwardly decomposed any nonlinear model can be approximated by a time-varying ha , () ha , () into two components, that is, and , ∂∂ D / δ ∂∂ E / δ tt tt parameter model, which is generally linear but still has suffi- where D denotes the total currency demand. It is apparent cient power to capture nonlinearities. Consequentially, during that the sign of each derivative basically depends on the past decade, time-varying parameter vector autoregres-  2  , which can be easily transformed sions (TVP-VARs) have gained widespread popularity among EW +− 4 fS S  ()  tt tt   applied macroeconomists and have become a standard 2   into 14 −WW +− fS S −1 . Thus, if the squared framework for analyzing macroeconomic time series in the ()  ()  tt tt   last decade due to their charm of tracking processes subject deviation of the exchange rate from its fundamental value to structural breaks or regime shifts (Baumeister & Peersman, 2013). While time variations in coefficients and SS − is no less than the threshold 14 / () f , then the () tt conditional higher moments have been separately well doc- () ha , signs of ∂∂ EMP / δ only rely on the effects of uncer- tt +1 umented in the literature, a TVP-VAR model unites time g g tainty on the fundamental exchange rate, that is, and , 1 2 variations jointly and concomitantly in coefficients and respectively. Therefore, if the condition is satisfied, an variances, allowing the data to speak freely and capturing a upsurge in domestic (foreign) economic uncertainty is wide range of time variation and nonlinearity (Lubik & expected to exert depreciating (appreciating) pressure on the Matthes, 2016; Nasir et al., 2018; Nasir & Simpson, 2018; () ha , exchange rate. Further, the signs of cannot ∂∂ EMP / δ tt +1 Nasir & Vo, 2020). be clearly identified when the condition is violated that Although a TVP-VAR model is very flexible and can , mainly on account of the uncertainty SS − < 14 / f () model nearly all nonlinear relationships among the variables () tt stemming from the impact of economic uncertainty on mar- of interest, it is highly parameterized and may risk overfit- ket demand. ting, leading to inaccurate estimation of impulse response By inspecting (10) and (11), the effect of economic uncer- functions. Recently, a growing literature has developed sev- tainty on market sentiment is certain that an unexpected eral methods to mitigate these over-parameterization con- increase in domestic (foreign) uncertainty would induce a cerns, see, for example, Koop and Korobilis (2013), Eisenstat negative (positive) response in market sentiment, making et al. (2016), Huber et al. (2019), Chan et al. (2020). In this fewer (more) traders using fundamental trading rule. part of the literature, a widely used method is to use global- However, the effect on market demand essentially lies in mar- local shrinkage priors to reduce estimation bias and improve ket sentiment E . More specifically, in a market dominated model performance. However, applying shrinkage to time- by fundamentalists, a rising shock in domestic (foreign) eco- varying parameter models cannot wholly eliminate estima- nomic uncertainty would cause the relative demand for the tion errors (Huber et al., 2021). Moreover, incorporating domestic currency to decline (increase). Otherwise, the shrinkage is less straightforward, and it often requires com- demand would increase (reduce) in accordance with domestic putationally demanding algorithms or approximate inference (foreign) economic uncertainty when chartists prevail. (Eisenstat et al., 2016). Eisenstat et al. (2016) propose a new Remarkably, as shown in (10) and (11), the impact of eco- approach that combines the stochastic model specification nomic uncertainty on EMP would be time-dependent due to search (SMSS) framework developed in Frühwirth-Schnatter Liu 7 and Wagner (2010) with a typical TVP-VAR to ensure model α α= αα ,, … Let denotes the initial states () 1 mmp+1 () parsimony. Comparatively, the methodology in Eisenstat γβ =− αω / of ββ and reparametrize () , for jt jt jj et al. (2016) is more flexible and efficient in that it allows the , so that (1)–(2) can be written as, jm =… 11 ,, () mp + model to automatically and endogenously choose between a time-varying parameter against a constant parameter for (4)  2 yX =+ α αγ XV γ + tt tt t each VAR coefficient. (5) γγγ =+ γη η tt−1 t A typical TVP-VAR model. The generic state-space form of a typical structural TVP-VAR with stochastic volatility can where η η ~,  0 I and  is independent of each () t t mmp+1 () be written as, other for all leads and lags, whereas . Finally, α αα ~,  () α A (4)–(5) constitute the conventional TVP-VAR model with By =+ X ββ  , (1) tt tt t stochastic volatility. βββ =+ βη η , (2) tt−1 t The SMSS framework. To alleviate the over-parameter- ization worries around ββ , Eisenstat et al. (2016) propose a where y is an m-dimensional vector of endogenous vari-  2 ables of interest at time , when the regressor tT =… 1, , Tobit prior on the diagonal elements of , that is, . For V ωω ′ ′ matrix Xy =⊗ I 1,,…, y consists of an intercept () tt tt −− 1 p jn =… 1, , , introduce a latent variable and assume to follow and p lags of y . is a lower unitriangular matrix contain- a normal distribution, ing contemporaneous relations. ββ is an vector mmp +1 () * 2 ωμ ~,  τ , of stacked coefficients at time t and is assumed to evolve as () jj j a random walk process with the errors , where ηη is ηη t t and relate to ω by the following indicator function, Gaussian innovations with zero mean and a diagonal covari- j j ance matrix V . In the matrix V, each element 00 , if ω ≤  j vj =… 11 ,, mmp + is the variance regarding coefficient () () j ω = ** (6) β , determining the degree of its time variability. ωω , if > 0  jj jt The residual follows an independent Normal distribu- In addition, in order to incorporate the Lasso structure, it tion with zero mean and a variance-covariance matrix ΣΣ , is assumed that τ follows an Exponential distribution, which is assumed to be time-varying and diagonal with ele- 2 2   ments σ = exp h , that is, λ () it it 2 τ ~   , where λλ ~,  λ , denotes the j ()  01 02      exp h L 0  () 1t Gamma distribution.   Σ Σ = MO M , It is clear to see that the Tobit prior automatically restricts t     as per ω , while still allowing for a straightforward 0 L exp h () ω ≥ 0 it j   Gibbs sampler, which is fast and suitable for implementing where h =… hh ,, is defined as the log-volatility () hierarchical shrinkage through the hyper-parameters of , tt 1 it ω evolving in a random walk fashion, that is, m and τ . hh =+ wR ,~ w  0, () (3) tt −1 tt Model estimation. Following Eisenstat et al. (2016), the MCMC Gibbs sampler is adopted for estimation and com- This setup is motivated by the well-documented evidence puting the posterior distributions for the parameters and of allowing variance-covariance of residuals to vary over hyper-parameters. Generally, the MCMC Gibbs sampler is time in modeling macroeconomic data, see Fernández- distinctively straightforward and efficient since it samples Villaverde et al. (2015), for instance. Following Eisenstat from the full conditional probability distribution. Crucially, et al. (2016), we assume that h is initialized with the MCMC is a smoothing method that produces smoothed hV ~,  0 and the prior on the transition covariance is () 00 estimates (Nasir & Morgan, 2018). Most importantly, the Rv ~, W R , where denotes the inverse Wishart () W MCMC Gibbs sampler is suitable for our case. As discussed distribution. earlier, introducing the Tobit prior to shrink model dimen- As the matrix is lower unitriangular, we can rewrite sionality would lead to a computationally feasible and fast (1)–(2) by rearranging X to include contemporaneous y Gibbs sampler. In addition, the posterior computations using t t and rearranging to include the free elements in B . Then, ββ MCMC are outlined to deliver an overview of the sampling t t the covariance matrix is no longer a diagonal but a full V τ procedure. First, for in =… 1, , , , , and are stacked ω ω i i 1 1 as vectors , ωω , and . Then, posterior draws are obtained ωω ττ  2 2 matrix that can be decomposed as , where VV = Φ ΦΦ Φ′V ΦΦ γγ by sequentially sampling by the following order: αα , , , 1 ΣΣ  2 is a lower unitriangular matrix, and , , , and . For details on the MCMC Gibbs sampler, the V =… diag ωω ,, ωω ττ λ () 1 n . reader is referred to Eisenstat et al. (2016). nm =+ () mp 1 8 SAGE Open Prior settings. The priors are set in line with Eisenstat non-transparent measures rather than intervening directly, et al. (2016), as all sample data will be standardized before especially after 2015. The central parity rate has been a usual estimation, making the priors standard to some extent. As tool since the People’s Bank of China (PBoC), China’s cen- shown above, the priors for the hyper-parameters, and tral bank, first announced it on August 11, 2015. Since the ΣΣ R are assumed to follow independent inverse-Wishart distri- reform on July 21, 2005, the PBoC has been operating a sys- butions, while ττ and are set to be distributed as Expo- tem that allows the daily exchange rate to fluctuate within a nential and Gamma distributions. In addition, the priors for narrow band, which was expanded from the initial 0.3% to parameters and are assumed to be normally distributed. 2% in March 2014, around a target called the central parity αα ω Further, we set the following hyper-parameters on these pri- rate. This central parity rate was based on the closing price ors to end the SMSS specification of the TVP-VAR model, on the last trading day before August 11, 2015, and thereaf- ter, is based on the last trading day’s closing price plus the needed changes, which refer to the needed adjustment of the α α == 00 ,, AI ,, hV == I vm =+11, 00 00 m 0 mm () p+1 RMB/USD exchange rate to offset the overall impact of the RI =− 00 . 11 vm − , μ = 00 0,. λλ == 1. () 0 0 m 01 02 fluctuations of the currencies in the currency basket against the dollar on the last trading day and overnight. In addition, a counter-cyclical adjustment factor (hereafter CCAF), initi- Finally, the above empirical model and framework will be ated on May 26, 2017, to mitigate irrational market senti- applied to a Chinese monthly dataset to exploring the roles of ment, has been added to the regime since then. The PBoC has economic uncertainty in China’s exchange market pressure. occasionally intervened in the market through the CCAF afterward. In this sense, the CCAF is viewed as an instru- Data ment for FX intervention (Das, 2019). Therefore, it is neces- sary to include this CCAF in the estimation of China’s EMP. We conduct our empirical investigation with a monthly data- Further, the PBoC had directly capitalized four big state- set from January 2001 to October 2020, primarily based on controlled banks with FX reserves between 2003 and 2008. data availability. Additionally, a principal consideration in Ignoring these expenditures for non-intervention purposes our decision to start our sample from January 2001 is that would bias the EMP estimates. Hence, we differ from previ- China’s macroeconomic data prior to 2001, during which ous studies in that we take these non-intervention purpose China underwent profound institutional and economic transi- expenditures back and accordingly recover the reserves. tions, were less reflective of the transmission dynamics Thus, the formula for calculating China’s EMP is defined among economic variables (Fernald et al., 2014). as follows, Our baseline model centers on five main variables indi- cated by our behavioral model presented in Section 3, that is, SS − RR − EMP, domestic and foreign economic uncertainties, market tt−1 tt−1 − ,, before August 2015 sentiment, and market demand in the FX market. S R t−1 t−1 EMP = , SS − − CCA AF () RR − tt−1 t  tt−1 Estimates of China’s EMP. Girton and Roper (1977) theo- − ,afterwards S R  t−1 t−1 rize that EMP is the sum of exchange rate fluctuations and official intervention. For the former, as discussed earlier, we where R represents the (recovered) FX reserves. Note that focus on the onshore RMB exchange rate against the dol- here is the onshore RMB/USD exchange rate expressed in lar since this trading pair dominates China’s FX market. the direct quotation. For the latter, the growth of FX reserves has been primar- However, since the PBoC does not make public the CCAF, ily considered in the literature, as the government usually we need to estimate it first. To this end, based on the forma- intervenes through transactions on FX reserves. In addition, tion regime of the central parity rate, we regress the differen- underpinned by the assumption that the central banks can tial between the central parity rate and the last closing price of intervene by adjusting short-term interest rates, Klaassen and the RMB/USD exchange rate (denoted by ) on the one- CR dt Jager (2011) include interest rate differentials in calculating trading-day lagged composite growth of the currencies in the EMP. However, we do not include interest rate differentials currency basket against the dollar (denoted by ) using AC dt−1 because China rarely utilizes interest rate instruments to a daily dataset from August 11, 2015, to October 30, 2020, intervene in the FX market (Das, 2019; Li et al., 2017). yielding the following regression equation, Moreover, the intervention might not be fully reflected in changes in China’s FX reserves, especially after the com- CR =+ ββ AC + λ , dt 01 dt−1 dt plete abolition of the mandatory FX settlement regime in April 2012. As documented in Li et al. (2017) and Das where a constant term β is included. We average the (2019), and more recently in The Economist (2020), China growth of the last closing price of the currencies against has many tools to intervene in the FX market and prefers to the U.S. dollar with the currencies’ weights. The residual do so by adjusting the central parity rate and other λ captures the unexplained component of . Based CR dt dt Liu 9 Figure 1. Estimated EMP and CCAF. EMP: exchange market pressure, CCAF: counter cyclical adjustment factor, USD base: the U.S. dollar index base, CFETS base: the Chinese currency basket designed by China Foreign Exchange Trade System. on the central parity rate formation regime, we can regard Compared to our EMP index, there are many outliers in λ as the CCAF, which we denote as a CFETS-based the Patnaik et al. (2017)’s measure. Moreover, our EMP dt CCAF. index is less volatile and generally smaller than Patnaik et al. Notably, to prepare AC , we calculate each currency’s (2017)’s. In the following section, we use the CFETS cur- dt−1 daily growth in the basket against the dollar and then average rency basket-based EMP to fulfill our empirical study. the growth by the corresponding weight of each currency. Finally, the average growth is multiplied by the weight of the Proxy for economic uncertainty. As stated in the introduc- dollar. In addition, the weights of the currencies in the basket tion section, we use the news-based EPU index to proxy are adjusted according to the PBoC. economic uncertainty. Motivated by Christou et al. (2018), Additionally, to provide some insights into the reason- Bartsch (2019), and Zhou et al. (2020), U.S. EPU is also ableness of our calculation of the composite growth, we also included. Accordingly, since we focus on the RMB/USD regress on the 1-day lagged growth of the U.S. dollar exchange rate pressure, we use the U.S. EPU constructed CR dt index to obtain a U.S. dollar index-based CCAF. by Baker et al. (2016) and China’s EPU computed by Davis The closing exchange rates are collected from Investing. et al. (2019), downloaded from policyuncertainty.com. com, while the central parity rate and the U.S. dollar index are retrieved from the WIND database. Proxies for market sentiment and market demand. Fol- We estimate the equation using the OLS method and sum lowing Das (2019), we proxy market sentiment with the the daily CCAFs to produce the monthly CCAFs. The result- one-year-ahead forward premium on the exchange rate. ing CCAFs are shown in Figure 1. As we have the CCAF, we Specifically, we compute the log-difference (multiplied compute the monthly EMP index from January 2001 to by 100) between the one-year-ahead non-deliverable for- October 2020, with the (recovered) FX reserves retrieved wards on the RMB/USD exchange rate and its onshore spot from the WIND database. Our calculated EMP indices are counterpart, yielding a market sentiment index that reveals also reported in Figure 1. Moreover, the EMP index com- an expectation of devaluation (appreciation) when it has a puted by Patnaik et al. (2017), downloaded from https://mac- positive (negative) value. rofinance.nipfp.org.in/releases/exchange_market_pressure. To approximate changes in currency demand, as investors html, is also presented for comparison. would trade according to their expectations (Cornell & As shown in Figure 1, the estimated CCAF and the EMP Dietrich, 1978), mainly through banks in China (Lin & index based on the CFETS currency basket are highly con- Schramm, 2003), we use customers’ net sell of foreign cur- sistent with those based on the U.S. dollar index in terms of rency against the RMB calculated by the log-difference magnitude and pattern, where the correlation between the between customers’ sell and purchase of FX through the com- two CCAFs exceeds 0.9. Centering on the estimated CCAF mercial banks. We believe that the overall trading volume yields an intuitively consistent result that the PBoC acts could disclose the changes in market demand of the RMB counter-cyclically when encountering significant exchange against the dollar since trading on this pair accounts for more market pressure. than 95% of the overall trading volume on average. 10 SAGE Open Table 1. Unit Root Tests. ADF without breaks ADF with breaks Test statistics Lag order Test statistics Allowed breaks EMP −9.476** 1 −7.158*** 1 U.S. EPU −6.141*** 0 −8.799*** 1 China’s EPU −3.036 3 −9.055*** 4 Market sentiment −3.103 0 −4.38** 3 Market demand −3.855** 1 −3.969*** 1 Note. *** and ** denote rejection of the null hypothesis at the 1% and 5% significance levels, respectively. Intercept and trend are included. The optimal lag is determined by the BIC. are stationary regardless of whether structural breaks are considered. While the conventional ADF test detects that China’s EPU and market sentiment are non-stationary, the ADF test incorporating breaks signifies a high significance of stationarity of these two variables. Thus, all variables can be viewed as statistically stationary in a general sense. Moreover, this finding highlights the importance of applying the TVP-VAR model that incorporates structural breaks and regime shifts. We set a lag length of six to ensure no serial autocorrela- tion and estimate the model by the procedure documented in Eisenstat et al. (2016). The MCMC Gibbs sampler in Eisenstat et al. (2016) is applied to attain 45,000 replicates with the first 15,000 draws as burn-in, obtaining 3,000 effec- tive posterior draws by recording every ten replicates. Our estimation incorporates three potentially beneficial Generalized Gibbs steps, which can produce the lowest inef- ficiency factors plotted in Figure 2. Clearly, most of the inef- Figure 2. Inefficiency factors of the estimated parameters. ficiency factors are less than 10, though the majority for is slightly greater than 10, indicating the good performance of Finally, all the data, except for the EPU measures, are our MCMC sampler. retrieved from the WIND database and seasonally adjusted (if Next, based on the effective MCMC draws, we estimate necessary), while the EPU measures were taken logarithm. the impulse response functions using the Cholesky decom- position to identify structural shocks. Notably, all impulse Results response functions (IRFs) are rescaled to match the original series. To evaluate the total impact of EPU, we compute Baseline Model cumulative impulse response functions at each time point of First of all, we perform unit root tests to elucidate some sta- 1% shocks in U.S. EPU and China’s EPU, respectively. tistical properties of our variables. Following Jebabli et al. We perform a preliminary analysis with the time-averaged (2014), we employ the conventional ADF unit root testing IRFs for a horizon of 12 months to provide an overall picture approach to validate the stationarity of our variables. of the EPU’s impact, see Figure 3: The 68% highest posterior However, the conventional ADF test may be underpowered, density (HPD) intervals (shaded areas) are constructed with given that the macroeconomic indicators may have multiple 16% and 84% percentiles of the posterior estimates, while potential structural breaks (Check & Piger, 2021). Therefore, the estimated IRFs (solid lines) are the posterior medians. As similar to Nasir (2021), we further apply the GLS-based expected, a rising shock in U.S. EPU would generally ADF test proposed by Carrion-i-Silvestre et al. (2009), which increase the relative demand for the renminbi against the dol- allows for multiple structural breaks in both the null and lar, exerting appreciation pressure on the RMB. However, alternative hypotheses. The corresponding results of unit this market demand channel does not hold for the shocks root testing are reported in Table 1. originating from China’s EPU that it has a positive but impre- As shown in Table 1, allowing for structural breaks in the cise effect on market demand. The counter-intuitive results variables improves the significance of the ADF test. It is sug- are also presented in the response of market sentiment. While gested that EMP, U.S. EPU, and market demand it exhibits a slightly negative response following a hike in Liu 11 Figure 3. Time-averaged dynamics of market sentiment, market demand, and EMP, following a 1% increase in U.S. EPU and China’s EPU, respectively. U.S. EPU, market sentiment responds with a more signifi- time-varying market responses at these horizons after a 1% cant negative pattern after an upsurge in China’s EPU, signi- shock in U.S. EPU and China’s EPU, respectively. fying that a higher China’s EPU may be associated with the At first glance, the IRFs in Figures 3 and 4 provide strong expectations of RMB appreciation and increased demand for evidence of the time-varying impact of EPU on China’s FX the RMB. Notably, this finding is likely relevant to our proxy market, particularly on EMP. Relatively, the IRFs are more for market sentiment, namely the spread between the one- volatile at longer horizons. One possible explanation is that year-ahead NDF exchange rate and the onshore spot since China’s FX market may not be fully efficient (Gupta & exchange rate. As the NDF exchange rate is the expected Plakandaras, 2019), the market may need time to recognize exchange rate 1 year later, if the uncertainty shocks cause the the shock, waiting to see the impact it will cause and respond- spot exchange rate to depreciate in the short run, the market ing accordingly afterward. Consequentially, the IRFs at the would expect the exchange rate to revert, inducing apprecia- horizons of 0-month and 1-month are more stable and less tion expectations. In addition, China’s asymmetric FX regu- precise, indicating market inefficiency. However, the IRFs at lations, where selling FX is less controlled than buying, may longer horizons generally depend on market conditions, such also contribute to the preposterous response in market as institutional and market regimes, which have dramatically demand. Nevertheless, an increase in China’s EPU is still changed in China over the past decades, resulting in the high followed by an approximate positive response in EMP, cor- volatility of these IRFs. Therefore, we scrutinize and discuss roborating roughly the findings in Liu (2020). the rationale of the time variability revealed by these IRFs in To dissect how the impact of EPU evolves over our sam- the following section. ple period, we turn to an analysis based on the time-varying As shown in the top panels in Figures 4 and 5, an increase IRFs. Since the FX market operates daily, it is unlikely that in either U.S. EPU or China’s EPU would trigger apprecia- the market consumes a long time to respond to an external tion sentiment on the renminbi against the dollar, which is shock. Accordingly, we accentuate our analysis on inspect- consistent with the findings by Li et al. (2020). Moreover, ing the market’s reactions in 3 months after the shock. the response of market sentiment to a U.S. EPU shock is con- Further, we detect the responses at 6-month after the shock as sistent with the common wisdom that a rise in U.S. EPU has the time-averaged IRFs shown in Figure 3 almost all con- a detrimental impact on the U.S. macroeconomy and pro- verge after then. More specifically, we look into the impulse vokes devaluation expectations on the dollar. However, the response functions at each time point at horizons of 0 to response to the China’s EPU shocks contradicts the uncer- 3 months and 6 months after the shock, where 0-month means tainty theory. As noted earlier, this could be explained by the moment when the shock occurs. Figures 4 and 5 paint the China’s FX regulations, which might cause the market to 12 SAGE Open Figure 4. Time-varying market responses at horizons of 0-month to 3-month, and 6 months following a 1% increase in U.S. EPU. The solid lines are posterior medians, while the shaded regions are the corresponding 68% HPD intervals. Figure 5. Time-varying market responses at horizons of 0 to 3, and 6 months following a 1% increase in China’s EPU. The solid lines are posterior medians, while the shaded regions are the corresponding 68% HPD intervals. Liu 13 expect the renminbi to appreciate against the dollar. Moreover, the IRFs at horizons of 2, 3, and 6 months after the Relatively, the response to China’s EPU shocks is more sta- shock exhibit noticeable time variations. Specifically, the ble than that to U.S. EPU shocks, though the latter’s corre- impact of EPU on EMP is stable until mid-2011 but then sponding HPD intervals are wide to contain zero, indicating enlarges till the end of 2017. Accordingly, our results provide insignificant responses of market sentiment to the U.S. EPU little evidence to support the findings by Kido (2016), who shocks. By contrast, following a hike in China’s EPU, there finds that the real effective exchange rate returns and U.S. would be a significant appreciation expectation on the ren- EPU were intensively correlated during the global financial minbi against the dollar from 3 months onward, especially in crisis. The explanation is straightforward. Albeit, indeed, the the post-2005 period. Before 2005, China carried out a de unconventional monetary policy undertaken by the U.S. facto pegged exchange rate regime, causing market senti- Federal Reserve during the financial crisis reduced the dol- ment to be less responsive to China’s EPU shocks. However, lar’s value (Neely, 2015), China had pegged the renminbi to following the reform in July 2005, China implemented a the dollar again during this period. Further, China had experi- gradual appreciation path for the renminbi against the dollar, enced large capital outflows during the crisis (Broner et al., leading to the widespread market expectations of the RMB 2013) due to the need to relieve the value-at-risk of U.S. appreciation. In addition, the exchange rate stability has been domestic assets caused by the crisis (Schmidt & Zwick, one of China’s principal objectives. Thus, the negative build- 2015). Thus, the impact of U.S. EPU on China’s EMP could ing impact of China’s EPU on the renminbi would stimulate retain its past pattern without being altered by the crisis. an expectation that the government will intervene in the mar- Intuitively, this time-varying property shown in the impact ket, inducing the appreciation expectations on the renminbi. of U.S. EPU on China’s EMP may be associated with the Further, the response of market sentiment to the U.S. EPU evolution of China’s dependence on the United States. Over shocks swells progressively after mid-2015, which could be the last decades, China has developed a high degree of trade ascribed to the increased exchange rate flexibility after intro- and financial linkages with the United States. Taking the for- ducing the quoting regime of the central parity rate in August eign trade as an example, Figure 6 displays the geographical 2015 (Das, 2019). Gini coefficient of China’s international trade and the share It is noteworthy that market demand expresses a signifi- of trade with the United States in China’s total trade from cant response following a rise in U.S. EPU, as shown in the 1992 to 2019. Generally, the dependence on the U.S. is high, middle panel in Figure 4. In the first month after the shock, especially for exports, although the geographical concentra- the relative demand for the renminbi increases significantly, tion of China’s trade is relatively low. The overall depen- mirroring the dollar devaluation expectations caused by the dence on the U.S. has declined gradually, but slightly, from U.S. EPU shocks. By contrast, market demand responds 2001 to 2011, resulting in a relatively small and stable impact counter-intuitively and insignificantly to China’s EPU of U.S. EPU during this period. Subsequently, from 2012 to shocks. As displayed in Figure 5, after a rising shock in 2017, China’s dependence on the U.S. has been fueled up, China’s EPU, we can see an increase in market demand for associated with the strengthening impact of U.S. EPU on the renminbi in the first month, but it shows a preference for China’s EMP. However, since 2018, the trade dependence the dollar 2 months after the shock. Therefore, it may take has plummeted because of the U.S.-China trade dispute, about 2 months for China’s EPU shocks to have the supposed making China’s EMP less responsive to the U.S. EPU shocks. impact on market demand. Notwithstanding, China’s EPU In addition, the recently heightened response of EMP to the exerts a less precise effect on market demand when com- U.S. EPU shocks may be attributed to the easing of the trade pared with U.S. EPU. One explanation for this is that China’s dispute and the outbreak of COVID-19. FX regulations make it much easier to sell foreign currencies As shown in Figures 4 and 5, it is discernible that the than to buy them. Importantly, as shown in Figure 5, the responses of China’s EMP to the shocks originating from response also has a downward trend after mid-2015, proba- domestic EPU and U.S. EPU are similar in dynamics but bly due to an increasingly flexible exchange rate regime. opposite in evolutionary path. Since EMP could be influ- This result supports the findings by Kozhan and Salmon enced by local and external uncertainties concurrently, we (2009) that agents in the foreign exchange market are uncer- find a trade-off mechanism between them, where the impact tainty averse. of domestic EPU prevails when the impact of foreign EPU While the responses of market sentiment and market fades and vice visa, corroborating the effect of the EPU dif- demand to China’s EPU shocks are somewhat at odds with ferentials on the exchange rate reported in Balcilar et al. the theory, the response of EMP is consistent with our expec- (2016), Christou et al. (2018), and Zhou et al. (2020). tations. As shown in the last panels in Figures 4 and 5, respec- Further, the time-varying nature of EMP response to the tively, a hike in U.S. EPU (China’s EPU) would systematically EPU shocks may also be related to the transmission through prompt appreciating (deprecating) pressure on the renminbi market sentiment and demand. To explore the roles of market against the dollar, leading to a negative (positive) response of sentiment and demand, we estimate the time-varying IRFs of EMP. Comparatively, China’s EPU shocks take a longer time EMP to the shocks in these two variables and plot the results than U.S. EPU shocks to generate an impact on China’s EMP. in Figure 7. 14 SAGE Open Figure 6. Gini coefficients of China’s trade and trade dependence on the United States. Following Liu et al. (2020), we measure the geographical concentration of trade with the Gini coefficients, calculated by , where X is the trade with country i in GX = / X it ti () tt denotes aggregate trade in year t. The trade data are retrieved from the UN COMTRADE database: (a) Gini coefficients year t, and X and (b) fractions of US in Chinese foreign trade. Figure 7. Time-varying responses of EMP to 1% shocks in market sentiment and market demand, respectively. The solid lines are posterior medians, while the shaded regions denote the corresponding 68% HPD intervals. Obviously, more time variation is detected in EMP would trigger appreciation pressure on the RMB. responses to the market sentiment and demand shocks. An Additionally, after August 2015, increasingly enhanced increase in market sentiment (i.e., devaluation expectations exchange rate flexibility raises investors’ risk exposure, on the renminbi against the dollar) would substantially causing the impact of market sentiment and market demand coerce the RMB to devaluate, while a rise in market demand to reduce sharply. Liu 15 Figure 8. Time-averaged responses of EMP to 1% shocks to U.S. EPU and China’s EPU. The solid lines are posterior medians, while the shaded regions are the 68% HPD intervals. As we can see in Figures 8 and 9, Rmodel1 and Rmodel2 Robustness Check produce similar results to our baseline model. However, We consider the following two alternative specifications to Rmodel1 derives a more precise estimate of the EMP implement robustness checks. (1) [Rmodel1] The EPU mea- response to China’s EPU shocks, particularly after 2010, sure we used for China is constructed based on information during which an upsurge in China’s EPU would cause sig- from only two newspapers that may be subject to media bias, nificant devaluation pressure on the onshore exchange rate. so we instead use the EPU measure for China constructed by This result may suggest that the EPU index constructed by Huang and Luk (2020), which is based on 10 newspapers Huang and Luk (2020) is a more credible indicator of eco- pressed in mainland China. (2) [Rmodel2] Since the funda- nomic uncertainty as it is less prone to media bias (Huang & mental price of the exchange rate is theoretically determined Luk, 2020). In addition, Rmodel2 renders a more remarkable by the macroeconomy, we include China’s leading macro- but more stable response of EMP to U.S. EPU shocks. economic indicators in the model, namely, economic growth (measured by real GDP growth), inflation (proxied by CPI Conclusion and Policy Implications growth), short-term interest rate (using the 7-day reverse repo rate) and money growth (measured with M2 growth). Our empirical findings show that an upsurge in U.S. eco- All the data are taken from Chang et al. (2016)’s China’s nomic policy uncertainty triggers appreciation pressures on Macroeconomy Database. All these variables are shown as the RMB/USD exchange rate, while a hike in China’s eco- stationary by the ADF unit root tests with and without struc- nomic policy uncertainty tends to exert devaluation pressures tural breaks. The estimation procedure for both models is the on the exchange rate. Moreover, the magnitude of the impact same as that for the baseline model. Here, we concentrate on of economic policy uncertainty on exchange market pressure the impact of unexpected rises in U.S. EPU and China’s EPU is time-dependent. Specifically, the impact is relatively stable on China’s EMP. Figures 8 and 9 show the time-averaged until mid-2011 but more volatile thereafter. In addition, we IRFs and the time-varying IRFs predicted by the two models, find a trade-off mechanism between the impact on EMP of respectively. domestic and foreign economic policy uncertainties. 16 SAGE Open Figure 9. Time-varying responses of EMP to 1% shocks in China’s EPU and U.S. EPU, respectively. The solid lines are posterior medians, while the shaded regions denote the 68% HPD intervals. Our results have the following policy implications. First, However, our study still has some limitations. First, to stabilize the exchange rate, China should enhance its man- because of the lack of statistics on market agents, our proxy agement of market expectations and market demand, reduc- for market sentiment in the empirical section is not entirely ing the unfavorable impact on the exchange rate of irrational consistent with that illustrated in the theoretical model, market sentiment to cope with spikes in domestic and foreign resulting in a rough understanding of market sentiment in a economic uncertainties. In addition, a foreign exchange broad sense. Second, our proxy for market demand is not a derivatives market should be established and advanced to pure index reflecting demand for the RMB/USD pair but also provide functional risk hedging tools and attain a better includes other trading pairs, despite the dominance of the understanding of market sentiment. Second, further liberal- RMB/USD pair in the market. Third, since China still adopts ization of the capital account should be proceeded prudently a managed floating exchange rate regime, we note that and sequenced with prerequisite reforms, including a solid China’s monetary authority releases scant interventions in and sound financial system and an appropriate macroeco- the foreign exchange market and intervenes the market by nomic regulatory framework. Third, China should further other hidden levers recently keeping the reserves unmoved, improve the flexibility of the RMB exchange rate and make limiting us to obtain a highly accurate estimate of China’s the exchange rate formation mechanism transparent, sending exchange market pressure. clear signals of two-way exchange rate fluctuations to avoid As for the future research, since capital flows are sensitive irrational market sentiment toward one-way movements. to economic uncertainty and exchange rate fluctuations, one Finally, China should further diversify the basket currency possible direction is to dissect the roles of capital flows in the structure and increase RMB transactions against non-USD relationships between economic uncertainty and exchange currencies. Furthermore, China could lower its dependence market pressure. Moreover, in recent years, China’s authority on the United States in trade and finance to a certain extent to has attempted to manage the foreign exchange market by reduce the detrimental impact on China’s real economy and adjusting the foreign exchange reserve requirement ratio, open- financial markets of economic uncertainty shocks originat- ing a topic that how and whether this policy affects market sen- ing from the United States. timent and market demand, and thereby the exchange rates. Liu 17 Declaration of Conflicting Interests Chang, C., Chen, K., Waggoner, D. F., & Zha, T. (2016). Trends and cycles in China’s macroeconomy. NBER Macroeconomics The author declared no potential conflicts of interest with respect to Annual, 30(1), 1–84. https://doi.org/10.1086/685949 the research, authorship, and/or publication of this article. Chan, J. C. C., Eisenstat, E., & Strachan, R. W. (2020). Reducing the state space dimension in a large TVP-VAR. Journal of Funding Econometrics, 218(1), 105–118. https://doi.org/10.1016/j.jeco- The author received no financial support for the research, author- nom.2019.11.006 ship, and/or publication of this article. Check, A., & Piger, J. (2021). Structural breaks in U.S. Macroeconomic time series: A Bayesian model averaging approach. Journal of Money Credit and Banking, forthcoming ORCID iD 53, 1999–2036. https://doi.org/10.1111/jmcb.12822 Lin Liu https://orcid.org/0000-0003-1258-1730 Chen, L., Du, Z., & Hu, Z. (2020). Impact of economic policy uncer- tainty on exchange rate volatility of China. Finance Research References Letters, 32, 101266. https://doi.org/10.1016/j.frl.2019.08.014 Aizenman, J., & Binici, M. (2016). Exchange market pressure in Chiarella, C., He, X.-Z., Shi, L., & Wei, L. (2017). A behavioural OECD and emerging economies: Domestic vs. External fac- model of investor sentiment in limit order markets. Quantitative tors and capital flows in the old and new normal. Journal of Finance, 17(1), 71–86. https://doi.org/10.1080/14697688.2016 International Money and Finance, 66, 65–87. https://doi. .1184756 org/10.1016/j.jimonfin.2015.12.008 Choi, M. J., Kim, G.-Y., & Lee, J. Y. (2018). An analysis of trade Bachmann, R., Elstner, S., & Sims, E. R. (2013). Uncertainty patterns and the effects of the real exchange rate movements and economic activity: Evidence from business survey data. in East Asia. Emerging Markets Finance and Trade, 54(7), American Economic Journal Macroeconomics, 5(2), 217–249. 1635–1652. https://doi.org/10.1080/1540496x.2017.1316712 https://doi.org/10.1257/mac.5.2.217 Christou, C., Gupta, R., Hassapis, C., & Suleman, T. (2018). The Baker, S. R., Bloom, N., & Davis, S. J. (2016). Measuring eco- role of economic uncertainty in forecasting exchange rate nomic policy uncertainty. The Quarterly Journal of Economics, returns and realized volatility: Evidence from quantile pre- 131(4), 1593–1636. https://doi.org/10.1093/qje/qjw024 dictive regressions. Journal of Forecasting, 37(7), 705–719. Balcilar, M., Gupta, R., Kyei, C., & Wohar, M. E. (2016). Does https://doi.org/10.1002/for.2539 economic policy uncertainty predict exchange rate returns and Cornell, W. B., & Dietrich, J. K. (1978). The efficiency of the mar- volatility? Evidence from a nonparametric causality-in-quan- ket for foreign exchange under floating exchange rates. The tiles test. Open Economies Review, 27(2), 229–250. https://doi. Review of Economics and Statistics, 60(1), 111. https://doi. org/10.1007/s11079-016-9388-x org/10.2307/1924339 Bartsch, Z. (2019). Economic policy uncertainty and dollar-pound Das, M. S. (2019). China’s evolving exchange rate regime. IMF exchange rate return volatility. Journal of International Working Paper No.19/50. https://www.imf.org/-/media/Files/ Money and Finance, 98, 17. https://doi.org/10.1016/j.jimon- Publications/WP/2019/WPIEA2019050.ashx fin.2019.102067 Davis, S. J., Liu, D., & Sheng, X. S. (2019). Economic policy Baumeister, C., & Peersman, G. (2013). Time-varying effects of uncertainty in china since 1946: The view from mainland oil supply shocks on the US economy. American Economic newspapers. Working Paper. https://stevenjdavis.com/s/EPU- Journal Macroeconomics, 5(4), 1–28. https://doi.org/10.1257/ in-China-View-from-Mainland-Newspapers-August-2019. mac.5.4.1 pdf Beckmann, J., & Czudaj, R. (2017a). Exchange rate expecta- De Grauwe, P., & Ji, Y. (2020). Structural reforms, animal spir- tions and economic policy uncertainty. European Journal of its, and monetary policies. European Economic Review, 124, Political Economy, 47, 148–162. https://doi.org/10.1016/j. 103395. https://doi.org/10.1016/j.euroecorev.2020.103395 ejpoleco.2016.06.003 de Jong, E., Verschoor, W. F. C., & Zwinkels, R. C. J. (2010). Beckmann, J., & Czudaj, R. (2017b). The impact of uncertainty on Heterogeneity of agents and exchange rate dynamics: professional exchange rate forecasts. Journal of International Evidence from the EMs. Journal of International Money and Money and Finance, 73, 296–316. https://doi.org/10.1016/j. Finance, 29(8), 1652–1669. https://doi.org/10.1016/j.jimon- jimonfin.2017.02.009 fin.2010.05.007 Bloom, N. (2009). The impact of uncertainty shocks. Econometrica, Dieci, R., & Westerhoff, F. (2010). Heterogeneous speculators, 77(3), 623–685. https://doi.org/10.3982/Ecta6248 endogenous fluctuations and interacting markets: A model of Bloom, N. (2014). Fluctuations in uncertainty. Journal of Economic stock prices and exchange rates. Journal of Economic Dynamics Perspectives, 28(2), 153–176. https://doi.org/10.1257/jep. and Control, 34(4), 743–764. https://doi.org/10.1016/j. 28.2.153 jedc.2009.11.002 Broner, F., Didier, T., Erce, A., & Schmukler, S. L. (2013). Gross cap- Dieci, R., & Westerhoff, F. (2013). On the inherent instability of ital flows: Dynamics and crises. Journal of Monetary Economics, international financial markets: Natural nonlinear interac- 60(1), 113–133. https://doi.org/10.1016/j.jmoneco.2012.12.004 tions between stock and foreign exchange markets. Applied Carrion-i-Silvestre, J. L., Kim, D., & Perron, P. (2009). GLS-based Mathematics and Computation, 221, 306–328. https://doi. unit root tests with multiple structural breaks under both the org/10.1016/j.amc.2013.06.042 null and the alternative hypotheses. Economic Theory, 25(6), Eisenstat, E., Chan, J. C. C., & Strachan, R. W. (2016). Stochastic 1754–1792. https://doi.org/10.1017/s0266466609990326 model specification search for time-varying parameter VARs. 18 SAGE Open Econometric Reviews, 35(8–10), 1638–1665. https://doi.org/10 stochastic volatility. Energy Economics, 45, 66–98. https:// .1080/07474938.2015.1092808 doi.org/10.1016/j.eneco.2014.06.008 Fernald, J. G., Spiegel, M. M., & Swanson, E. T. (2014). Monetary Jurado, K., Ludvigson, S. C., & Ng, S. (2015). Measuring uncer- policy effectiveness in China: Evidence from a FAVAR model. tainty. American Economic Review, 105(3), 1177–1216. Journal of International Money and Finance, 49, 83–103. https://doi.org/10.1257/aer.20131193 https://doi.org/10.1016/j.jimonfin.2014.05.007 Kido, Y. (2016). On the link between the US economic policy Fernández-Villaverde, J., Guerrón-Quintana, P., & Rubio-Ramírez, uncertainty and exchange rates. Economics Letters, 144, 49– J. F. (2015). Estimating dynamic equilibrium models with sto- 52. https://doi.org/10.1016/j.econlet.2016.04.022 chastic volatility. Journal of Econometrics, 185(1), 216–229. Klaassen, F., & Jager, H. (2011). Definition-consistent measure- https://doi.org/10.1016/j.jeconom.2014.08.010 ment of exchange market pressure. Journal of International Frühwirth-Schnatter, S., & Wagner, H. (2010). Stochastic model Money and Finance, 30(1), 74–95. https://doi.org/10.1016/j. specification search for Gaussian and partial non-Gaussian jimonfin.2010.07.003 state space models. Journal of Econometrics, 154(1), 85–100. Koop, G., & Korobilis, D. (2013). Large time-varying parameter https://doi.org/10.1016/j.jeconom.2009.07.003 VARs. Journal of Econometrics, 177(2), 185–198. https://doi. Girton, L., & Roper, D. (1977). A monetary model of exchange org/10.1016/j.jeconom.2013.04.007 market pressure applied to the postwar Canadian experience. Kozhan, R., & Salmon, M. (2009). Uncertainty aversion in a het- American Economic Review, 67(4), 537–548. https://www. erogeneous agent model of foreign exchange rate formation. jstor.org/stable/1813387 Journal of Economic Dynamics and Control, 33(5), 1106– Granger, C. W. J. (2008). Non-linear models: Where do we go 1122. https://doi.org/10.1016/j.jedc.2008.11.008 next - time varying parameter models? Studies in Nonlinear Krol, R. (2014). Economic policy uncertainty and exchange rate Dynamics and Econometrics, 12(3), 1639. https://doi.org/10. volatility. International Finance, 17(2), 241–256. https://doi. 2202/1558-3708 org/10.1111/infi.12049 Gupta, R., Olasehinde-Williams, G., & Wohar, M. E. (2020). The Lengnick, M., & Wohltmann, H.-W. (2013). Agent-based financial impact of US uncertainty shocks on a panel of advanced and markets and new Keynesian macroeconomics: A synthesis. emerging market economies. Journal of International Trade & Journal of Economic Interaction and Coordination, 8(1), 1–32. Economic Development, 29(6), 711–721. https://doi.org/10.10 https://doi.org/10.1007/s11403-012-0100-y 80/09638199.2020.1720785 Li, H., Yu, Z., Zhang, C., & Zhang, Z. (2017). Determination of Gupta, R., & Plakandaras, V. (2019). Efficiency in BRICS cur- china’s foreign exchange intervention: Evidence from the rency markets using long-spans of data: Evidence from model- yuan/dollar market. Studies in Economics and Finance, 34(1), free tests of directional predictability. Journal of Economics 62–81. https://doi.org/10.1108/sef-10-2015-0249 and Behavioral Studies, 11(1(J)), 152–165. https://doi. Lin, G., & Schramm, R. M. (2003). China’s foreign exchange org/10.22610/jebs.v11i1(j).2756 policies since 1979: A review of developments and an assess- Huang, Y., & Luk, P. (2020). Measuring economic policy uncer- ment. China Economic Review, 14(3), 246–280. https://doi. tainty in China. China Economic Review, 59, 101367. https:// org/10.1016/j.chieco.2003.09.006 doi.org/10.1016/j.chieco.2019.101367 Liu, J. (2020). Impact of uncertainty on foreign exchange market Huber, F., Kastner, G., & Feldkircher, M. (2019). Should I stay or stability: Based on the LT-TVP-VAR model. China Finance should I go? A latent threshold approach to large-scale mix- Review International, 11(1), 53–72. https://doi.org/10.1108/ ture innovation models. Journal of Applied Economics, 34(5), cfri-07-2019-0112 621–640. https://doi.org/10.1002/jae.2680 Liu, L. (2021). U.S. Economic uncertainty shocks and china’s Huber, F., Koop, G., & Onorante, L. (2021). Inducing sparsity economic activities: A time-varying perspective. Sage Open, and shrinkage in time-varying parameter models. Journal of 11(3), 21582440211032672. https://doi.org/10.1177/215824 Business and Economic Statistics, 39(3), 669–683. https://doi. 40211032672 org/10.1080/07350015.2020.1713796 Liu, L., Zhang, M., & Li, W. (2020). China’s economic policy Hu, X. (1997). Macroeconomic uncertainty and the risk premium in uncertainty shocks and South Korea’s exports: A TVP-VAR the foreign exchange market. Journal of International Money approach with an SMSS structure. Journal of Korea Trade, and Finance, 16(5), 699–718. https://doi.org/10.1016/s0261- 24(4), 1–17. https://doi.org/10.35611/jkt.2020.24.4.1 5606(97)00027-2 Li, X.-L., Li, X., & Si, D.-K. (2020). Investigating asymmetric Huynh, T. L., Nasir, M. A., & Nguyen, D. K. (2020). Spillovers and determinants of the CNY–CNH exchange rate spreads: The connectedness in foreign exchange markets: The role of trade role of economic policy uncertainty. Economics Letters, 186, policy uncertainty. The Quarterly Review of Economics and 4. https://doi.org/10.1016/j.econlet.2019.108827 Finance. Advance online publication. https://doi.org/10.1016/j. Lubik, T., & Matthes, C. (2016). Time-varying parameter vec- qref.2020.09.001 tor autoregressions: Specification, estimation, and an appli- Imbs, J., Mumtaz, H., Ravn, M. O., & Rey, H. (2003). cation. Economic Quarterly, 101(04), 323–352. https://doi. Nonlinearities and real exchange rate dynamics. Journal of the org/10.21144/eq1010403 European Economic Association, 1(2-3), 639–649. https://doi. Mahajan, A., & Wagner, A. J. (1999). Nonlinear dynamics in for- org/10.1162/154247603322391279 eign exchange rates. Global Finance Journal, 10(1), 1–23. Jebabli, I., Arouri, M., & Teulon, F. (2014). On the effects of https://doi.org/10.1016/s1044-0283(99)00002-2 world stock market and oil price shocks on food prices: An Mumtaz, H., & Theodoridis, K. (2018). The changing transmis- empirical investigation based on TVP-VAR models with sion of uncertainty shocks in the U.S. Journal of Business and Liu 19 Economic Statistics, 36(2), 239–252. https://doi.org/10.1080/0 Olasehinde-Williams, G., & Olanipekun, I. (2020). Unveiling the 7350015.2016.1147357 causal impact of US economic policy uncertainty on exchange Mussa, M. (1984). The theory of exchange rate determination. In market pressure of African economies. Journal of Public F. B. John & C. M. Richard (Eds.), Exchange rate theory and Affairs. Advance online publication. https://doi.org/10.1002/ practice (pp. 13–78). University of Chicago Press. pa.2278 Nakagawa, H. (2010). Investigating nonlinearities in real exchange Pástor, & Veronesi, P. (2013). Political uncertainty and risk premia. rate adjustment: Threshold cointegration and the dynamics of Journal of Financial Economics, 110(3), 520–545. https://doi. exchange rates and relative prices. Journal of International org/10.1016/j.jfineco.2013.08.007 Money and Finance, 29(5), 770–790. https://doi.org/10.1016/j. Patnaik, I., Felman, J., & Shah, A. (2017). An exchange mar- jimonfin.2010.03.002 ket pressure measure for cross country analysis. Journal of Nasir, M. A., Naidoo, M. A., Shahbaz, L., & Amoo, L. M. N. (2018). International Money and Finance, 73, 62–77. https://doi. Implications of oil prices shocks for the major emerging econo- org/10.1016/j.jimonfin.2017.02.004 mies: A comparative analysis of BRICS. Energy Economics, Qin, B., Strömberg, D., & Wu, Y. (2018). Media bias in China. 76, 76–88. https://doi.org/10.1016/j.eneco.2018.09.023 American Economic Review, 108(9), 2442–2476. https://doi. Nasir, M. A. (2020). Forecasting inflation under uncertainty: The org/10.1257/aer.20170947 forgotten dog and the frisbee. Technological Forecasting and Schmidt, T., & Zwick, L. (2015). Uncertainty and episodes of Social Change, 158, 120172. https://doi.org/10.1016/j.tech- extreme capital flows in the euro area. Economic Modelling, fore.2020.120172 48, 343–356. https://doi.org/10.1016/j.econmod.2014.10.047 Nasir, M. A. (2021). Zero lower bound and negative interest rates: Sengupta, J. K., & Sfeir, R. E. (1998). Nonlinear dynamics in Choices for monetary policy in the UK. Journal of Policy foreign exchange markets. International Journal of Systems Modeling, 43(1), 200–229. https://doi.org/10.1016/j.jpol- Science, 29(11), 1213–1224. https://doi.org/10.1080/0020 mod.2020.03.014 7729808929610 Nasir, M. A., & Morgan, J. (2018). Pre-Brexit: The EU referendum Shin, M., Zhang, B., Zhong, M., & Lee, D. J. (2018). Measuring as an illustration of the effects of uncertainty on the sterling international uncertainty: The case of Korea. Economics exchange rate. Journal of Economic Studies, 45(5), 910–921. Letters, 162, 22–26. https://doi.org/10.1016/j.econlet.2017. https://doi.org/10.1108/jes-07-2017-0205 10.014 Nasir, M. A., & Simpson, J. (2018). Brexit associated sharp depre- Simo-Kengne, B. D., Ababio, K. A., Mba, J., Koumba, U., & ciation and implications for UK’s inflation and balance of pay- Molepo, M. (2018). Risk, uncertainty and exchange rate behav- ments. Journal of Economic Studies, 45(2), 231–246. https:// ior in South Africa. Journal of African Business, 19(2), 262– doi.org/10.1108/jes-02-2017-0051 278. https://doi.org/10.1080/15228916.2017.1413627 Nasir, M. A., & Vo, X. V. (2020). A quarter century of inflation Taylor, M. P. (1989). Expectations, risk and uncertainty in targeting & structural change in exchange rate pass-through: the foreign exchange market: Some results based on sur- Evidence from the first three movers. Structural Change and vey data. Manchester School, 57(2), 142–153. https://doi. Economic Dynamics, 54, 42–61. https://doi.org/10.1016/j. org/10.1111/j.1467-9957.1989.tb00807.x strueco.2020.03.010 Ter Ellen, S., Verschoor, W. F. C., & Zwinkels, R. C. J. (2013). Neely, C. J. (2015). Unconventional monetary policy had large Dynamic expectation formation in the foreign exchange mar- international effects. Journal of Banking & Finance, 52, 101– ket. Journal of International Money and Finance, 37, 75–97. 111. https://doi.org/10.1016/j.jbankfin.2014.11.019 https://doi.org/10.1016/j.jimonfin.2013.06.001 Nilavongse, R., Rubaszek, M., & Uddin, G. S. (2020). Economic Tervala, J. (2019). U.S. Monetary policy and China’s exchange policy uncertainty shocks, economic activity, and exchange rate policy during the great recession. International Journal rate adjustments. Economics Letters, 186, 4. https://doi. of Finance & Economics, 24(1), 113–130. https://doi. org/10.1016/j.econlet.2019.108765 org/10.1002/ijfe.1652 Olanipekun, I. O., Güngör, H., & Olasehinde-Williams, G. (2019a). The Economist. (2020). Is China suppressing the yuan?, October Unraveling the causal relationship between economic policy 31st–November 6th 2020. uncertainty and exchange market pressure in BRIC countries: Wang, Y., Li, X., Huang, D., & Wang, A. (2020). Is there evidence Evidence from bootstrap panel granger causality. Sage Open, of “fear of appreciation” in China’s foreign exchange market 9(2), 13. https://doi.org/10.1177/2158244019853903 intervention? Applied Economics, 52(6), 537–551. https://doi. Olanipekun, I. O., Olasehinde-Williams, G., & Güngör, org/10.1080/00036846.2019.1646878 H. (2019b). Impact of economic policy uncertainty on Zheng, L., Jiang, Y., & Long, H. (2019). Exchange rates change, exchange market pressure. Sage Open, 9(3), 13. https://doi. asset-denominated currency difference and stock price fluctua- org/10.1177/2158244019876275 tion. Applied Economics, 51(60), 6517–6534. https://doi.org/1 Olanipekun, I. O., & Olasehinde-Williams, G. (2021). Unveiling the 0.1080/00036846.2019.1624920 causal interactions among the US monetary policy uncertainty, Zhou, Z., Fu, Z., Jiang, Y., Zeng, X., & Lin, L. (2020). Can eco- exchange market pressure, and FDI-flows in large emerging nomic policy uncertainty predict exchange rate volatility? New market economies. Journal of Public Affairs. Advance online evidence from the GARCH-MIDAS model. Finance Research publication. https://doi.org/10.1002/pa.2684 Letters, 34, 101258. https://doi.org/10.1016/j.frl.2019.08.006 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png SAGE Open SAGE

Economic Uncertainty and Exchange Market Pressure: Evidence From China:

SAGE Open , Volume 12 (1): 1 – Jan 18, 2022

Loading next page...
 
/lp/sage/economic-uncertainty-and-exchange-market-pressure-evidence-from-china-xkchilB7Vd

References (90)

Publisher
SAGE
Copyright
Copyright © 2022 by SAGE Publications Inc, unless otherwise noted. Manuscript content on this site is licensed under Creative Commons Licenses.
ISSN
2158-2440
eISSN
2158-2440
DOI
10.1177/21582440211068485
Publisher site
See Article on Publisher Site

Abstract

This paper evaluates the impact of local and external economic uncertainty shocks on China’s exchange market pressure from a time-varying perspective. We first construct a simple behavioral model to provide some economic background for our empirical analysis. The model identifies two channels, namely market sentiment and market demand, through which economic uncertainty has a time-varying impact on exchange market pressure. Notably, we calculate a new exchange market pressure index for China by considering China’s practice. Using the economic policy uncertainty index as a proxy for economic uncertainty and our new estimates of China’s exchange market pressure, we employ a novel TVP-VAR model controlling for over-parameterization to a monthly dataset from January 2001 to October 2020 to analyze the impact of China’s and U.S. economic policy uncertainties on China’s exchange market pressure. Our empirical findings robustly show that an upsurge in U.S. EPU is followed by appreciation pressure of the RMB against the dollar, while a hike in China’s EPU triggers devaluation pressure on the RMB. In addition, the impact of economic policy uncertainty has considerable time variation in magnitude, especially after mid-2011, showing a trade-off mechanism between the effects of domestic and foreign economic uncertainties on China’s exchange market pressure. Moreover, we attribute the time- varying features to the changes in China’s dependence on the U.S. and in the exchange rate flexibility. Finally, China should further improve the exchange rate flexibility, reduce its dependence on the U.S. and develop a more diversified currency basket in the exchange rate formation mechanism. JEL: F31 C50 E70 Keywords economic Uncertainty, EMP, EPU, TVP-VAR, Over-parameterization pressures on the exchange rate to depreciate or appreciate. Introduction Ideally, these pressures will be fully reflected in the There is an anecdotal story about globalization that eco- exchange rate movements for the countries with floating nomic activity in an open economy is primarily driven by exchange rate regimes. However, this would not happen in foreign shocks, while domestic shocks are reflected in its the economies like China and Japan, where the central currency value (Nilavongse et al., 2020). Since economic banks are known for intervening in the foreign exchange uncertainty is a representative shock that conveys informa- market (hereafter FX market). tion about overall economic conditions, various attempts In China, government intervention in the FX market is a have been made to assess the impact of economic uncertainty stylized fact (Das, 2019; Wang et al., 2020). Accordingly, on the exchange rate, see Kido (2016), Simo-Kengne et al. since the exchange rate is managed to change, depreciation (2018), Nasir and Morgan (2018), Nilavongse et al. (2020), or appreciation pressures on the renminbi against other cur- among others. Granted that most of the literature argues that rencies cannot be manifested totally in the exchange rate the exchange rate is affected by local and external uncertain- movements. Thus, uncertainty shocks may have a negligible ties, but Simo-Kengne et al. (2018) and Nasir and Morgan impact on the realized exchange rate in China owing to cen- (2018) highlight that foreign uncertainty plays a small role in tral bank intervention, although it has been argued that driving the exchange rate. Theoretically, economic uncertainty is the primary source of risk of holding a currency (Hu, 1997). Elevated Nantong University, Jiangsu, China economic uncertainty can make it challenging to predict Corresponding Author: the exchange rates (Beckmann & Czudaj, 2017a, 2017b), Lin Liu, Nantong University, Room 515, Teaching building #10, No.9, leading agents to reduce their demand for the currency or Seyuan Road, Chongchuan District, Nantong, Jiangsu 226019, China. raise the risk premium (Taylor, 1989), thereby putting Email: liulintyu@outlook.com Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License (https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage). 2 SAGE Open market expectations are affected by the uncertainty shocks 2013), exchange rate dynamics are often interpreted as non- (Beckmann & Czudaj, 2017a, 2017b; Ter Ellen et al., 2013). linear processes in the literature; see Sengupta and Sfeir Naturally, when analyzing the impact of uncertainty for (1998), Mahajan and Wagner (1999), Imbs et al. (2003), the economies that resort to FX intervention, the realized Nakagawa (2010), among others. Second, the impact of eco- exchange rate needs to be restored to incorporate govern- nomic uncertainty shocks may be nonlinear due to changing ment intervention. A well-known measure of such an dynamics, policy regimes, and economic shocks (Mumtaz & exchange rate recovery is the exchange market pressure Theodoridis, 2018). Third, China has undergone many dra- (hereafter EMP) index. While most previous studies have matic phases over the past decades (Chang et al., 2016) and focused on the realized exchange rates and exchange rate has substantially reformed its exchange rate system, raising expectations, the impact of economic uncertainty on EMP the possibility of potential nonlinearities and multi-equilibria has been less touched upon, except for two path-breaking (Liu, 2021). Therefore, it is reasonable and desirable to apply works by Olanipekun, Güngör et al. (2019) and Olanipekun, a nonlinear model when exploring the relationship between Olasehinde-Williams et al. (2019). They, however, only aim the foreign exchange market and the macroeconomy. at the causal relationships between economic uncertainty and Theoretically, any nonlinear model can be approximated EMP. Two recent studies by Olasehinde-Williams and by a time-varying parameter (TVP) model (Granger, 2008). Olanipekun (2020) and Olanipekun and Olasehinde-Williams Given the appeal of TVP models, a few studies regarding (2021) provide empirical evidence that U.S. uncertainty exchange rate dynamics have utilized TVP-VAR models to affects EMP in some emerging market countries and African examine the effects and causes of exchange rate movements; economies, respectively. Nonetheless, these two studies do see Choi et al. (2018), Nasir and Morgan (2018), Zheng et al. not consider the impact of domestic uncertainty on EMP, (2019), Nasir and Vo (2020), and others. However, assuming which may be jointly motivated by local and external factors, that all model parameters are time-dependent increases as derived from the theory of exchange rate determination model complexity and may suffer from the “curse of dimen- (Mussa, 1984). A similar issue also exists in the study by Liu sionality” of over-parameterization, which would eventually (2020), who explores the impact of domestic uncertainty bias the estimates. Recent studies have shown their concerns shocks on China’s EMP, overlooking the potential effects over this issue; see, for example, Koop and Korobilis (2013), exerted by foreign uncertainty. However, because of China’s Eisenstat et al. (2016), Huber et al. (2019), Chan et al. (2020). high dependence on the United States, China’s financial mar- On this account, using a monthly dataset running from kets may endure the significant impact of uncertainty shocks January 2001 to October 2020, we conduct our empirical from the United States (Gupta et al., 2020; Liu, 2021). study with a TVP-VAR model by combing the approach Against this backdrop, we attempt to fill this gap by inspect- developed by Eisenstat et al. (2016) to control for ing China’s EMP dynamics following unexpected rises in over-parameterization. domestic and foreign economic uncertainties. Nevertheless, an essential issue that setting a proper proxy In this paper, we explicitly assess the role of economic for economic uncertainty arises in our empirical study. uncertainty in governing EMP with China’s monthly data. Currently, there are three optional proxies for economic China has been pursuing the globalization of the renminbi uncertainty in the literature; the news or reports-based eco- and building a responsible national image. Hence, maintain- nomic policy uncertainty index (EPU) (Baker et al., 2016), ing the exchange rate stability is a pressing concern for the real data-based macroeconomic uncertainty index (Jurado China. As the exchange rate stability is critical to the macro- et al., 2015), and the survey-based uncertainty index economy and plays a crucial part in the Chinese govern- (Bachmann et al., 2013). Comparatively, the news-based ment’s objectives (Das, 2019), evaluating the impact of uncertainty index is computed by text-mining newspaper economic uncertainty on EMP would benefit the Chinese reports available equally to every market participant. policymakers to better understand the underlying causes of Therefore, the EPU index would catch more the uncertainty exchange rate fluctuations. caused by specific events (Shin et al., 2018), as these events Specifically, as documented in Pástor and Veronesi (2013) tend to evoke public or political concerns about the economic and Krol (2014), the impact of economic uncertainty on consequences, leading to an increase in the frequency of EMP might be state-dependent, and thus motivated by economic-relevant news and thus a higher EPU. In addition, Bartsch (2019), we construct a simple behavioral model with the news reports used to construct EPU are usually easy to heterogeneous agents to demonstrate that the impact of eco- understand for market participants and may sometimes sig- nomic uncertainty on EMP, mainly through the market senti- nal policy stance, especially for countries where the govern- ment and market demand channels, evolves over time. ment controls public media in some way. Davis et al. (2019) Importantly, as our analytical result indicates, it is neces- construct an EPU index for China by retrieving news reports sary to consider nonlinearity when modeling EMP and eco- from two influential newspapers circulating in mainland nomic uncertainty. First, since the FX market has been China, namely the People’s Daily and the Guangming Daily. surveyed as dominated by heterogeneous agents using differ- Since these two government-run newspapers are the fore- ent trading strategies (de Jong et al., 2010; Ter Ellen et al., most government mouthpieces in China (Qin et al., 2018), Liu 3 market participants may translate an upsurge in the frequency index representing China’s exchange market pressure by of economic-relevant coverage into a rise in economic uncer- considering the occasional expenditure of FX reserves and tainty and policy concerns. Therefore, in this paper, we use the intervention implemented in recent years via the counter- EPU as a proxy for economic uncertainty. cyclical adjustment factor. Doing so distinguishes our paper Additionally, as the largest member holding dollar assets, in a significant way from several previous studies, for exam- China has been witnessed as a dollar-pegged country for a ple, Olanipekun, Güngör et al. (2019), Olanipekun, long time (Tervala, 2019), even though moving toward a Olasehinde-Williams et al. (2019), and Liu (2020), who all more flexible exchange rate regime after the reform initiated use slightly biased EMP estimates for China. in July 2005. Moreover, trading of the renminbi against the Third, in line with the theory of exchange rate determina- U.S. dollar dominates China’s onshore spot FX market, tion, an exchange rate is determined concomitantly by accounting for more than 95% of the overall FX transactions domestic and foreign fundamentals, which have been well- on average. Therefore, in our empirical section, we estimate documented to be affected by the uncertainty shocks (Bloom, a new EMP index of the RMB against the dollar by consider- 2009, 2014). Moreover, the uncertainty has been found to ing the contingent spending of FX reserves and the interven- spill over globally, especially from advanced economies to tion through the counter-cyclical adjustment factor, which less developed countries (Liu, 2021). Hence, in this paper, have been overlooked in most previous related studies, see, unlike Liu (2020), Huynh et al. (2020) and others who only for example, Olanipekun, Güngör et al. (2019), Olanipekun, consider the impact of local (or external) uncertainty, we Olasehinde-Williams et al. (2019), and Liu (2020). To this assess the impact of China’s and U.S. uncertainties on end, we estimate a new EMP index for China using daily China’s EMP in a consistent model. exchange rate data from August 11, 2015, when China Finally, as outlined above, we resort our analysis to a authority announced adopting a new regime for formulating novel TVP-VAR model to incorporate potential nonlineari- the central parity of the RMB/USD exchange rate, to October ties. Liu (2020) also uses a classic TVP-VAR model to evalu- 30, 2020, the last trading day of our sample. ate the impact of macroeconomic and financial uncertainties The empirical findings based on a monthly dataset for on EMP for China. Differently, considering the over-param- China from January 2001 to October 2020 show that a ris- eterization problem arisen from the time-varying specifica- ing shock in U.S. EPU would trigger the appreciation pres- tions, we apply a typical TVP-VAR model with the stochastic sure on the RMB against the dollar, while a hike in China’s model specification (SMSS) framework proposed by EPU is followed by the devaluation pressure of the RMB Eisenstat et al. (2016) to implement our empirical analysis. against the dollar. Moreover, the magnitude of the EPU The rest of our paper is arranged in the following manner. impact on EMP is found to be time-dependent. Specifically, The next section briefly reviews the current literature rele- the effect is relatively stable until mid-2011 but more vola- vant to this paper. Section 3 illustrates the economic intuition tile thereafter. As for the rationale for this time-varying that economic uncertainty affects exchange market pressure impact, we attribute it to changes in China’s dependence on through a simple behavioral exchange rate model. Section 4 the U.S. and increased exchange rate flexibility. Further, describes the empirical methodology and the data we use. the empirical evidence indicates a trade-off mechanism The empirical results and discussions, as well as robustness between the effects of China’s and U.S. economic policy checks, are analyzed in Section 5. The last section concludes uncertainties on China’s EMP. the whole paper and highlights the policy implications. Our marginal contribution to the growing literature is that we focus on the time-varying impact of local and external Related Literature economic uncertainties on China’s exchange market pres- sure. Our work reaches a step closer to an understanding of Overall, our work is related to two strands of the existing how the FX market responds to the shocks originating from literature. One of these strands has been devoted to uncov- domestic and foreign economic uncertainties. First, unlike ering the link between economic uncertainty and exchange Kido (2016), Nasir and Morgan (2018), Huynh et al. (2020), rate movements, whereby the focus, however, has been pri- Nasir (2020), Chen et al. (2020), among others, who center marily on exchange rate returns. Another direction of the on the roles of uncertainty in the exchange rate dynamics, we literature, which is the closest to our work, has been to attempt to explore the impact of uncertainty on China’s EMP. assess the role of economic uncertainty in provoking pres- Since the RMB exchange rate remains under control, its fluc- sures on the exchange rate. tuations are less informative in revealing market pressures. The impact of economic uncertainty on exchange rate Second, as documented in The Economist (2020), China’s movements (both returns and volatilities) is increasingly intervention in the FX market has become less pronounced in well documented. Earlier studies find pricing effects of recent years, keeping FX reserves fairly intact. Therefore, a uncertainty on the currency risk premium; see, for example, better index capturing exchange market pressure should con- Taylor (1989) and Hu (1997). Recently, there has been evi- sider the changes in FX reserves and the veiled measures as dence for the roles of home and foreign uncertainties in well. Consequently, we estimate a new and more accurate explaining exchange rate movements in industrial and 4 SAGE Open emerging economies, albeit analyzing mainly the spillovers particular, a bidirectional relation between global EPU and of U.S. EPU. However, while the impact of home-grown EMP and a one-way causality from EMP to EPU were found uncertainty on exchange rate stability is well understood, for China. Further, Olanipekun, Olasehinde-Williams et al. the findings on the influence of foreign uncertainty are (2019) expand the sample to 20 countries and use four EMP mixed. Compared to developed countries, Krol (2014) finds measures as in Aizenman and Binici (2016) to conclude that that exchange rate volatilities in emerging economies are domestic EPU and EMP are cointegrated in the long run. less affected by U.S. EPU during recessions. The author Most recently, Olasehinde-Williams and Olanipekun (2020) attributes the explanation to the low level of financial open- and Olanipekun and Olasehinde-Williams (2021) report ness of these economies to the United States. However, more evidence about the causality from uncertainty to EMP Huynh et al. (2020) offer empirical findings that exchange for African economies and emerging market countries, rate returns and volatilities of nine international currencies respectively. Moreover, based on China’s data, Liu (2020) against the U.S. dollar are associated with U.S. trade policy finds that macroeconomic uncertainty and financial uncer- uncertainty and global economic policy uncertainty. Chen tainty weaken (strengthen) EMP in a state of RMB apprecia- et al. (2020) show that the EPU shocks resulting from the tion (depreciation). United States, Europe, and Japan have significant and The recent study by Liu (2020) is undoubtedly the closest asymmetric effects on the Chinese onshore RMB/USD to ours. As we will do in our paper, Liu (2020) uses a TVP- exchange rate volatility. Nevertheless, the exchange rate VAR model to investigate the impact of China’s domestic volatilities of developed countries may be less exposed to uncertainty on EMP and the jump risk of the RMB/USD U.S. uncertainty shocks. Nasir and Morgan (2018) and exchange rate. However, there are several differences. First, Nilavongse et al. (2020) highlight the significant devalua- Liu (2020) jointly estimates the effects of macroeconomic tion effect of uncertainty locally from the UK on the ster- uncertainty and financial uncertainty measures, estimated ling exchange rate. Notably, the latter and Simo-Kengne using the methodology developed by Jurado et al. (2015). et al. (2018) reckon that foreign EPU is not a determinant Moreover, Liu (2020) models the impact of domestic uncer- of the exchange rate. In addition, Bartsch (2019) documents tainty only, but it is necessary to incorporate foreign uncer- that UK EPU rather than U.S. EPU impairs the stability of tainty since, in theory, the exchange rate is determined by the USD/GBP exchange rate. domestic economic conditions and foreign counterparts However, a growing literature demonstrates the salient simultaneously. Second, Liu (2020) directly uses the EMP impact of foreign uncertainty on the exchange rate returns. measure estimated by Patnaik et al. (2017). A closer scrutiny Kido (2016) finds that the real effective exchange rate is of Patnaik et al. (2017)’s EMP estimates for China reveals negatively correlated with U.S. EPU for Australia, Brazil, that they have overlooked China’s intervention through Korea, and Mexico except for Japan, which exhibits a posi- adjusting the central parity rate in recent years and the irregu- tive pattern, indicating that the yen plays as a safe-haven cur- lar spending of FX reserves. However, we estimate a new rency in the face of the U.S. uncertainty. Beckmann and Chinese EMP index by considering these practices. Third, Liu Czudaj (2017a) further corroborate the yen’s safe-haven (2020) relies on a four-variable TVP-VAR model of uncer- standing, though they concentrate on the aftermath of uncer- tainty and market stability to achieve an empirical analysis. tainty on the exchange rate expectations. Addressing the Nevertheless, as documented in Aizenman and Binici (2016), spread between the onshore and offshore RMB/USD variables regarding market sentiment and market demand, as exchange rates, a recent paper by Li et al. (2020) documents well as macroeconomic fundamentals, which might change in a finding that a rise in their constructed composite EPU the light of economic uncertainty, could be important factors (extracted from the EPU indices of China and G7) widens affecting EMP. Therefore, we will consider these variables to the spread. In addition, a closely related literature in this line assess the impact of economic uncertainty on EMP. In doing of research centers on the impact of relative uncertainty. so, we can better explore the underlying mechanism of how Focusing on the relative value of difference or ratio of economic uncertainty shapes EMP. Finally, we differ mark- domestic EPU to external EPU, Balcilar et al. (2016), edly from previous studies in that we illustrate some eco- Christou et al. (2018), and Zhou et al. (2020) show that the nomic background for our empirical study through a simple relative EPU has much power in predicting exchange rate behavioral exchange rate model with heterogeneous agents. movements. Another strand of the literature has attempted to dissect A Simple Behavioral Exchange Rate the relation between economic uncertainty and EMP. Model Olanipekun, Güngör et al. (2019) study a case of four BRIC countries but give somewhat puzzling results. In the paper, We extend the FX market model in Dieci and Westerhoff Olanipekun, Güngör et al. (2019) find the one-way causation (2010) to incorporate economic uncertainty. Specifically, we from global EPU to EMP and mutual interplays between consider a world consisting of two-country and a domestic domestic EPU and EMP in all four countries, which is incon- FX market in which two currencies are traded. Two types of sistent with the results of the country-specific analysis. In traders, namely fundamentalists and chartists, invest in the Liu 5 market. Moreover, investors switch between these two trad- 1 W = , ing strategies depending on market conditions. To conve- (5) 1+− fS S () tt niently describe exchange market pressure, we assume no trading restrictions in the market (e.g., capital controls and f > 0 where parameter is a sensitive parameter controlling FX intervention). chartists’ insistence on the chartist trading strategy for a given exchange rate misalignment. A higher f is accompa- nied by a lower , indicating more traders switch to funda- Fundamentalists t mentalists. Furthermore, (4) suggests that more traders turn Following Dieci and Westerhoff (2010), fundamentalists are fundamentalists when the exchange rate deviates more from usually assumed to formalize their demand as, its fundamental price. Recall that the FX market is supposed to be perfect and FF (1) DS =− β S , () free of intervention, so following the EMP definition in t tt Girton and Roper (1977), we can define exchange market where S is the indirectly quoted (log) nominal exchange t pressure in our context as, rate and represents the fundamental exchange rate. S β > 0 reflects the belief in mean reversion. EMPS =− S . (6) tt ++ 11 t It is indicated in (1) that if the exchange rate is overvalued (undervalued), fundamentalists would expect the exchange Using (6), we can obtain EMP from (3), rate to move toward its underlying fundamental price, pro- F C moting them to reduce (increase) their demand for the local EMPW =− α 1 DW + D . () () (7) tt +1 t tt currency against the foreign one. Clearly, (7) provides a theoretical basis for an intuition that EMP is jointly determined by market sentiment and cur- Chartists rency demand. For simplicity, follow Dieci and Westerhoff (2010), we define the chartists’ demand as follows, Role of Economic Uncertainty CC DS =− β S , (2) () t tt Our model suggests that investors’ demands for currency partially depend on the fundamental exchange rate, which where parameter β > 0 governs the confidence in the per- is determined by domestic and foreign macroeconomic sistence of deviations. variables in traditional theory. Since economic uncer- tainty has detrimental effects on the macroeconomy (Bloom, 2009, 2014; Jurado et al., 2015), investors’ Evolution of the Exchange Rate F C demands D , D and market sentiment would be t t As in Dieci and Westerhoff (2010, 2013), the exchange rate influenced by economic uncertainty through the funda- h a at time t + 1 is determined by the excess demand formed by mental exchange rate S . Therefore, let and denote δ δ t t t heterogeneous investors in period t. Thus, the exchange rate unexpected rising shocks in domestic and foreign eco- is developed by, nomic uncertainties, respectively, we can model S as a binary function of δ and δ , t t F C SS =+ α 1−WD +WD , () (3) () tt +1 tt tt h a Sg = δδ ,, () tt t (8) where parameter α> 0 controls the price adjustment. The gg <> 00 , proportion of chartists in period t is denoted by , while where the assumptions on the first-order partial deriva- fundamentalists occupy the portion of because we 1−W g > 0 g > 0 tive that and imply that an increase in domes- 1 2 assume only these two types of investors in the market. tic uncertainty leads to a decline in the exchange rate, while Lengnick and Wohltmann (2013) and De Grauwe and Ji a rise in foreign uncertainty causes the foreign currency to (2020) refer to these portions as market sentiment, reflecting depreciate. which trading rule prevails in the market. Motivated by De FC ββ = For simplicity, let , substituting (4) into (7) Grauwe and Ji (2020), we define market sentiment as a mea- yields, sure of the dominance of fundamentalists, EMPE =− αβ SS , EW =− 11 −= WW − 2. (4) () (9) () t+1 tt t tt tt In addition, following Dieci and Westerhoff (2010), we We hence can derive the impact of economic uncertainty define the proportion as, on EMP by taking partial derivatives based on equation (9), t 6 SAGE Open the variabilities in the exchange rate misalignments and the     share of chartists, as well as market sentiment. Moreover, as gE +− 4gW fS S () ∂EMP 11 t tt t   t+1 F analyzed by Chiarella et al. (2017), traders’ beliefs and sen- = αβ 12 4443 444 4  ∂D  ∂δ ∂E t sitivity to price misalignments, which we have assumed to be t h  ∂δ h  (10) ∂δ   constant, may also change over time, bringing additional   sources to the time variability of the impact. F 2 =+ αβ gE 4. Wf SS − () 1 tt tt () Empirical Methodology and Data Similarly, Empirical Model We work with a VAR framework to allow for potential   endogeneity generated by the self-filling mechanism and the   gE +− 4gW fS S ∂EMP ()  22 t tt t  t+1 F autocorrelations. Since a large body of evidence indicates = αβ 12 4443 444 4  ∂D  ∂δ ∂E various nonlinearities in the macroeconomic time series, it t a  ∂δ a  ∂δ (11) is quite desirable to take these nonlinearities into account     when studying the behavior of the macroeconomy. While F 2 many studies have contributed to nonlinear modeling, the =+ αβ gE 4. Wf SS − () 2 tt tt () results obtained from nonlinear models may conflict with economic theory. However, Granger (2008) demonstrates that From (10)–(11), the impact is straightforwardly decomposed any nonlinear model can be approximated by a time-varying ha , () ha , () into two components, that is, and , ∂∂ D / δ ∂∂ E / δ tt tt parameter model, which is generally linear but still has suffi- where D denotes the total currency demand. It is apparent cient power to capture nonlinearities. Consequentially, during that the sign of each derivative basically depends on the past decade, time-varying parameter vector autoregres-  2  , which can be easily transformed sions (TVP-VARs) have gained widespread popularity among EW +− 4 fS S  ()  tt tt   applied macroeconomists and have become a standard 2   into 14 −WW +− fS S −1 . Thus, if the squared framework for analyzing macroeconomic time series in the ()  ()  tt tt   last decade due to their charm of tracking processes subject deviation of the exchange rate from its fundamental value to structural breaks or regime shifts (Baumeister & Peersman, 2013). While time variations in coefficients and SS − is no less than the threshold 14 / () f , then the () tt conditional higher moments have been separately well doc- () ha , signs of ∂∂ EMP / δ only rely on the effects of uncer- tt +1 umented in the literature, a TVP-VAR model unites time g g tainty on the fundamental exchange rate, that is, and , 1 2 variations jointly and concomitantly in coefficients and respectively. Therefore, if the condition is satisfied, an variances, allowing the data to speak freely and capturing a upsurge in domestic (foreign) economic uncertainty is wide range of time variation and nonlinearity (Lubik & expected to exert depreciating (appreciating) pressure on the Matthes, 2016; Nasir et al., 2018; Nasir & Simpson, 2018; () ha , exchange rate. Further, the signs of cannot ∂∂ EMP / δ tt +1 Nasir & Vo, 2020). be clearly identified when the condition is violated that Although a TVP-VAR model is very flexible and can , mainly on account of the uncertainty SS − < 14 / f () model nearly all nonlinear relationships among the variables () tt stemming from the impact of economic uncertainty on mar- of interest, it is highly parameterized and may risk overfit- ket demand. ting, leading to inaccurate estimation of impulse response By inspecting (10) and (11), the effect of economic uncer- functions. Recently, a growing literature has developed sev- tainty on market sentiment is certain that an unexpected eral methods to mitigate these over-parameterization con- increase in domestic (foreign) uncertainty would induce a cerns, see, for example, Koop and Korobilis (2013), Eisenstat negative (positive) response in market sentiment, making et al. (2016), Huber et al. (2019), Chan et al. (2020). In this fewer (more) traders using fundamental trading rule. part of the literature, a widely used method is to use global- However, the effect on market demand essentially lies in mar- local shrinkage priors to reduce estimation bias and improve ket sentiment E . More specifically, in a market dominated model performance. However, applying shrinkage to time- by fundamentalists, a rising shock in domestic (foreign) eco- varying parameter models cannot wholly eliminate estima- nomic uncertainty would cause the relative demand for the tion errors (Huber et al., 2021). Moreover, incorporating domestic currency to decline (increase). Otherwise, the shrinkage is less straightforward, and it often requires com- demand would increase (reduce) in accordance with domestic putationally demanding algorithms or approximate inference (foreign) economic uncertainty when chartists prevail. (Eisenstat et al., 2016). Eisenstat et al. (2016) propose a new Remarkably, as shown in (10) and (11), the impact of eco- approach that combines the stochastic model specification nomic uncertainty on EMP would be time-dependent due to search (SMSS) framework developed in Frühwirth-Schnatter Liu 7 and Wagner (2010) with a typical TVP-VAR to ensure model α α= αα ,, … Let denotes the initial states () 1 mmp+1 () parsimony. Comparatively, the methodology in Eisenstat γβ =− αω / of ββ and reparametrize () , for jt jt jj et al. (2016) is more flexible and efficient in that it allows the , so that (1)–(2) can be written as, jm =… 11 ,, () mp + model to automatically and endogenously choose between a time-varying parameter against a constant parameter for (4)  2 yX =+ α αγ XV γ + tt tt t each VAR coefficient. (5) γγγ =+ γη η tt−1 t A typical TVP-VAR model. The generic state-space form of a typical structural TVP-VAR with stochastic volatility can where η η ~,  0 I and  is independent of each () t t mmp+1 () be written as, other for all leads and lags, whereas . Finally, α αα ~,  () α A (4)–(5) constitute the conventional TVP-VAR model with By =+ X ββ  , (1) tt tt t stochastic volatility. βββ =+ βη η , (2) tt−1 t The SMSS framework. To alleviate the over-parameter- ization worries around ββ , Eisenstat et al. (2016) propose a where y is an m-dimensional vector of endogenous vari-  2 ables of interest at time , when the regressor tT =… 1, , Tobit prior on the diagonal elements of , that is, . For V ωω ′ ′ matrix Xy =⊗ I 1,,…, y consists of an intercept () tt tt −− 1 p jn =… 1, , , introduce a latent variable and assume to follow and p lags of y . is a lower unitriangular matrix contain- a normal distribution, ing contemporaneous relations. ββ is an vector mmp +1 () * 2 ωμ ~,  τ , of stacked coefficients at time t and is assumed to evolve as () jj j a random walk process with the errors , where ηη is ηη t t and relate to ω by the following indicator function, Gaussian innovations with zero mean and a diagonal covari- j j ance matrix V . In the matrix V, each element 00 , if ω ≤  j vj =… 11 ,, mmp + is the variance regarding coefficient () () j ω = ** (6) β , determining the degree of its time variability. ωω , if > 0  jj jt The residual follows an independent Normal distribu- In addition, in order to incorporate the Lasso structure, it tion with zero mean and a variance-covariance matrix ΣΣ , is assumed that τ follows an Exponential distribution, which is assumed to be time-varying and diagonal with ele- 2 2   ments σ = exp h , that is, λ () it it 2 τ ~   , where λλ ~,  λ , denotes the j ()  01 02      exp h L 0  () 1t Gamma distribution.   Σ Σ = MO M , It is clear to see that the Tobit prior automatically restricts t     as per ω , while still allowing for a straightforward 0 L exp h () ω ≥ 0 it j   Gibbs sampler, which is fast and suitable for implementing where h =… hh ,, is defined as the log-volatility () hierarchical shrinkage through the hyper-parameters of , tt 1 it ω evolving in a random walk fashion, that is, m and τ . hh =+ wR ,~ w  0, () (3) tt −1 tt Model estimation. Following Eisenstat et al. (2016), the MCMC Gibbs sampler is adopted for estimation and com- This setup is motivated by the well-documented evidence puting the posterior distributions for the parameters and of allowing variance-covariance of residuals to vary over hyper-parameters. Generally, the MCMC Gibbs sampler is time in modeling macroeconomic data, see Fernández- distinctively straightforward and efficient since it samples Villaverde et al. (2015), for instance. Following Eisenstat from the full conditional probability distribution. Crucially, et al. (2016), we assume that h is initialized with the MCMC is a smoothing method that produces smoothed hV ~,  0 and the prior on the transition covariance is () 00 estimates (Nasir & Morgan, 2018). Most importantly, the Rv ~, W R , where denotes the inverse Wishart () W MCMC Gibbs sampler is suitable for our case. As discussed distribution. earlier, introducing the Tobit prior to shrink model dimen- As the matrix is lower unitriangular, we can rewrite sionality would lead to a computationally feasible and fast (1)–(2) by rearranging X to include contemporaneous y Gibbs sampler. In addition, the posterior computations using t t and rearranging to include the free elements in B . Then, ββ MCMC are outlined to deliver an overview of the sampling t t the covariance matrix is no longer a diagonal but a full V τ procedure. First, for in =… 1, , , , , and are stacked ω ω i i 1 1 as vectors , ωω , and . Then, posterior draws are obtained ωω ττ  2 2 matrix that can be decomposed as , where VV = Φ ΦΦ Φ′V ΦΦ γγ by sequentially sampling by the following order: αα , , , 1 ΣΣ  2 is a lower unitriangular matrix, and , , , and . For details on the MCMC Gibbs sampler, the V =… diag ωω ,, ωω ττ λ () 1 n . reader is referred to Eisenstat et al. (2016). nm =+ () mp 1 8 SAGE Open Prior settings. The priors are set in line with Eisenstat non-transparent measures rather than intervening directly, et al. (2016), as all sample data will be standardized before especially after 2015. The central parity rate has been a usual estimation, making the priors standard to some extent. As tool since the People’s Bank of China (PBoC), China’s cen- shown above, the priors for the hyper-parameters, and tral bank, first announced it on August 11, 2015. Since the ΣΣ R are assumed to follow independent inverse-Wishart distri- reform on July 21, 2005, the PBoC has been operating a sys- butions, while ττ and are set to be distributed as Expo- tem that allows the daily exchange rate to fluctuate within a nential and Gamma distributions. In addition, the priors for narrow band, which was expanded from the initial 0.3% to parameters and are assumed to be normally distributed. 2% in March 2014, around a target called the central parity αα ω Further, we set the following hyper-parameters on these pri- rate. This central parity rate was based on the closing price ors to end the SMSS specification of the TVP-VAR model, on the last trading day before August 11, 2015, and thereaf- ter, is based on the last trading day’s closing price plus the needed changes, which refer to the needed adjustment of the α α == 00 ,, AI ,, hV == I vm =+11, 00 00 m 0 mm () p+1 RMB/USD exchange rate to offset the overall impact of the RI =− 00 . 11 vm − , μ = 00 0,. λλ == 1. () 0 0 m 01 02 fluctuations of the currencies in the currency basket against the dollar on the last trading day and overnight. In addition, a counter-cyclical adjustment factor (hereafter CCAF), initi- Finally, the above empirical model and framework will be ated on May 26, 2017, to mitigate irrational market senti- applied to a Chinese monthly dataset to exploring the roles of ment, has been added to the regime since then. The PBoC has economic uncertainty in China’s exchange market pressure. occasionally intervened in the market through the CCAF afterward. In this sense, the CCAF is viewed as an instru- Data ment for FX intervention (Das, 2019). Therefore, it is neces- sary to include this CCAF in the estimation of China’s EMP. We conduct our empirical investigation with a monthly data- Further, the PBoC had directly capitalized four big state- set from January 2001 to October 2020, primarily based on controlled banks with FX reserves between 2003 and 2008. data availability. Additionally, a principal consideration in Ignoring these expenditures for non-intervention purposes our decision to start our sample from January 2001 is that would bias the EMP estimates. Hence, we differ from previ- China’s macroeconomic data prior to 2001, during which ous studies in that we take these non-intervention purpose China underwent profound institutional and economic transi- expenditures back and accordingly recover the reserves. tions, were less reflective of the transmission dynamics Thus, the formula for calculating China’s EMP is defined among economic variables (Fernald et al., 2014). as follows, Our baseline model centers on five main variables indi- cated by our behavioral model presented in Section 3, that is, SS − RR − EMP, domestic and foreign economic uncertainties, market tt−1 tt−1 − ,, before August 2015 sentiment, and market demand in the FX market. S R t−1 t−1 EMP = , SS − − CCA AF () RR − tt−1 t  tt−1 Estimates of China’s EMP. Girton and Roper (1977) theo- − ,afterwards S R  t−1 t−1 rize that EMP is the sum of exchange rate fluctuations and official intervention. For the former, as discussed earlier, we where R represents the (recovered) FX reserves. Note that focus on the onshore RMB exchange rate against the dol- here is the onshore RMB/USD exchange rate expressed in lar since this trading pair dominates China’s FX market. the direct quotation. For the latter, the growth of FX reserves has been primar- However, since the PBoC does not make public the CCAF, ily considered in the literature, as the government usually we need to estimate it first. To this end, based on the forma- intervenes through transactions on FX reserves. In addition, tion regime of the central parity rate, we regress the differen- underpinned by the assumption that the central banks can tial between the central parity rate and the last closing price of intervene by adjusting short-term interest rates, Klaassen and the RMB/USD exchange rate (denoted by ) on the one- CR dt Jager (2011) include interest rate differentials in calculating trading-day lagged composite growth of the currencies in the EMP. However, we do not include interest rate differentials currency basket against the dollar (denoted by ) using AC dt−1 because China rarely utilizes interest rate instruments to a daily dataset from August 11, 2015, to October 30, 2020, intervene in the FX market (Das, 2019; Li et al., 2017). yielding the following regression equation, Moreover, the intervention might not be fully reflected in changes in China’s FX reserves, especially after the com- CR =+ ββ AC + λ , dt 01 dt−1 dt plete abolition of the mandatory FX settlement regime in April 2012. As documented in Li et al. (2017) and Das where a constant term β is included. We average the (2019), and more recently in The Economist (2020), China growth of the last closing price of the currencies against has many tools to intervene in the FX market and prefers to the U.S. dollar with the currencies’ weights. The residual do so by adjusting the central parity rate and other λ captures the unexplained component of . Based CR dt dt Liu 9 Figure 1. Estimated EMP and CCAF. EMP: exchange market pressure, CCAF: counter cyclical adjustment factor, USD base: the U.S. dollar index base, CFETS base: the Chinese currency basket designed by China Foreign Exchange Trade System. on the central parity rate formation regime, we can regard Compared to our EMP index, there are many outliers in λ as the CCAF, which we denote as a CFETS-based the Patnaik et al. (2017)’s measure. Moreover, our EMP dt CCAF. index is less volatile and generally smaller than Patnaik et al. Notably, to prepare AC , we calculate each currency’s (2017)’s. In the following section, we use the CFETS cur- dt−1 daily growth in the basket against the dollar and then average rency basket-based EMP to fulfill our empirical study. the growth by the corresponding weight of each currency. Finally, the average growth is multiplied by the weight of the Proxy for economic uncertainty. As stated in the introduc- dollar. In addition, the weights of the currencies in the basket tion section, we use the news-based EPU index to proxy are adjusted according to the PBoC. economic uncertainty. Motivated by Christou et al. (2018), Additionally, to provide some insights into the reason- Bartsch (2019), and Zhou et al. (2020), U.S. EPU is also ableness of our calculation of the composite growth, we also included. Accordingly, since we focus on the RMB/USD regress on the 1-day lagged growth of the U.S. dollar exchange rate pressure, we use the U.S. EPU constructed CR dt index to obtain a U.S. dollar index-based CCAF. by Baker et al. (2016) and China’s EPU computed by Davis The closing exchange rates are collected from Investing. et al. (2019), downloaded from policyuncertainty.com. com, while the central parity rate and the U.S. dollar index are retrieved from the WIND database. Proxies for market sentiment and market demand. Fol- We estimate the equation using the OLS method and sum lowing Das (2019), we proxy market sentiment with the the daily CCAFs to produce the monthly CCAFs. The result- one-year-ahead forward premium on the exchange rate. ing CCAFs are shown in Figure 1. As we have the CCAF, we Specifically, we compute the log-difference (multiplied compute the monthly EMP index from January 2001 to by 100) between the one-year-ahead non-deliverable for- October 2020, with the (recovered) FX reserves retrieved wards on the RMB/USD exchange rate and its onshore spot from the WIND database. Our calculated EMP indices are counterpart, yielding a market sentiment index that reveals also reported in Figure 1. Moreover, the EMP index com- an expectation of devaluation (appreciation) when it has a puted by Patnaik et al. (2017), downloaded from https://mac- positive (negative) value. rofinance.nipfp.org.in/releases/exchange_market_pressure. To approximate changes in currency demand, as investors html, is also presented for comparison. would trade according to their expectations (Cornell & As shown in Figure 1, the estimated CCAF and the EMP Dietrich, 1978), mainly through banks in China (Lin & index based on the CFETS currency basket are highly con- Schramm, 2003), we use customers’ net sell of foreign cur- sistent with those based on the U.S. dollar index in terms of rency against the RMB calculated by the log-difference magnitude and pattern, where the correlation between the between customers’ sell and purchase of FX through the com- two CCAFs exceeds 0.9. Centering on the estimated CCAF mercial banks. We believe that the overall trading volume yields an intuitively consistent result that the PBoC acts could disclose the changes in market demand of the RMB counter-cyclically when encountering significant exchange against the dollar since trading on this pair accounts for more market pressure. than 95% of the overall trading volume on average. 10 SAGE Open Table 1. Unit Root Tests. ADF without breaks ADF with breaks Test statistics Lag order Test statistics Allowed breaks EMP −9.476** 1 −7.158*** 1 U.S. EPU −6.141*** 0 −8.799*** 1 China’s EPU −3.036 3 −9.055*** 4 Market sentiment −3.103 0 −4.38** 3 Market demand −3.855** 1 −3.969*** 1 Note. *** and ** denote rejection of the null hypothesis at the 1% and 5% significance levels, respectively. Intercept and trend are included. The optimal lag is determined by the BIC. are stationary regardless of whether structural breaks are considered. While the conventional ADF test detects that China’s EPU and market sentiment are non-stationary, the ADF test incorporating breaks signifies a high significance of stationarity of these two variables. Thus, all variables can be viewed as statistically stationary in a general sense. Moreover, this finding highlights the importance of applying the TVP-VAR model that incorporates structural breaks and regime shifts. We set a lag length of six to ensure no serial autocorrela- tion and estimate the model by the procedure documented in Eisenstat et al. (2016). The MCMC Gibbs sampler in Eisenstat et al. (2016) is applied to attain 45,000 replicates with the first 15,000 draws as burn-in, obtaining 3,000 effec- tive posterior draws by recording every ten replicates. Our estimation incorporates three potentially beneficial Generalized Gibbs steps, which can produce the lowest inef- ficiency factors plotted in Figure 2. Clearly, most of the inef- Figure 2. Inefficiency factors of the estimated parameters. ficiency factors are less than 10, though the majority for is slightly greater than 10, indicating the good performance of Finally, all the data, except for the EPU measures, are our MCMC sampler. retrieved from the WIND database and seasonally adjusted (if Next, based on the effective MCMC draws, we estimate necessary), while the EPU measures were taken logarithm. the impulse response functions using the Cholesky decom- position to identify structural shocks. Notably, all impulse Results response functions (IRFs) are rescaled to match the original series. To evaluate the total impact of EPU, we compute Baseline Model cumulative impulse response functions at each time point of First of all, we perform unit root tests to elucidate some sta- 1% shocks in U.S. EPU and China’s EPU, respectively. tistical properties of our variables. Following Jebabli et al. We perform a preliminary analysis with the time-averaged (2014), we employ the conventional ADF unit root testing IRFs for a horizon of 12 months to provide an overall picture approach to validate the stationarity of our variables. of the EPU’s impact, see Figure 3: The 68% highest posterior However, the conventional ADF test may be underpowered, density (HPD) intervals (shaded areas) are constructed with given that the macroeconomic indicators may have multiple 16% and 84% percentiles of the posterior estimates, while potential structural breaks (Check & Piger, 2021). Therefore, the estimated IRFs (solid lines) are the posterior medians. As similar to Nasir (2021), we further apply the GLS-based expected, a rising shock in U.S. EPU would generally ADF test proposed by Carrion-i-Silvestre et al. (2009), which increase the relative demand for the renminbi against the dol- allows for multiple structural breaks in both the null and lar, exerting appreciation pressure on the RMB. However, alternative hypotheses. The corresponding results of unit this market demand channel does not hold for the shocks root testing are reported in Table 1. originating from China’s EPU that it has a positive but impre- As shown in Table 1, allowing for structural breaks in the cise effect on market demand. The counter-intuitive results variables improves the significance of the ADF test. It is sug- are also presented in the response of market sentiment. While gested that EMP, U.S. EPU, and market demand it exhibits a slightly negative response following a hike in Liu 11 Figure 3. Time-averaged dynamics of market sentiment, market demand, and EMP, following a 1% increase in U.S. EPU and China’s EPU, respectively. U.S. EPU, market sentiment responds with a more signifi- time-varying market responses at these horizons after a 1% cant negative pattern after an upsurge in China’s EPU, signi- shock in U.S. EPU and China’s EPU, respectively. fying that a higher China’s EPU may be associated with the At first glance, the IRFs in Figures 3 and 4 provide strong expectations of RMB appreciation and increased demand for evidence of the time-varying impact of EPU on China’s FX the RMB. Notably, this finding is likely relevant to our proxy market, particularly on EMP. Relatively, the IRFs are more for market sentiment, namely the spread between the one- volatile at longer horizons. One possible explanation is that year-ahead NDF exchange rate and the onshore spot since China’s FX market may not be fully efficient (Gupta & exchange rate. As the NDF exchange rate is the expected Plakandaras, 2019), the market may need time to recognize exchange rate 1 year later, if the uncertainty shocks cause the the shock, waiting to see the impact it will cause and respond- spot exchange rate to depreciate in the short run, the market ing accordingly afterward. Consequentially, the IRFs at the would expect the exchange rate to revert, inducing apprecia- horizons of 0-month and 1-month are more stable and less tion expectations. In addition, China’s asymmetric FX regu- precise, indicating market inefficiency. However, the IRFs at lations, where selling FX is less controlled than buying, may longer horizons generally depend on market conditions, such also contribute to the preposterous response in market as institutional and market regimes, which have dramatically demand. Nevertheless, an increase in China’s EPU is still changed in China over the past decades, resulting in the high followed by an approximate positive response in EMP, cor- volatility of these IRFs. Therefore, we scrutinize and discuss roborating roughly the findings in Liu (2020). the rationale of the time variability revealed by these IRFs in To dissect how the impact of EPU evolves over our sam- the following section. ple period, we turn to an analysis based on the time-varying As shown in the top panels in Figures 4 and 5, an increase IRFs. Since the FX market operates daily, it is unlikely that in either U.S. EPU or China’s EPU would trigger apprecia- the market consumes a long time to respond to an external tion sentiment on the renminbi against the dollar, which is shock. Accordingly, we accentuate our analysis on inspect- consistent with the findings by Li et al. (2020). Moreover, ing the market’s reactions in 3 months after the shock. the response of market sentiment to a U.S. EPU shock is con- Further, we detect the responses at 6-month after the shock as sistent with the common wisdom that a rise in U.S. EPU has the time-averaged IRFs shown in Figure 3 almost all con- a detrimental impact on the U.S. macroeconomy and pro- verge after then. More specifically, we look into the impulse vokes devaluation expectations on the dollar. However, the response functions at each time point at horizons of 0 to response to the China’s EPU shocks contradicts the uncer- 3 months and 6 months after the shock, where 0-month means tainty theory. As noted earlier, this could be explained by the moment when the shock occurs. Figures 4 and 5 paint the China’s FX regulations, which might cause the market to 12 SAGE Open Figure 4. Time-varying market responses at horizons of 0-month to 3-month, and 6 months following a 1% increase in U.S. EPU. The solid lines are posterior medians, while the shaded regions are the corresponding 68% HPD intervals. Figure 5. Time-varying market responses at horizons of 0 to 3, and 6 months following a 1% increase in China’s EPU. The solid lines are posterior medians, while the shaded regions are the corresponding 68% HPD intervals. Liu 13 expect the renminbi to appreciate against the dollar. Moreover, the IRFs at horizons of 2, 3, and 6 months after the Relatively, the response to China’s EPU shocks is more sta- shock exhibit noticeable time variations. Specifically, the ble than that to U.S. EPU shocks, though the latter’s corre- impact of EPU on EMP is stable until mid-2011 but then sponding HPD intervals are wide to contain zero, indicating enlarges till the end of 2017. Accordingly, our results provide insignificant responses of market sentiment to the U.S. EPU little evidence to support the findings by Kido (2016), who shocks. By contrast, following a hike in China’s EPU, there finds that the real effective exchange rate returns and U.S. would be a significant appreciation expectation on the ren- EPU were intensively correlated during the global financial minbi against the dollar from 3 months onward, especially in crisis. The explanation is straightforward. Albeit, indeed, the the post-2005 period. Before 2005, China carried out a de unconventional monetary policy undertaken by the U.S. facto pegged exchange rate regime, causing market senti- Federal Reserve during the financial crisis reduced the dol- ment to be less responsive to China’s EPU shocks. However, lar’s value (Neely, 2015), China had pegged the renminbi to following the reform in July 2005, China implemented a the dollar again during this period. Further, China had experi- gradual appreciation path for the renminbi against the dollar, enced large capital outflows during the crisis (Broner et al., leading to the widespread market expectations of the RMB 2013) due to the need to relieve the value-at-risk of U.S. appreciation. In addition, the exchange rate stability has been domestic assets caused by the crisis (Schmidt & Zwick, one of China’s principal objectives. Thus, the negative build- 2015). Thus, the impact of U.S. EPU on China’s EMP could ing impact of China’s EPU on the renminbi would stimulate retain its past pattern without being altered by the crisis. an expectation that the government will intervene in the mar- Intuitively, this time-varying property shown in the impact ket, inducing the appreciation expectations on the renminbi. of U.S. EPU on China’s EMP may be associated with the Further, the response of market sentiment to the U.S. EPU evolution of China’s dependence on the United States. Over shocks swells progressively after mid-2015, which could be the last decades, China has developed a high degree of trade ascribed to the increased exchange rate flexibility after intro- and financial linkages with the United States. Taking the for- ducing the quoting regime of the central parity rate in August eign trade as an example, Figure 6 displays the geographical 2015 (Das, 2019). Gini coefficient of China’s international trade and the share It is noteworthy that market demand expresses a signifi- of trade with the United States in China’s total trade from cant response following a rise in U.S. EPU, as shown in the 1992 to 2019. Generally, the dependence on the U.S. is high, middle panel in Figure 4. In the first month after the shock, especially for exports, although the geographical concentra- the relative demand for the renminbi increases significantly, tion of China’s trade is relatively low. The overall depen- mirroring the dollar devaluation expectations caused by the dence on the U.S. has declined gradually, but slightly, from U.S. EPU shocks. By contrast, market demand responds 2001 to 2011, resulting in a relatively small and stable impact counter-intuitively and insignificantly to China’s EPU of U.S. EPU during this period. Subsequently, from 2012 to shocks. As displayed in Figure 5, after a rising shock in 2017, China’s dependence on the U.S. has been fueled up, China’s EPU, we can see an increase in market demand for associated with the strengthening impact of U.S. EPU on the renminbi in the first month, but it shows a preference for China’s EMP. However, since 2018, the trade dependence the dollar 2 months after the shock. Therefore, it may take has plummeted because of the U.S.-China trade dispute, about 2 months for China’s EPU shocks to have the supposed making China’s EMP less responsive to the U.S. EPU shocks. impact on market demand. Notwithstanding, China’s EPU In addition, the recently heightened response of EMP to the exerts a less precise effect on market demand when com- U.S. EPU shocks may be attributed to the easing of the trade pared with U.S. EPU. One explanation for this is that China’s dispute and the outbreak of COVID-19. FX regulations make it much easier to sell foreign currencies As shown in Figures 4 and 5, it is discernible that the than to buy them. Importantly, as shown in Figure 5, the responses of China’s EMP to the shocks originating from response also has a downward trend after mid-2015, proba- domestic EPU and U.S. EPU are similar in dynamics but bly due to an increasingly flexible exchange rate regime. opposite in evolutionary path. Since EMP could be influ- This result supports the findings by Kozhan and Salmon enced by local and external uncertainties concurrently, we (2009) that agents in the foreign exchange market are uncer- find a trade-off mechanism between them, where the impact tainty averse. of domestic EPU prevails when the impact of foreign EPU While the responses of market sentiment and market fades and vice visa, corroborating the effect of the EPU dif- demand to China’s EPU shocks are somewhat at odds with ferentials on the exchange rate reported in Balcilar et al. the theory, the response of EMP is consistent with our expec- (2016), Christou et al. (2018), and Zhou et al. (2020). tations. As shown in the last panels in Figures 4 and 5, respec- Further, the time-varying nature of EMP response to the tively, a hike in U.S. EPU (China’s EPU) would systematically EPU shocks may also be related to the transmission through prompt appreciating (deprecating) pressure on the renminbi market sentiment and demand. To explore the roles of market against the dollar, leading to a negative (positive) response of sentiment and demand, we estimate the time-varying IRFs of EMP. Comparatively, China’s EPU shocks take a longer time EMP to the shocks in these two variables and plot the results than U.S. EPU shocks to generate an impact on China’s EMP. in Figure 7. 14 SAGE Open Figure 6. Gini coefficients of China’s trade and trade dependence on the United States. Following Liu et al. (2020), we measure the geographical concentration of trade with the Gini coefficients, calculated by , where X is the trade with country i in GX = / X it ti () tt denotes aggregate trade in year t. The trade data are retrieved from the UN COMTRADE database: (a) Gini coefficients year t, and X and (b) fractions of US in Chinese foreign trade. Figure 7. Time-varying responses of EMP to 1% shocks in market sentiment and market demand, respectively. The solid lines are posterior medians, while the shaded regions denote the corresponding 68% HPD intervals. Obviously, more time variation is detected in EMP would trigger appreciation pressure on the RMB. responses to the market sentiment and demand shocks. An Additionally, after August 2015, increasingly enhanced increase in market sentiment (i.e., devaluation expectations exchange rate flexibility raises investors’ risk exposure, on the renminbi against the dollar) would substantially causing the impact of market sentiment and market demand coerce the RMB to devaluate, while a rise in market demand to reduce sharply. Liu 15 Figure 8. Time-averaged responses of EMP to 1% shocks to U.S. EPU and China’s EPU. The solid lines are posterior medians, while the shaded regions are the 68% HPD intervals. As we can see in Figures 8 and 9, Rmodel1 and Rmodel2 Robustness Check produce similar results to our baseline model. However, We consider the following two alternative specifications to Rmodel1 derives a more precise estimate of the EMP implement robustness checks. (1) [Rmodel1] The EPU mea- response to China’s EPU shocks, particularly after 2010, sure we used for China is constructed based on information during which an upsurge in China’s EPU would cause sig- from only two newspapers that may be subject to media bias, nificant devaluation pressure on the onshore exchange rate. so we instead use the EPU measure for China constructed by This result may suggest that the EPU index constructed by Huang and Luk (2020), which is based on 10 newspapers Huang and Luk (2020) is a more credible indicator of eco- pressed in mainland China. (2) [Rmodel2] Since the funda- nomic uncertainty as it is less prone to media bias (Huang & mental price of the exchange rate is theoretically determined Luk, 2020). In addition, Rmodel2 renders a more remarkable by the macroeconomy, we include China’s leading macro- but more stable response of EMP to U.S. EPU shocks. economic indicators in the model, namely, economic growth (measured by real GDP growth), inflation (proxied by CPI Conclusion and Policy Implications growth), short-term interest rate (using the 7-day reverse repo rate) and money growth (measured with M2 growth). Our empirical findings show that an upsurge in U.S. eco- All the data are taken from Chang et al. (2016)’s China’s nomic policy uncertainty triggers appreciation pressures on Macroeconomy Database. All these variables are shown as the RMB/USD exchange rate, while a hike in China’s eco- stationary by the ADF unit root tests with and without struc- nomic policy uncertainty tends to exert devaluation pressures tural breaks. The estimation procedure for both models is the on the exchange rate. Moreover, the magnitude of the impact same as that for the baseline model. Here, we concentrate on of economic policy uncertainty on exchange market pressure the impact of unexpected rises in U.S. EPU and China’s EPU is time-dependent. Specifically, the impact is relatively stable on China’s EMP. Figures 8 and 9 show the time-averaged until mid-2011 but more volatile thereafter. In addition, we IRFs and the time-varying IRFs predicted by the two models, find a trade-off mechanism between the impact on EMP of respectively. domestic and foreign economic policy uncertainties. 16 SAGE Open Figure 9. Time-varying responses of EMP to 1% shocks in China’s EPU and U.S. EPU, respectively. The solid lines are posterior medians, while the shaded regions denote the 68% HPD intervals. Our results have the following policy implications. First, However, our study still has some limitations. First, to stabilize the exchange rate, China should enhance its man- because of the lack of statistics on market agents, our proxy agement of market expectations and market demand, reduc- for market sentiment in the empirical section is not entirely ing the unfavorable impact on the exchange rate of irrational consistent with that illustrated in the theoretical model, market sentiment to cope with spikes in domestic and foreign resulting in a rough understanding of market sentiment in a economic uncertainties. In addition, a foreign exchange broad sense. Second, our proxy for market demand is not a derivatives market should be established and advanced to pure index reflecting demand for the RMB/USD pair but also provide functional risk hedging tools and attain a better includes other trading pairs, despite the dominance of the understanding of market sentiment. Second, further liberal- RMB/USD pair in the market. Third, since China still adopts ization of the capital account should be proceeded prudently a managed floating exchange rate regime, we note that and sequenced with prerequisite reforms, including a solid China’s monetary authority releases scant interventions in and sound financial system and an appropriate macroeco- the foreign exchange market and intervenes the market by nomic regulatory framework. Third, China should further other hidden levers recently keeping the reserves unmoved, improve the flexibility of the RMB exchange rate and make limiting us to obtain a highly accurate estimate of China’s the exchange rate formation mechanism transparent, sending exchange market pressure. clear signals of two-way exchange rate fluctuations to avoid As for the future research, since capital flows are sensitive irrational market sentiment toward one-way movements. to economic uncertainty and exchange rate fluctuations, one Finally, China should further diversify the basket currency possible direction is to dissect the roles of capital flows in the structure and increase RMB transactions against non-USD relationships between economic uncertainty and exchange currencies. Furthermore, China could lower its dependence market pressure. Moreover, in recent years, China’s authority on the United States in trade and finance to a certain extent to has attempted to manage the foreign exchange market by reduce the detrimental impact on China’s real economy and adjusting the foreign exchange reserve requirement ratio, open- financial markets of economic uncertainty shocks originat- ing a topic that how and whether this policy affects market sen- ing from the United States. timent and market demand, and thereby the exchange rates. Liu 17 Declaration of Conflicting Interests Chang, C., Chen, K., Waggoner, D. F., & Zha, T. (2016). Trends and cycles in China’s macroeconomy. NBER Macroeconomics The author declared no potential conflicts of interest with respect to Annual, 30(1), 1–84. https://doi.org/10.1086/685949 the research, authorship, and/or publication of this article. Chan, J. C. C., Eisenstat, E., & Strachan, R. W. (2020). Reducing the state space dimension in a large TVP-VAR. Journal of Funding Econometrics, 218(1), 105–118. https://doi.org/10.1016/j.jeco- The author received no financial support for the research, author- nom.2019.11.006 ship, and/or publication of this article. Check, A., & Piger, J. (2021). Structural breaks in U.S. Macroeconomic time series: A Bayesian model averaging approach. Journal of Money Credit and Banking, forthcoming ORCID iD 53, 1999–2036. https://doi.org/10.1111/jmcb.12822 Lin Liu https://orcid.org/0000-0003-1258-1730 Chen, L., Du, Z., & Hu, Z. (2020). Impact of economic policy uncer- tainty on exchange rate volatility of China. Finance Research References Letters, 32, 101266. https://doi.org/10.1016/j.frl.2019.08.014 Aizenman, J., & Binici, M. (2016). Exchange market pressure in Chiarella, C., He, X.-Z., Shi, L., & Wei, L. (2017). A behavioural OECD and emerging economies: Domestic vs. External fac- model of investor sentiment in limit order markets. Quantitative tors and capital flows in the old and new normal. Journal of Finance, 17(1), 71–86. https://doi.org/10.1080/14697688.2016 International Money and Finance, 66, 65–87. https://doi. .1184756 org/10.1016/j.jimonfin.2015.12.008 Choi, M. J., Kim, G.-Y., & Lee, J. Y. (2018). An analysis of trade Bachmann, R., Elstner, S., & Sims, E. R. (2013). Uncertainty patterns and the effects of the real exchange rate movements and economic activity: Evidence from business survey data. in East Asia. Emerging Markets Finance and Trade, 54(7), American Economic Journal Macroeconomics, 5(2), 217–249. 1635–1652. https://doi.org/10.1080/1540496x.2017.1316712 https://doi.org/10.1257/mac.5.2.217 Christou, C., Gupta, R., Hassapis, C., & Suleman, T. (2018). The Baker, S. R., Bloom, N., & Davis, S. J. (2016). Measuring eco- role of economic uncertainty in forecasting exchange rate nomic policy uncertainty. The Quarterly Journal of Economics, returns and realized volatility: Evidence from quantile pre- 131(4), 1593–1636. https://doi.org/10.1093/qje/qjw024 dictive regressions. Journal of Forecasting, 37(7), 705–719. Balcilar, M., Gupta, R., Kyei, C., & Wohar, M. E. (2016). Does https://doi.org/10.1002/for.2539 economic policy uncertainty predict exchange rate returns and Cornell, W. B., & Dietrich, J. K. (1978). The efficiency of the mar- volatility? Evidence from a nonparametric causality-in-quan- ket for foreign exchange under floating exchange rates. The tiles test. Open Economies Review, 27(2), 229–250. https://doi. Review of Economics and Statistics, 60(1), 111. https://doi. org/10.1007/s11079-016-9388-x org/10.2307/1924339 Bartsch, Z. (2019). Economic policy uncertainty and dollar-pound Das, M. S. (2019). China’s evolving exchange rate regime. IMF exchange rate return volatility. Journal of International Working Paper No.19/50. https://www.imf.org/-/media/Files/ Money and Finance, 98, 17. https://doi.org/10.1016/j.jimon- Publications/WP/2019/WPIEA2019050.ashx fin.2019.102067 Davis, S. J., Liu, D., & Sheng, X. S. (2019). Economic policy Baumeister, C., & Peersman, G. (2013). Time-varying effects of uncertainty in china since 1946: The view from mainland oil supply shocks on the US economy. American Economic newspapers. Working Paper. https://stevenjdavis.com/s/EPU- Journal Macroeconomics, 5(4), 1–28. https://doi.org/10.1257/ in-China-View-from-Mainland-Newspapers-August-2019. mac.5.4.1 pdf Beckmann, J., & Czudaj, R. (2017a). Exchange rate expecta- De Grauwe, P., & Ji, Y. (2020). Structural reforms, animal spir- tions and economic policy uncertainty. European Journal of its, and monetary policies. European Economic Review, 124, Political Economy, 47, 148–162. https://doi.org/10.1016/j. 103395. https://doi.org/10.1016/j.euroecorev.2020.103395 ejpoleco.2016.06.003 de Jong, E., Verschoor, W. F. C., & Zwinkels, R. C. J. (2010). Beckmann, J., & Czudaj, R. (2017b). The impact of uncertainty on Heterogeneity of agents and exchange rate dynamics: professional exchange rate forecasts. Journal of International Evidence from the EMs. Journal of International Money and Money and Finance, 73, 296–316. https://doi.org/10.1016/j. Finance, 29(8), 1652–1669. https://doi.org/10.1016/j.jimon- jimonfin.2017.02.009 fin.2010.05.007 Bloom, N. (2009). The impact of uncertainty shocks. Econometrica, Dieci, R., & Westerhoff, F. (2010). Heterogeneous speculators, 77(3), 623–685. https://doi.org/10.3982/Ecta6248 endogenous fluctuations and interacting markets: A model of Bloom, N. (2014). Fluctuations in uncertainty. Journal of Economic stock prices and exchange rates. Journal of Economic Dynamics Perspectives, 28(2), 153–176. https://doi.org/10.1257/jep. and Control, 34(4), 743–764. https://doi.org/10.1016/j. 28.2.153 jedc.2009.11.002 Broner, F., Didier, T., Erce, A., & Schmukler, S. L. (2013). Gross cap- Dieci, R., & Westerhoff, F. (2013). On the inherent instability of ital flows: Dynamics and crises. Journal of Monetary Economics, international financial markets: Natural nonlinear interac- 60(1), 113–133. https://doi.org/10.1016/j.jmoneco.2012.12.004 tions between stock and foreign exchange markets. Applied Carrion-i-Silvestre, J. L., Kim, D., & Perron, P. (2009). GLS-based Mathematics and Computation, 221, 306–328. https://doi. unit root tests with multiple structural breaks under both the org/10.1016/j.amc.2013.06.042 null and the alternative hypotheses. Economic Theory, 25(6), Eisenstat, E., Chan, J. C. C., & Strachan, R. W. (2016). Stochastic 1754–1792. https://doi.org/10.1017/s0266466609990326 model specification search for time-varying parameter VARs. 18 SAGE Open Econometric Reviews, 35(8–10), 1638–1665. https://doi.org/10 stochastic volatility. Energy Economics, 45, 66–98. https:// .1080/07474938.2015.1092808 doi.org/10.1016/j.eneco.2014.06.008 Fernald, J. G., Spiegel, M. M., & Swanson, E. T. (2014). Monetary Jurado, K., Ludvigson, S. C., & Ng, S. (2015). Measuring uncer- policy effectiveness in China: Evidence from a FAVAR model. tainty. American Economic Review, 105(3), 1177–1216. Journal of International Money and Finance, 49, 83–103. https://doi.org/10.1257/aer.20131193 https://doi.org/10.1016/j.jimonfin.2014.05.007 Kido, Y. (2016). On the link between the US economic policy Fernández-Villaverde, J., Guerrón-Quintana, P., & Rubio-Ramírez, uncertainty and exchange rates. Economics Letters, 144, 49– J. F. (2015). Estimating dynamic equilibrium models with sto- 52. https://doi.org/10.1016/j.econlet.2016.04.022 chastic volatility. Journal of Econometrics, 185(1), 216–229. Klaassen, F., & Jager, H. (2011). Definition-consistent measure- https://doi.org/10.1016/j.jeconom.2014.08.010 ment of exchange market pressure. Journal of International Frühwirth-Schnatter, S., & Wagner, H. (2010). Stochastic model Money and Finance, 30(1), 74–95. https://doi.org/10.1016/j. specification search for Gaussian and partial non-Gaussian jimonfin.2010.07.003 state space models. Journal of Econometrics, 154(1), 85–100. Koop, G., & Korobilis, D. (2013). Large time-varying parameter https://doi.org/10.1016/j.jeconom.2009.07.003 VARs. Journal of Econometrics, 177(2), 185–198. https://doi. Girton, L., & Roper, D. (1977). A monetary model of exchange org/10.1016/j.jeconom.2013.04.007 market pressure applied to the postwar Canadian experience. Kozhan, R., & Salmon, M. (2009). Uncertainty aversion in a het- American Economic Review, 67(4), 537–548. https://www. erogeneous agent model of foreign exchange rate formation. jstor.org/stable/1813387 Journal of Economic Dynamics and Control, 33(5), 1106– Granger, C. W. J. (2008). Non-linear models: Where do we go 1122. https://doi.org/10.1016/j.jedc.2008.11.008 next - time varying parameter models? Studies in Nonlinear Krol, R. (2014). Economic policy uncertainty and exchange rate Dynamics and Econometrics, 12(3), 1639. https://doi.org/10. volatility. International Finance, 17(2), 241–256. https://doi. 2202/1558-3708 org/10.1111/infi.12049 Gupta, R., Olasehinde-Williams, G., & Wohar, M. E. (2020). The Lengnick, M., & Wohltmann, H.-W. (2013). Agent-based financial impact of US uncertainty shocks on a panel of advanced and markets and new Keynesian macroeconomics: A synthesis. emerging market economies. Journal of International Trade & Journal of Economic Interaction and Coordination, 8(1), 1–32. Economic Development, 29(6), 711–721. https://doi.org/10.10 https://doi.org/10.1007/s11403-012-0100-y 80/09638199.2020.1720785 Li, H., Yu, Z., Zhang, C., & Zhang, Z. (2017). Determination of Gupta, R., & Plakandaras, V. (2019). Efficiency in BRICS cur- china’s foreign exchange intervention: Evidence from the rency markets using long-spans of data: Evidence from model- yuan/dollar market. Studies in Economics and Finance, 34(1), free tests of directional predictability. Journal of Economics 62–81. https://doi.org/10.1108/sef-10-2015-0249 and Behavioral Studies, 11(1(J)), 152–165. https://doi. Lin, G., & Schramm, R. M. (2003). China’s foreign exchange org/10.22610/jebs.v11i1(j).2756 policies since 1979: A review of developments and an assess- Huang, Y., & Luk, P. (2020). Measuring economic policy uncer- ment. China Economic Review, 14(3), 246–280. https://doi. tainty in China. China Economic Review, 59, 101367. https:// org/10.1016/j.chieco.2003.09.006 doi.org/10.1016/j.chieco.2019.101367 Liu, J. (2020). Impact of uncertainty on foreign exchange market Huber, F., Kastner, G., & Feldkircher, M. (2019). Should I stay or stability: Based on the LT-TVP-VAR model. China Finance should I go? A latent threshold approach to large-scale mix- Review International, 11(1), 53–72. https://doi.org/10.1108/ ture innovation models. Journal of Applied Economics, 34(5), cfri-07-2019-0112 621–640. https://doi.org/10.1002/jae.2680 Liu, L. (2021). U.S. Economic uncertainty shocks and china’s Huber, F., Koop, G., & Onorante, L. (2021). Inducing sparsity economic activities: A time-varying perspective. Sage Open, and shrinkage in time-varying parameter models. Journal of 11(3), 21582440211032672. https://doi.org/10.1177/215824 Business and Economic Statistics, 39(3), 669–683. https://doi. 40211032672 org/10.1080/07350015.2020.1713796 Liu, L., Zhang, M., & Li, W. (2020). China’s economic policy Hu, X. (1997). Macroeconomic uncertainty and the risk premium in uncertainty shocks and South Korea’s exports: A TVP-VAR the foreign exchange market. Journal of International Money approach with an SMSS structure. Journal of Korea Trade, and Finance, 16(5), 699–718. https://doi.org/10.1016/s0261- 24(4), 1–17. https://doi.org/10.35611/jkt.2020.24.4.1 5606(97)00027-2 Li, X.-L., Li, X., & Si, D.-K. (2020). Investigating asymmetric Huynh, T. L., Nasir, M. A., & Nguyen, D. K. (2020). Spillovers and determinants of the CNY–CNH exchange rate spreads: The connectedness in foreign exchange markets: The role of trade role of economic policy uncertainty. Economics Letters, 186, policy uncertainty. The Quarterly Review of Economics and 4. https://doi.org/10.1016/j.econlet.2019.108827 Finance. Advance online publication. https://doi.org/10.1016/j. Lubik, T., & Matthes, C. (2016). Time-varying parameter vec- qref.2020.09.001 tor autoregressions: Specification, estimation, and an appli- Imbs, J., Mumtaz, H., Ravn, M. O., & Rey, H. (2003). cation. Economic Quarterly, 101(04), 323–352. https://doi. Nonlinearities and real exchange rate dynamics. Journal of the org/10.21144/eq1010403 European Economic Association, 1(2-3), 639–649. https://doi. Mahajan, A., & Wagner, A. J. (1999). Nonlinear dynamics in for- org/10.1162/154247603322391279 eign exchange rates. Global Finance Journal, 10(1), 1–23. Jebabli, I., Arouri, M., & Teulon, F. (2014). On the effects of https://doi.org/10.1016/s1044-0283(99)00002-2 world stock market and oil price shocks on food prices: An Mumtaz, H., & Theodoridis, K. (2018). The changing transmis- empirical investigation based on TVP-VAR models with sion of uncertainty shocks in the U.S. Journal of Business and Liu 19 Economic Statistics, 36(2), 239–252. https://doi.org/10.1080/0 Olasehinde-Williams, G., & Olanipekun, I. (2020). Unveiling the 7350015.2016.1147357 causal impact of US economic policy uncertainty on exchange Mussa, M. (1984). The theory of exchange rate determination. In market pressure of African economies. Journal of Public F. B. John & C. M. Richard (Eds.), Exchange rate theory and Affairs. Advance online publication. https://doi.org/10.1002/ practice (pp. 13–78). University of Chicago Press. pa.2278 Nakagawa, H. (2010). Investigating nonlinearities in real exchange Pástor, & Veronesi, P. (2013). Political uncertainty and risk premia. rate adjustment: Threshold cointegration and the dynamics of Journal of Financial Economics, 110(3), 520–545. https://doi. exchange rates and relative prices. Journal of International org/10.1016/j.jfineco.2013.08.007 Money and Finance, 29(5), 770–790. https://doi.org/10.1016/j. Patnaik, I., Felman, J., & Shah, A. (2017). An exchange mar- jimonfin.2010.03.002 ket pressure measure for cross country analysis. Journal of Nasir, M. A., Naidoo, M. A., Shahbaz, L., & Amoo, L. M. N. (2018). International Money and Finance, 73, 62–77. https://doi. Implications of oil prices shocks for the major emerging econo- org/10.1016/j.jimonfin.2017.02.004 mies: A comparative analysis of BRICS. Energy Economics, Qin, B., Strömberg, D., & Wu, Y. (2018). Media bias in China. 76, 76–88. https://doi.org/10.1016/j.eneco.2018.09.023 American Economic Review, 108(9), 2442–2476. https://doi. Nasir, M. A. (2020). Forecasting inflation under uncertainty: The org/10.1257/aer.20170947 forgotten dog and the frisbee. Technological Forecasting and Schmidt, T., & Zwick, L. (2015). Uncertainty and episodes of Social Change, 158, 120172. https://doi.org/10.1016/j.tech- extreme capital flows in the euro area. Economic Modelling, fore.2020.120172 48, 343–356. https://doi.org/10.1016/j.econmod.2014.10.047 Nasir, M. A. (2021). Zero lower bound and negative interest rates: Sengupta, J. K., & Sfeir, R. E. (1998). Nonlinear dynamics in Choices for monetary policy in the UK. Journal of Policy foreign exchange markets. International Journal of Systems Modeling, 43(1), 200–229. https://doi.org/10.1016/j.jpol- Science, 29(11), 1213–1224. https://doi.org/10.1080/0020 mod.2020.03.014 7729808929610 Nasir, M. A., & Morgan, J. (2018). Pre-Brexit: The EU referendum Shin, M., Zhang, B., Zhong, M., & Lee, D. J. (2018). Measuring as an illustration of the effects of uncertainty on the sterling international uncertainty: The case of Korea. Economics exchange rate. Journal of Economic Studies, 45(5), 910–921. Letters, 162, 22–26. https://doi.org/10.1016/j.econlet.2017. https://doi.org/10.1108/jes-07-2017-0205 10.014 Nasir, M. A., & Simpson, J. (2018). Brexit associated sharp depre- Simo-Kengne, B. D., Ababio, K. A., Mba, J., Koumba, U., & ciation and implications for UK’s inflation and balance of pay- Molepo, M. (2018). Risk, uncertainty and exchange rate behav- ments. Journal of Economic Studies, 45(2), 231–246. https:// ior in South Africa. Journal of African Business, 19(2), 262– doi.org/10.1108/jes-02-2017-0051 278. https://doi.org/10.1080/15228916.2017.1413627 Nasir, M. A., & Vo, X. V. (2020). A quarter century of inflation Taylor, M. P. (1989). Expectations, risk and uncertainty in targeting & structural change in exchange rate pass-through: the foreign exchange market: Some results based on sur- Evidence from the first three movers. Structural Change and vey data. Manchester School, 57(2), 142–153. https://doi. Economic Dynamics, 54, 42–61. https://doi.org/10.1016/j. org/10.1111/j.1467-9957.1989.tb00807.x strueco.2020.03.010 Ter Ellen, S., Verschoor, W. F. C., & Zwinkels, R. C. J. (2013). Neely, C. J. (2015). Unconventional monetary policy had large Dynamic expectation formation in the foreign exchange mar- international effects. Journal of Banking & Finance, 52, 101– ket. Journal of International Money and Finance, 37, 75–97. 111. https://doi.org/10.1016/j.jbankfin.2014.11.019 https://doi.org/10.1016/j.jimonfin.2013.06.001 Nilavongse, R., Rubaszek, M., & Uddin, G. S. (2020). Economic Tervala, J. (2019). U.S. Monetary policy and China’s exchange policy uncertainty shocks, economic activity, and exchange rate policy during the great recession. International Journal rate adjustments. Economics Letters, 186, 4. https://doi. of Finance & Economics, 24(1), 113–130. https://doi. org/10.1016/j.econlet.2019.108765 org/10.1002/ijfe.1652 Olanipekun, I. O., Güngör, H., & Olasehinde-Williams, G. (2019a). The Economist. (2020). Is China suppressing the yuan?, October Unraveling the causal relationship between economic policy 31st–November 6th 2020. uncertainty and exchange market pressure in BRIC countries: Wang, Y., Li, X., Huang, D., & Wang, A. (2020). Is there evidence Evidence from bootstrap panel granger causality. Sage Open, of “fear of appreciation” in China’s foreign exchange market 9(2), 13. https://doi.org/10.1177/2158244019853903 intervention? Applied Economics, 52(6), 537–551. https://doi. Olanipekun, I. O., Olasehinde-Williams, G., & Güngör, org/10.1080/00036846.2019.1646878 H. (2019b). Impact of economic policy uncertainty on Zheng, L., Jiang, Y., & Long, H. (2019). Exchange rates change, exchange market pressure. Sage Open, 9(3), 13. https://doi. asset-denominated currency difference and stock price fluctua- org/10.1177/2158244019876275 tion. Applied Economics, 51(60), 6517–6534. https://doi.org/1 Olanipekun, I. O., & Olasehinde-Williams, G. (2021). Unveiling the 0.1080/00036846.2019.1624920 causal interactions among the US monetary policy uncertainty, Zhou, Z., Fu, Z., Jiang, Y., Zeng, X., & Lin, L. (2020). Can eco- exchange market pressure, and FDI-flows in large emerging nomic policy uncertainty predict exchange rate volatility? New market economies. Journal of Public Affairs. Advance online evidence from the GARCH-MIDAS model. Finance Research publication. https://doi.org/10.1002/pa.2684 Letters, 34, 101258. https://doi.org/10.1016/j.frl.2019.08.006

Journal

SAGE OpenSAGE

Published: Jan 18, 2022

Keywords: economic Uncertainty; EMP; EPU; TVP-VAR; Over-parameterization

There are no references for this article.