The Dynamic Effects of Oil Price Shocks on Exchange Rates—From a Time-Varying Perspective

Abstract

This study investigates the time-varying effects of three types of oil price shocks (oil demand, supply, and risk shock) on exchange rates by applying the time-varying parameter structural vector autoregression stochastic volatility (TVP-SVAR-SV) model. Through examining the impulse response of exchange rates to oil price shocks at different lag periods and time points, this paper contributes to the existing literature on the dynamic relationship between oil shocks and exchange rates. From the response at different lag periods, we find that oil price shocks have a significant time-varying impact on the exchange rate, among which oil demand shock has the most significant effect. The response of the exchange rate market to oil price shock shows an obvious short-term time-varying effect and is positive and negative alternately, with a certain periodicity. From the response at different time points, the time-varying effect of oil price shock on exchange rates is related to external shock, and is more intense during the global economic and political turmoil. This is the first empirical study using a novel method to examine the time-varying effects of oil price shock from different sources on exchange rates, providing investors and policy makers assistance to manage foreign exchange during global turmoil periods.

1. Introduction

The oil price has always been considered as a leading indicator of exchange rate movements in the world economy [1,2]. In recent years, with the intense of global trade wars and events such as Brexit and Sino-US trade frictions, the impact of oil prices on exchange rates has become more complex, and the dynamic relationship between the two has become more and more diversified. There are significant country differences and changes in the scope and intensity of influence before and after the crisis [3,4]. Many scholars have begun to pay attention to the dynamic relationship between oil price and the exchange rate market [5], but there is little research on its time-varying characteristics. In this context, exploring the time-varying effect of oil prices on exchange rates for both financial investors and policy makers is of great significance.

The relationship between oil price and exchange rate seems not to be the same over time. Ji et al., (2020) believe that the impact of oil shocks on the real exchange rate increases over time, especially after the global financial crisis (GFC) [6]. César and Rebeca (2020) also found the same conclusion when exploring the dynamic interaction between oil price and US dollar exchange rate, and they also stressed the importance of considering the time frame of oil price shocks [7]. This corresponds to the research results of Wen et al., (2018) to a certain extent [8]. In their opinion, a major shock leading to structural fracture will increase the risk diffusion between the two markets. For example, the impact of the financial crisis, the Iranian nuclear issue, the eurozone debt crisis, and the new round of quantitative easing policy in the United States, etc., formed greater volatility transmission or risk contagion in the crude oil market and exchange rate market. Because the impact of oil prices varies over time, it is important to establish a precise classification of oil price shocks to distinguish them.

According to several authors such as Kilian (2009), Kilian and Murphy (2012), Peersman and Van Robays (2009), and Baumeister and Peersman (2013), the impact of oil price shocks greatly depends on the source of oil price fluctuations [9,10,11,12]. Specifically, oil price shocks affect the exchange rate market through three aspects, the basic supply and demand side and financial speculation. First, due to the rapid growth of the global economy and demand, the oil demand shock has been proved to be an important factor in explaining exchange rate changes. For example, real exchange rates in the BRIC countries have a strong appreciation response to aggregate demand shocks [13], while aggregate global demand and oil-specific demand shocks can explain depreciation in OECD countries [14]. Second, the oil supply shock is closely related to future supply or expected output. Although all the current literature believes that oil supply shock has little explanatory power on exchange rate fluctuations, Askari and Krichene (2010) indicate that oil supply has deficient price elasticity, which will cause violent price fluctuations of oil commodities [15]. Failure to take this into account could easily lead to inflation, further affecting the exchange rate. Cunado et al., (2015) also show that the impact of oil supply shock is marginal [16]. The third aspect is the speculative oil shock. Most literatures tend to study the impact of the speculative oil shock on futures or stock markets [17,18], while there are few literature studies on the impact of the speculative oil shock on exchange rates [19]. In recent years, some people have begun to doubt the effect of speculative shocks on the exchange rate. For example, Zhang et al., (2020) believe that the increase in financialization and speculation in the oil market may lead to a higher correlation between oil prices and exchange rates, but they do not draw specific conclusions [20]. The impact of the speculative oil shock on exchange rate fluctuations will be shown in detail in the follow-up of our article.

Oil affects exchange rates in different theoretical ways by influencing inflation, wealth transfers, investment portfolios and terms of trade. From the perspective of supply and demand, Darby (1982) believed that the fluctuation of oil prices was closely related to the rise of inflation [21]. According to the Fisher effect, interest rates will go up depending on the performance of inflationary pressure, which affects the value of domestic money, which is consistent with the real interest rate parity hypothesis. When oil prices rise, it is expected that the currencies of countries with high oil dependence in the trade sector will depreciate. In addition, based on the investment portfolio and wealth channel [22,23], oil price fluctuation will lead to wealth transfer between oil-importing and oil-exporting countries, which is consistent with the theory of balance of payments. Among them, the wealth channel mainly focuses on the short-term effects, while the portfolio channel assesses the medium-term and long-term effects. From the perspective of speculation, Amano and Van Norden (1998) proposed the trade channel condition [24]. Based on the international trade theory, they pointed out that the relationship between oil price and the real exchange rate could be explained by terms of trade and purchasing power parity, both of which are driven by the power of arbitrage. Balance of payments theory focuses on nominal exchange rates, while international trade theory studies the effect of oil prices on real exchange rates. What is clear is that for countries with freely floating exchange rates, the channels through which oil price shock affects exchange rates are well understood. For countries with a central exchange rate, the key variable between the oil price and the exchange rate is considered to be the overall level of prices [25]. Moreover, the oil-exchange rate relationship can be affected by other shocks, such as monetary policy, economic uncertainty, or equity prices [26,27].

The empirical literature on the relationship between oil prices and the exchange rate has evolved in multiple directions. Early empirical papers that investigate the impact of oil prices on exchange rates often use the vector autoregression (VAR) model, structural VAR, GARCH or co-integration methods [28,29,30,31,32]. However, there are few works of literature on time-varying effects in these studies, and the subject of exchange rate studies is biased towards a single currency. This does not help financial investors diversify their hedging strategies in the dynamic process. Therefore, several recent papers have used nonlinear methods to assess the dynamic relationship between oil and foreign exchange markets. Basher et al., (2016) used the Markov switch model to study the impact of oil shocks on real exchange rates [33]. They suggested that oil exporters had stronger exchange rate appreciation pressure after experiencing oil demand shocks, but like most studies, the impact of oil supply shocks was insignificant. Xu et al., (2019) indicate that the dependence between oil and foreign exchange will change over time [34]. Yue Liu et al., (2020) studied the dynamic relationship between crude oil and the US dollar exchange rate under the background of structural break detection [35]. They show that oil shocks have a direct and short-term impact on the exchange rate. The research of Malik and Umar (2019) once again verifies the dynamic correlation between oil price shock and the exchange rate market [36]. Demand shock and risk shock have significant impacts on the exchange rate market, while supply shock has rare impacts. Furthermore, they show that the dynamic relationship between oil price shocks and exchange rates increased significantly after the financial crisis. Unfortunately, their study does not indicate when the effect is the strongest, or how soon it wears off, so it would provide vague guidance for policymakers and investors during decision-making.

Using the time-varying parameter structural vector autoregression stochastic volatility (TVP-SVAR-SV) model, scholars discuss the time-varying effects of oil shocks on stock returns, macroeconomics, financial intermediation and other financial variables [37,38,39]. Their work not only demonstrates the time-lag effects of oil shocks on other variables, but also shows responses at different lag periods. We find that the TVP-SVAR-SV model allows time-varying variances of the transmission mechanism and structural shocks. This time-varying aspect reflects the impact of economic and political turbulence and the financialization of oil markets in particular periods. In fact, compared with the constant parameter SVAR model, the TVP-SVAR-SV model allows the detection of changes in the response behavior of the economic system over time, considering the identified structural shocks, while retaining complete sample information specificity [40].

In summary, the main objective of this research is to analyze the time-varying impact of oil price shocks on exchange rate by applying the TVP-SVAR-SV model. Specifically, this paper contributes to the existing literature in different respects. First, this is the first paper to explore the time-varying effects of the oil price shock on exchange rates using the TVP-SVAR-SV methodology. According to the Ready (2018) method, we decompose the oil price shocks into three components: demand, supply, and risk shock [41]. By analyzing the stochastic volatility estimation, the time-varying response of six exchange rates to oil price shocks is studied, which benefits investors and policymakers in asset allocation, diversification, and policy making. Second, our findings will help illuminate how exchange rate responses to oil price shock have changed over time, how these responses have changed at different levels (different lag periods and different time points) and the importance of these responses. Finally, the paper considers potential structural changes caused by several key important events of endogenous or exogenous nature, which are thought to influence the volatility of shocks and the underlying economic dynamics. By considering random fluctuations, it provides a way to manage foreign exchange in times of crisis, thus providing advice to foreign exchange management in times of crisis.

2. Data and Methodology

2.1. Data

The sample period of this paper extends from 19 December 2005 to 19 October 2021, which ultimately yields 5783 daily data observations. The sample period includes the financial crisis of 2008, the Arab Spring of 2010, the election of Donald Trump in 2017, and the COVID-19 pandemic. The related variables are described as follows.

Oil price shock. The original data were extracted from Thomson Reuters DataStream. The index of oil-producing firms, changes in oil prices and changes in expected returns were used in the study. Following the method of Ready (2018), the change in risk is measured by the VIX index, and the demand shock is obtained through the simultaneous regression residual of Oil Production Index returns against the detrended VIX innovation [41]. Supply shocks are collected as independent parts of changes in demand and risk. Figure 1 shows the time evolution of the oil shocks over the whole sample period. It is not hard to see that the financial crisis near 2008 and the raging COVID-19 pandemic in 2020 caused a large-scale outbreak of demand shocks. As for the supply shock, it is obvious that the recent epidemic period, which is still prevalent, has generated the greatest negative shock in the past 15 years. Speculative shocks always undergo frequent volatility and deserve further attention due to the market-driven risks.

Exchange rate. The analysis considers six currencies to the US dollar, including the Australian dollar, Canadian dollar, Swiss franc, euro, British pound, and Japanese yen (AUD, CAD, CHF, EUR, GBP, JPY). They are extracted from the WIND Information database. Figure 2 shows the time evolution of changes in the exchange rate. Overall, a depreciation occurs in both CAD and GBP, while a significant appreciation appears in CHF throughout the sample period. As can be seen, most exchange rates experienced severe shocks in 2008, obviously because of the financial crisis, with the AUD being the most obvious. Subsequently, the global supply chain disruption caused by the COVID-19 pandemic led to a sharp fall in the Australian dollar, but the mechanism still needs further exploration. Over the sample period, the volatility of the JPY and CHF was significantly stronger than that of other currencies. The two currencies are typically safe havens in foreign exchange markets, favored by investors in times of economic turmoil. Various evidence make us doubt the relationship between oil shocks and the foreign exchange market, so the impact of oil shocks on exchange rates at different lag periods still needs further discussion. Before modeling, all variables are converted into a first-order difference form. A summary of the descriptive statistical parameters and unit root tests are provided in

Table 1. Descriptive statistics and unit root tests.

Variables	Mean	Std.Dev.	Skewness	Kurtosis	ADF	PP
DS	0.001	1.800	−0.333	17.026	−58.236 ***	−58.227 ***
SS	0.001	5.662	−42.023	2183.091	−46.274 ***	−45.164 ***
RS	0.016	7.609	1.249	9.873	−62.834 ***	−63.163 ***
CAD	0.000	0.007	−0.208	10.318	−62.878 ***	−62.879 ***
JPY	−0.000	0.648	−0.126	9.222	−66.004 ***	−66.099 ***
EUR	0.000	0.005	−0.021	5.555	−63.218 ***	−63.218 ***
GBP	0.000	0.004	1.432	22.431	−60.604 ***	−60.561 ***
CHF	−0.000	0.007	−4.500	129.679	−64.398 ***	−64.453 ***
AUD	0.000	0.011	0.614	27.882	−69.826 ***	−69.719 ***

Notes: DS, SS and RS denote demand shock, supply shock, and risk shock, respectively. ADF and PP denote augmented Dickey-Fuller and Phillips-Perron, respectively. Asterisks indicate statistical significance at the 1% (***) level.

.

2.2. Time Varying Structural VAR Model with Stochastic Volatility

In this paper, the TVP-SVAR-SV model proposed by Primiceri (2005) is used to assess the time-varying impact of oil shocks on exchange rates [42]. This model, which is an extension of the SVAR model, obtains the instability relationship through time-varying coefficient estimation and solves the heteroscedasticity through time-varying volatility, thereby improving the accuracy of model estimation [43]. Considering the effects of oil demand, supply, and risk shock on exchange rates, the four-variable SVAR model below was first constructed.

$$A{y}_t=\sum _{n=1}^p{\beta }_i{y}_{t-i}+v_t$$

(1)

where $y_t$ = (DS, SS, RS, ER) ′, DS, SS, RS, and RE represent oil demand shock, supply shock, risk shock, and exchange rate. $y_t$ is the 4$\times 1$vector of observed variables, and $A$, ${\beta }_i$are 4$\times 4$ matrices of coefficients. The disturbance $v_t$ is a 4$\times 1$ structural shock and the reduced form innovations and the structural shocks can be represented as follows:

$$\left({\epsilon }_t^{\mathrm{DS}}{\epsilon }_t^{\mathrm{SS}}{\epsilon }_t^{\mathrm{R}\mathrm{S}}{\epsilon }_t^{\mathrm{ER}}\right)=\left(\begin{array}{cc}\begin{array}{cc}{a}_{11}& 0\\ a_{21}& a_{22}\end{array}& \begin{array}{cc}0 & 0\\ 0 & 0\end{array}\\ \begin{array}{cc}{a}_{31}& a_{32}\\ a_{41}& a_{42}\end{array}& \begin{array}{cc}{a}_{33}& 0\\ a_{43}& a_{44}\end{array}\end{array}\right) \left(\begin{array}{c}\begin{array}{c}{v}_t^{\mathrm{DS}}\\ v_t^{\mathrm{SS}}\end{array}\\ \begin{array}{c}{v}_t^{\mathrm{RS}}\\ v_t^{\mathrm{ER}}\end{array}\end{array}\right)$$

(2)

we assume that$v_t~N\left(0,\mathsf{\Psi }\mathsf{\Psi }\right)$, where

$$\mathsf{\Psi }=\left(\begin{array}{cc}\begin{array}{cc}{\sigma }_1& 0\\ 0& \ddots \end{array}& \begin{array}{cc}\dots & 0\\ \ddots & ⋮\end{array}\\ \begin{array}{cc}⋮& \ddots \\ 0& \dots \end{array}& \begin{array}{cc}\ddots & 0\\ 0& {\sigma }_k\end{array}\end{array}\right)$$

(3)

$\sigma$is the standard deviation. By assuming that $A$ is reversible, Equation (1) can be transformed into the following reduced form VAR model:

$$y_t=\sum _{i=1}^p{A}^{-1}{\beta }_i{y}_{t-i}+A^{-1}{v}_t=\sum _{i=1}^p{\mathsf{\Phi }}_i{y}_{t-i}+A^{-1}\mathsf{\Psi }{\epsilon }_t, {\epsilon }_t~N\left(0,I_K\right)$$

(4)

where ${\mathsf{\Phi }}_i=A^{-1}{\beta }_i$, for $i=1,\dots ,s$. Stacking the elements in the rows of the ${\mathsf{\Psi }}_i$s to form $\beta \left({\mathsf{\kappa }}^2 s\times 1 \mathrm{vector}\right)$, and defining $X_t=I_k\otimes \left(y_{t-1}^{\text{'}},\dots ,y_{t-s}^{\text{'}}\right)$, where$\otimes$ is the Kronecker product, the model can be written as

$$y_t=X_t\beta +A^{-1}\mathsf{\Psi }{\epsilon }_t$$

(5)

where all the parameters in Equation (5) are time-invariant. We consider a time-varying parameter condition. The TVP-SVAR-SV model with stochastic volatility can be expressed as:

$$y_t=X_t{\beta }_t+A_t^{-1}{\mathsf{\Psi }}_t{\epsilon }_t , t=s+1,\dots ,n$$

(6)

where the coefficients${\beta }_t$ and the parameters $A_t$ and${\mathsf{\Psi }}_t$ are all time-varying.

According to study of Primiceri (2005), the lower-triangular elements in $A_t$ can be transformed and expressed as $a_t=\left(a_{21},a_{31},a_{32,}{a}_{41,\dots ,}{a}_{k,k-1}\right)\text{'}$ and $h_t={\left(h_{1,t},\dots ,h_{k,t}\right)}^{\text{'}},h_{j,t}=\log {\sigma }_{jt}^2,$ for $=1,\dots ,k,t=s+1,\dots ,n$. We assume that the parameters in Equation (6) follow a random walk process as follows:

$$\left\{\begin{array}{c}{\beta }_{t+1}={\beta }_t+u_{\beta t}\\ a_{t+1}=a_t+u_{at}\\ h_{t+1}=h_t+u_{ht}\end{array}\right. \left\{\begin{array}{c}{\beta }_{s+1}\sim N\left({\mu }_{{\beta }_0},{\mathsf{\Psi }}_{{\beta }_0}\right)\\ a_{s+1}\sim N\left({\mu }_{a_0},{\mathsf{\Psi }}_{a_0}\right)\\ h_{s+1}\sim N\left({\mu }_{h_0},{\mathsf{\Psi }}_{h_0}\right)\end{array}\right.\left(\begin{array}{c}{\epsilon }_t\\ u_{\beta t}\\ u_{at}\\ u_{ht}\end{array}\right)\sim N\left(0,\left(\begin{array}{cccc}I& 0& 0& 0\\ 0& {\mathsf{\Psi }}_{\beta }& 0& 0\\ 0& 0& {\mathsf{\Psi }}_a& 0\\ 0& 0& 0& {\mathsf{\Psi }}_h\end{array}\right)\right)$$

(7)

We further assume that${\mathsf{\Psi }}_{\beta }$, ${\mathsf{\Psi }}_a$ and ${\mathsf{\Psi }}_h$ are all diagonal matrices. The prior distribution of the parameters and its density function is determined first. Taking account of our data, the initial settings for prior parameters are as follows:

$$\left\{\begin{array}{c}{\mu }_{b_0}={\mu }_{a_0}={\mu }_{h_0}=0\\ {\psi }_{b_0}={\psi }_{a_0}=10I\\ {\psi }_{h_0}=100I\\ {\left({\psi }_b\right)}_t^{-2}~\mathrm{G}\mathrm{a}\mathrm{m}\mathrm{m}\mathrm{a}\left(40,0.02\right)\\ {\left({\psi }_a\right)}_t^{-2}~\mathrm{G}\mathrm{a}\mathrm{m}\mathrm{m}\mathrm{a}\left(4,0.02\right)\\ {\left({\psi }_h\right)}_t^{-2}~\mathrm{G}\mathrm{a}\mathrm{m}\mathrm{m}\mathrm{a}\left(40,0.02\right)\end{array}\right.$$

(8)

In order to obtain parameters to be estimated, the Markov chain Monte Carlo (MCMC) method is implemented to generate samples from the posterior distributions [40], and the sampling frequency is 10,000 times.

3. Results and Discussion

3.1. Estimation of Selected Parameters

The optimal lag order is set to be two according to the likelihood ratio (LR) test, Akaike information criterion (AIC), and Schwarz information criterion (SC). To estimate the parameters in the model, the Markov Chain Monte Carlo (MCMC) method is used for simulation sampling to obtain the posterior distribution of the parameters.

Table 2. Estimation Result (CHF).

Parameter	Mean	Stdev	95%L	95%U	Geweke	Inef.
sb1	0.0180	0.0015	0.0154	0.0212	0.622	91.46
sb2	0.0184	0.0014	0.0158	0.0215	0.236	76.00
sa1	0.0373	0.0048	0.0286	0.0476	0.006	192.97
sa2	0.1479	0.0179	0.1142	0.1829	0.013	175.94
sh1	0.3245	0.0193	0.2884	0.3630	0.575	41.26
sh2	0.3895	0.0219	0.3463	0.4330	0.086	52.19

shows the mean, standard deviation, 95% confidence interval, and diagnostic statistics of the posterior distribution of the parameters in the TVP-SVAR-SV model. The mean is in the confidence interval, and the inefficiency factor is relatively low, with a maximum value of 192.97. The Geweke statistics indicate that the parameter converges to the posterior distribution. In summary, simulation sampling using the MCMC method and the subsequent estimation of the TVP-SVAR-SV model parameters is valid.

3.2. Results of Impulse Response

3.2.1. Impulse Response at Different Time Horizons

Figure 3, Figure 4 and Figure 5 show the impulse responses of exchange rates to oil demand, supply, and risk shock for the 1-, 4-, and 8-day lag periods. There is significant variation in the impulse responses over time, which supports applying the TVP-SVAR-SV model. The results show that the impacts of the oil shocks on exchange rates are the highest for the 1-day lag period, followed by the 4-day lag period, while the impacts are minimal for the 8-day lag period.

First, the impacts of an oil demand shock on exchange rates are analyzed. As shown in Figure 3, the impacts of an oil demand shock on the exchange rate weaken as the lag period increases and decreases to almost zero for the 8-day lag period. Figure 3 shows that, to oil demand shock, the impacts on AUD are mostly positive at the 1-day horizon, and the positive responses peak in 2018. It is clear that the trade war between China and the US has dampened demand for crude oil, raising investors’ concerns about geopolitics and energy markets [44]. As a typical commodity-dependent economy, Australia is particularly sensitive to this, and its currency suffered a severe devaluation. The impulse response of oil demand shock on CAD is negative until 2013, and it has since gradually turned positive, indicating that the Canadian central bank’s previous decision to give up austerity preferences has led to resistance on CAD. In addition, weak oil prices and broad declines in commodity currencies accelerated CAD depreciation in the short term after 2013. The response trajectory of CHF to oil demand shock is similar to that of EUR, which is positive. Finally, for the last two graphs in Figure 3, the oil demand shock has the least impact on the GBP and, on the contrary, the most on the JPY. As an oil-importing country, Japan’s low interest makes it more vulnerable to oil-demand shocks. This is consistent with the study of Cunado et al., (2015), who argue that oil demand shock has a significant positive impact on Asian oil importers such as Japan [16]. Moreover, they show that for a country like Japan, policy tools such as interest rate policies and exchange rate policies help mitigate such effects.

Second, we analyzed the impacts of oil supply shock on exchange rates. Similar to the above results, the impacts of oil supply shock on exchange rates also weaken as the lag duration increases, decreasing to almost zero for the 8-day lag period. The impact of the oil supply shock on AUD is positive in the sample period and reaches its peak at the end of 2013. Australia is currently the only net oil importer in the IEA, which explains its heavy dependence on oil supplies. The International Energy Agency (IEA) said in 2013 that OPEC production cuts were already tightening global oil supplies. In the short term, the rising international oil price has led to a rise in the value of the commodity currency, the Australian dollar. More interestingly, CAD shows an obvious cyclical response to the oil supply shock. Specifically, it approximately had a positive peak every two years since 2007, and this cycle became more frequent after the arrival of the COVID-19 pandemic. This implies the need for a cyclical portfolio strategy. In addition, we observe that the medium-term effects of supply shock are much stronger than demand and speculative shocks for CAD, highlighting the long duration (4-day period) of the impact of oil supply shocks on CAD compared to other currencies. The response trajectories of CHF and EUR to supply shocks remain similar. However, in contrast to oil demand shocks, the impulse response of these two exchange rates to oil supply shocks is negative in the short run. For the EU, all oil data are strongly influenced by Russia, which accounts for 32 percent of imports, and the overall risk is high [45]. We suspect that the EUR will fall sharply as investors worry about energy supply problems in the eurozone and a potential recession. The response pattern of GBP to oil supply shock is positive and negative alternately. The response of the yen to supply shock also shows positive and negative alternation, but obviously the fluctuation degree is smoother, and the influence intensity is larger. However, Kim et al., (2020) also said that the Japanese foreign exchange market is highly resistant to the impact of international oil prices, and more attention should be paid to the speculative impact in the following text [46].

Thirdly, we analyze the impact of oil risk shock on the exchange rate. In general, like demand and supply shocks, the impact of risk shock gradually weakens with the increase of the lag period, but the direction, intensity, and size of risk shocks are different. Obviously, risk shock leads to more frequent exchange rate fluctuations. This is particularly prominent in the AUD. As a commodity currency, the Australian dollar is highly speculative. The fluctuation and flexibility of the Australian exchange rate under the impact of oil speculation are also huge. In the whole sample period, the impact of oil speculation on CAD is positive. That may be because Canada has become an increasingly important oil supplier to the United States. As a result, the risk interdependence between oil prices and CAD becomes even stronger [47]. From 2008 to 2011, oil speculative shocks played a more significant role in CHF fluctuations, especially reaching a positive peak in late 2010. This is because speculative capital re-entered the oil futures market after the financial crisis, which intensified the volatility of the oil price and made it have a certain speculative premium rise process. It also depends on the volatility of investor sentiment and perception in financial markets [48]. After the oil risk shock, EUR and GBP showed alternating positive and negative reactions in the short term, but this effect was not very strong compared with other exchange rates. JPY, by contrast, has the strongest response. JPY’s response to oil shocks was positive until 2011, after which there was a cyclical response. The unique speculative shock of global demand and oil markets has a prominent impact on Japan’s macro economy [49].

3.2.2. Impulse Response at Different Time Points

Figure 6, Figure 7 and Figure 8 give the pulse response of the exchange rate market to three types of oil shocks at different points in time. These include four time points, the financial crisis of 2008, the Arab Spring of 2010, the election of Donald Trump in 2017, and the COVID-19 pandemic of 2020. In general, the results show that the impact of oil shocks on the exchange rate market at different points have positive and negative alternating shock effects with time variability and a significant short period impact. The impact is significantly reduced as the number of lag periods becomes longer and disappears within a 5-day lag period.

As shown in Figure 6, we find that oil demand shock has completely different impacts on exchange rates at different time points. First, during the Arab Spring and Trump administration, the impact of the oil demand shock on the AUD and CAD had the same trend in size and direction, while the impact during the COVID-19 pandemic is in the opposite direction after a 1-day lag period. This shows that oil demand shock has similar effects on AUD and CAD when political events occur, while emergencies like the COVID-19 pandemic are uncertain. Second, as observed in time-varying impulse responses at different time horizons in Figure 3, the responses of CHF and EUR at different time points are generally similar, which also shows that the connection characteristics of the two currencies are far from separable. Third, during the financial crisis, the impact of an oil demand shock on JPY reached a positive maximum at the beginning, then began to decline, and at the same time reached a negative maximum in the first period, then gradually fell to zero in the third period. As a major advanced economy, the impact of an oil demand shock on the JPY during the financial crisis is the most significant. Diniz-Maganini et al., (2021) indicate that after the subprime crisis, currencies following a free-float regime experienced a greater deterioration in price efficiency than those under a managed-float regime [50]. They also show that the efficiency of free-floating currencies did not fully recover even a decade after the crisis.

As shown in Figure 7, the impact of an oil supply shock on the exchange rate is much higher in the period of sharp political risk than at other time points. First, compared with the financial crisis and epidemic, the impact of political events is obviously stronger through observing the impact of the oil supply shock on CAD. During Trump’s presidency, CAD’s response to the oil supply shock reached a positive peak, while the Arab Spring was at a negative peak. This is consistent with the research of Lee et al., (2017), who said that economic and political risks significantly impact the supply side of net oil-exporting countries such as Canada [51]. Second, unlike the general similar responses of CHF and EUR in the previous study, oil supply shock has opposite effects on the two exchange rates during the epidemic, with CHF having a positive short-term response while EUR has a negative short-term response. Exogenous cuts in OPEC and non-OPEC oil supply shock would reduce industrial production and raise unemployment in the eurozone [52]. According to the viewpoint of Western classical economics, the rise in the unemployment rate will lead to a decrease in foreign investment, further leading to the devaluation of currency. Enhancing energy security in the EU is important. Third, in the first period after the Arab Spring event, GBP reached a positive maximum and then decreased, and then reached a negative maximum in the second period and went to zero in the fifth period. As North Sea oil production has declined, the UK has turned into a net oil importer. The political risk in the Middle East will lead to the decline of oil supply expectations in the future, which will lead to a rise in oil prices. People’s market expectation psychology is bearish, and the pound will depreciate in the short term, but it will rise steadily in the medium and long term.

As shown in Figure 8, apart from JPY, the responses of the five exchange rates to the oil risk shock caused by political events are similar. In fact, the oil shock is not rooted in the supply and demand of narrow economic concepts, but it closely relates to the security- related political development to a greater extent [53]. National debt and political tensions have special significance for exchange rate changes [54]. However, as political events, the impact of the Arab Spring and the election of Donald Trump are just the opposite. The former has a positive and negative peak, while the latter fluctuates near the zero axis. It is understandable that in the core region of the Middle East, the geopolitical crisis and the arrival of the Arab Spring strongly hit the oil market. Similar cases can also be seen in the Gulf War. In the medium and long term, the political development of the Middle East region with important strategic significance may lead to a new oil crisis. In this regard, European diplomatic and economic policymakers will be wisely advised to take action to prepare for this situation. The impact of the oil risk shock caused by exogenous events such as the financial crisis and COVID-19 on the exchange rate is similar. The strongest reaction to this is JPY. During the financial crisis, the impact of oil risk shock on JPY was much greater than that of the demand shock, and the impact direction is opposite. In this regard, Kohler (2010) has stated that structural changes have taken place in the factors determining exchange rate movements due to the increased role of carrying trade activities [55]. On the one hand, the typical pattern of risk aversion affects crisis flows. On the other hand, interest rate differentials explain more of the crisis-related exchange rate movements in 2008–2009. We speculate that this is likely to be related to the impact of changes in terms of trade. Real oil prices will capture exogenous terms of trade shocks, and in the long run, such shocks may be an important factor in determining the real exchange rate [56].

3.3. Variance Decompositions

The variance decompositions of six exchange rates because of oil shocks are reported in

Table 3. Variance decompositions of exchange rate due to oil shocks (7-day lag period).

	DS	SS	RS	Other Shocks
CAD	18.50%	0.27%	7.15%	74.08%
JPY	2.33%	0.07%	11.21%	86.39%
EUR	4.72%	0.08%	1.55%	93.65%
GBP	6.57%	0.12%	3.49%	89.82%
CHF	2.00%	0.06%	0.24%	97.70%
AUD	17.67%	0.03%	12.79%	69.51%

Notes: DS, SS, and RS denote demand shock, supply shock, and risk shock, respectively.

. This study intends to focus on the influence of three types of oil shocks on the exchange rate. Therefore, we limit our discussion of the forecast error variances explained by oil shock only. The variance decompositions result largely confirms the findings from impulse response, that is, with a few exceptions, oil demand shock has the most significant effect on the exchange rate, whereas supply shock is the weakest. Interestingly enough, the empirical results suggest that the oil price shock explains almost the same forecast error variation (see Table 3) for each group mentioned in the previous section. The estimated results indicate that oil demand shock explains approximately 18.5% and 17.67% of the variation in CAD and AUD. This is related to Canada’s status as a net importer and Australia’s status as a commodity country mentioned in the previous section, resulting in their consistent sensitivity to oil demand. The impact of oil supply shock on the exchange rate is generally small, less than 1%, and even the maximum value is only 0.27% for CAD. However, the contribution rate of oil risk shock to exchange rates varies greatly, with the maximum value being 12.79% for AUD and the minimum value being only 0.24% for CHF. It is worth noting that for Japan, the effect of risk shock on exchange rates is more than about five times that of a demand shock. This means that policymakers and regulators in the Japanese foreign exchange market need to be particularly careful about risk diversification when faced with the risks associated with oil speculation and arbitrage processes.

3.4. Further Discussion

The result of this paper holds that oil has a time-varying effect on the exchange rate, and the dynamic effect is mainly in the short term, while the long-term time-varying effect almost disappears. As a key factor of short-term exchange rate fluctuations, international trade is often used to explain complex exchange rate fluctuations [57]. Oil has always been a high circulation item in the international commodity market and has a large trade flow in international settlements. Therefore, it can affect trade in a short period of time and can quickly impact the exchange rates of many countries. In recent years, oil has been endowed with increasingly enhanced financial attributes. As oil tends to be settled in US dollars in international trade, it further increases the risk exposure of exchange rates to currencies with developed forward markets. In contrast to spot delivery, which requires processing and transportation of a series of time costs, the development of the financialization of petroleum commodities has significantly accelerated the process of influencing the foreign exchange market. The more profound the financial attribute, the more obvious the short-term time change.

By comparing oil shocks from different sources, the demand-driven and risk-driven shocks are clearly more noteworthy. Demand-side driven oil shocks are more rigid and can work through relative productivity differences between both traded and non-traded goods sectors. When the non-oil export competitive sectors of the domestic economy are crowded out by the oil and non-trading sectors, further appreciation of the real exchange rate and Dutch disease occur [58]. Exogenous shocks caused by changes in financial market conditions are also a key determinant of oil prices [59]. Through traditional trade channels, the impact of oil shocks on foreign exchange is reflected in positive terms-of-trade shocks pushing up the prices and real exchange rates of non-tradable goods in the domestic economy. For oil exporters, the resulting valuation effect can also be explained by the monetary composition of net foreign assets. Under the investment trading and spot trading system, the loss of valuation effect can relieve the pressure of currency appreciation.

4. Conclusions and Implications

4.1. Conclusions

Based on the TVP-SVAR-SV model of Primiceri (2005), this paper discusses and analyzes the time-varying response of exchange rates to oil shocks. Specifically, this paper uses the methodology proposed by Ready (2018) to measure oil shocks (oil demand, supply, and risk shock). The exchange rate market is represented by: AUD, CAD, CHF, EUR, GBP, and JPY. The sample period runs from 19 December 2005 to 19 October 2021. By allowing the transmission mechanism of oil shocks to be time-varying, the empirical investigation produced expected results to add contributions to the existing literature.

First, we find that oil shocks have significant time-varying effects on the exchange rate market, as the duration of the lag increases, the impact gradually weakens. The time-varying effect of oil shocks on exchange rates is strongest in the lag period of 1 day. Second, by observing the impulse response at different time horizons, we find that the oil demand shock has the most significant time-varying effect on exchange rates, while the supply shock is the weakest. Most exchange rates respond positively and negatively to oil shocks alternately, reflecting a certain periodicity. Oil shocks caused by speculative risk make exchange rates more volatile, and nowhere is this more evident than in JPY. Third, by examining impulse responses at different points in time, we find that a series of exchange rates tend to respond to exogenous contingencies and political risks. When political risk rises sharply, the impact of an oil supply shock on countries’ exchange rates rises significantly. This effect will gradually disappear within 5 days, that is, the long-term time-varying effect is leveled off. This adds to our understanding of the impact of the oil market on the foreign exchange market in times of emergencies.

4.2. Theoretical Implications

By studying the time-varying effect of oil shocks on exchange rates, this paper supplements and extends the existing theories. First, we illustrate the complex dynamic role of oil in commodity markets on exchange rates compared to traditional factors affecting exchange rates, such as international borrowing, inflation rate, national interest rate level, and balance of payments level. This not only makes up for the role of commodity market in the exchange rate determination theory, but also provides a new perspective for the future study of the exchange rate market, jumping out of the traditional cognition. Secondly, in the context of the complex time-varying dynamic impact of oil on the exchange rate, we clearly classify oil shocks from different sources. In terms of the traditional supply and demand shock, it is consistent with the purchasing power parity hypothesis and the interest rate parity hypothesis, that is, the oil supply and demand shock will affect the inflation rate, and then cause the foreign exchange fluctuation. In addition to the role of bulk commodities in the international trade market, the mechanism of oil’s financial properties on foreign exchange is more deeply verified, that is, oil price changes are driven by speculation and arbitrage improves the investment portfolio and terms of trade and affect the exchange rate market, which is a good extension of the international financial theory.

4.3. Policy Implications

From a policy perspective, our results offer the following important suggestions for risk aversion, investment decisions, and policy making in different markets. First of all, investors and policymakers can make investment decisions and policy adjustments according to the characteristics of the short-term impact of the oil shock on the exchange rate market. Given the short-term behavior of exchange rates following oil shocks, it is recommended that policymakers of all six exchange rates take measures to protect against the negative effects of instantaneous shocks. Second, investors and policy makers can obtain excess returns and formulate policies to stabilize the market according to the heterogeneity of the response of different exchange rates to oil shocks. For example, for commodity countries, Australia should pay special attention to the exchange rate fluctuations caused by the impact of oil supply and demand, and further improve the market mechanism. For Japan, financial market investors can consider the portfolio income brought by the impact of oil speculation. Third, a stable economic and political environment is an important condition for alleviating the adverse impact of oil price shocks. In times of acute political risk, central banks and monetary policymakers should pay more attention to the instantaneous impact of oil supply shocks. In order to prevent external shocks from being transmitted to the capital market through the oil market and interfering with the normal operation of the national capital market, countries need to strengthen the monitoring and management of the import and export of international hot money. If necessary, appropriate interventions should be made to release exchange rate expectations.

Although our results benefit the time-varying effects of oil shocks on exchange rates and the method we use, our research is limited. Due to the limitation of data availability, we do not distinguish the specific source of oil shocks from which country, so the results are not targeted enough. Therefore, we look forward to further addressing this issue in future studies.