The Dynamic Impact of Structural Oil Price Shocks on the Macroeconomy

Ping LI; Jie LI; Ziyi ZHANG

doi:10.21078/JSSI-2021-469-29

PDF(1247 KB)

Journal of Systems Science and Information ›› 2021, Vol. 9 ›› Issue (5) : 469-497. DOI: 10.21078/JSSI-2021-469-29

The Dynamic Impact of Structural Oil Price Shocks on the Macroeconomy

Author information +

History +

Abstract

In this paper, we apply the structural vector autoregression (SVAR) model to decompose the international oil price shock into oil supply shocks, aggregate demand shocks and oil-specific demand shocks, and then use the DCC-GARCH model to analyse the dynamic correlations between these three kinds of oil price shocks and the macroeconomic variables of several oil importing and exporting countries. To quantify the intensity of the effect of oil shocks on these variables, we propose a measure, conditional expectation (CoE), to capture the percent change of the economic variable under oil price shocks relative to the median state. The time-varying copula model is employed to estimate the proposed measure through time. The empirical results show that, for instance, the impacts of oil price shocks on macroeconomic variables are different in different periods, showing the time-varying characteristics. Additionally, the impacts of oil price shocks on macroeconomic variables show great differences and some similarities among different countries. Finally, we give some policy suggestions for these countries, in particular for China's special results.

Key words

structured oil shocks / SVAR / DCC-GARCH / macro-economy / conditional expectation (CoE) / time-varying copula

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

Ping LI , Jie LI , Ziyi ZHANG. The Dynamic Impact of Structural Oil Price Shocks on the Macroeconomy. Journal of Systems Science and Information, 2021, 9(5): 469-497 https://doi.org/10.21078/JSSI-2021-469-29

1 Introduction

Since oil has been dominating the sources of the world energy and playing an critical role in economic activities and industrial production for several centuries, all countries in the world concentrate on oil reserve and pricing while market participants are active in investment and speculation in the oil market. The imbalance between supply and demand and the other various market factors produce effects on the oil prices changes caused by the different shock origins which may have diversified influence on macroeconomic indicators in different economies. Under such circumstance, it is critical to figure out that whether and how international oil price fluctuation could lead to the effects on the macroeconomics and what the policy makers and market participants could do to deal with these effects, and to provide some valuable policy implications.

Most researches about the international oil fluctuation focus on two aspects: The reason of the international oil price fluctuation and the influence of the oil price changes on the macroeconomy. In earlier researches, scholars put forward a point of view that the international oil price is mainly affected by supply factors and take the macroeconomic factors as exogenous, such that the research about the influence of the international oil price changes on the macroeconomic is limited to the oil price itself. Since Hamilton^[1] pointed out that the oil price hikes is the Granger cause of the recession of U.S. economy, the effects of oil price shocks on the macroeconomy have increasingly been analyzed in a number of studies, such as economic growth^[2-6], commodity price^[7], exchange rate^[8-10], stock index returns^[11-13] etc.

With the in-depth study on the dependence between the oil price and the economy, scholars find that oil price is born economical and affect the world economy conversely. Kilian^[14] proposed the Structural Vector Autoregression (SVAR) model with the global crude oil production, global real economic activity, and real oil prices to distinguish among different sources of oil shocks, and decomposed the effects on oil price into oil supply shocks, aggregate demand shocks and oil-specific demand shocks. This model provides a new thinking to analyze the effects of oil price on the macroeconomy. Apergis and Miller^[15] applied SVAR model to investigate the sample of eight countries and show how the structural oil price shocks affect the stock market return. Kilian and Park^[16] affiliated U.S. stock market return into the three-variable SVAR model and establish four-variables SVAR model to analyze the effect of the structural oil shocks on the U.S. stock market. Peersman and Robays^[17] use SVAR model to investigate the effects of structural oil price shocks on the macroeconomy of different industrial countries. Ji, et al.^[18] analyzed the influence of structural oil price shocks on production, exchange rate and inflation in BRICS countries. Li, et al.^[19] affiliated domestic demand shock into three-variable SVAR model to study how structural oil price shocks affect Chinese stock market return. However, the researches mentioned above all apply SVAR model to build the mean model and use the impulse response analysis and variance decomposition, and don't take the dynamic effects of the structural oil price on the economic variables into consideration. Furthermore, there exist some studies on the time-varying characteristics, for instance, Broadstock and Filis^[20] combined the SVAR model and the Scalar-BEKK model to investigate the relationship between structural oil price shocks and stock market returns. Aastveit, et al.^[21] and Baumeister and Peersman^[22] applied the time-varying parameter VAR model (TVP-VAR) to study the effect of the oil price shock on the macroeconomy. An alternative dynamic method is the DCC-GARCH model which is applied by Filis, et al.^[23] to analyze the dynamics between stock market returns of oil importing and exporting countries and the international oil price.

In this paper, we employ Kilian's SVAR model^[16] to decompose the international oil price shocks into oil supply shocks, aggregate demand shocks and oil-specific demand shocks. Then the DCC-GARCH model is used to investigate the dynamic relationship between these three oil shocks and macroeconomic indicators. Then to quantify the intensity of effects of oil shocks on these variables, we propose a new measure, conditional expectation (CoE) capturing the difference between the expected level of the macroeconomic variable conditional on the oil price shock and that conditional on the benchmark state of the oil price. Although the TVP-VAR model is useful to study the dynamic characteristics of the influence between two variables, it is not flexible to capture the different effect modes (e.g., CoE). In this paper we propose to use the time-varying copula (see [24]) to character the joint distribution between the oil shocks and the macroeconomic variables, to estimate the proposed measure. As the proxy of the macroeconomy, we choose the industrial output, inflation and exchange rate which are usually considered as the main macroeconomic variables by most researchers, as well as the economic policy uncertainty. As argued by many scholars (see [25-28]), oil price increases driven by different kinds of oil shocks have different consequences on the U.S. economic policy uncertainty. In this study, we extend the focused objects to several main oil importing and exporting countries in the world.

The main contributions of this paper are as follows: 1) Combining SVAR model with DCC-GARCH model, we temporally study the dynamic correlation between oil price shocks and a country's macro-economy in different time periods from the perspective of structural impact of oil price, rather than just regarding the oil price as a variable born out of the global economy; 2) We propose a new measure, to quantify the intensity of effects of oil shocks on macroeconomic variables, estimated by the time-varying copula model; 3) From the perspective of oil importing and exporting countries, the spatially comparative analysis is carried out among the major economies in the world, and the commonness and differences between these countries are discussed.

The remainder of this paper is organized as follows. Section 2 outlines the methodology. Section 3 introduces the data for major oil importing and exporting countries along with some descriptive statistics. Section 4 presents empirical results and the final section concludes and gives some policy suggestion.

2 Methodologies

With the regard to the research methodology, we firstly apply the Kilian's SVAR model^[16] to decompose the oil price shocks into the oil supply shocks, aggregate demand shocks and oil-specific demand shocks. Then, the DCC-GARCH model is exploited to analyze the dynamic dependence between three kinds of oil shocks and macroeconomic variables of different economies. Finally, we apply the conditional expectation to analyze the dynamic impact of oil shocks on the macroeconomic indicators.

2.1 SVAR Model

Consider a VAR model based on the data

z_{t} = (Δ {p r o d}_{t}, {r e a}_{t}, {r p o}_{t})^{'}

, where

Δ {p r o d}_{t}

is the percent changes in global crude oil production,

{r e a}_{t}

denotes the index of real economic activity, and

{r p o}_{t}

defers to the real price of the oil. Kilian's SVAR model^[16] is given by

A_{0} z_{t} = α + \sum_{i = 1}^{p} A_{i} z_{t - i} + ε_{t},

(1)

where

ε_{t}

denotes the vector of serially and mutually uncorrelated structural innovations. We assume that

A_{0}^{- 1}

has a recursive structure such that the reduced-form errors

e_{t}

can be decomposed according to

e_{t} = A_{0}^{- 1} ε_{t}

e_{t} \equiv (\begin{matrix} e_{t}^{Δ p r o d} \\ e_{t}^{r e a} \\ e_{t}^{r p o} \end{matrix}) = [\begin{matrix} a_{1} 1 & 0 & 0 \\ a_{2} 1 & a_{2} 2 & 0 \\ a_{3} 1 & a_{3} 2 & a_{3} 3 \end{matrix}] (\begin{matrix} ε_{t}^{o i l s u p p l y s h o c k} \\ ε_{t}^{a g g r e g a t e d e m a n d s h o c k} \\ ε_{t}^{o i l - s p e c i f i c d e m a n d s h o c k} \end{matrix}) .

(2)

This model postulates a vertical short-run supply curve of crude oil. Changes of the demand curve caused by either of the two oil demand shocks leads to an immediate shifts in the real price of oil, as do unpredictable oil supply shocks that change the vertical supply curve. The limitations

A_{0}^{- 1}

on may be motivated as follows.

Crude oil supply shocks are defined as unanticipated innovations to global production. It is assumed that crude oil supply does not respond to innovations to the demand for oil within the same month. Innovations to global real economic activity that cannot be explained based on crude oil supply shocks will be referred to as shocks to the global demand for industrial commodities, that is, the aggregate demand shocks. Furthermore, innovations to real price of oil that cannot be explained based on oil supply shocks or aggregate demand shocks by construction will reflect changes in the demand for oil as opposed to changes in the demand for all industrial commodities. The latter structural shock will reflect, in particular, fluctuations in precautionary demand for oil driven by uncertainty about future oil supply shortfalls. Whereas this shock potentially could also reflect other oil market-specific demand shocks, there are strong reason to believe that it effectively represents exogenous shifts in precautionary demand.

Firstly, there are no other plausible candidates for exogenous oil market-specific demand shocks. For instance, Asian consumers' preference for durable goods (such as automobiles) has increased since 2003, leading to a sharp rise in real oil prices. This preference may seem to be an exogenous special demand shock. However, the rise in oil prices is not associated to the rise of oil-specific demand, but is caused by global economic prosperity, allowing us to exclude this interpretation.

Secondly, the empirical results of Kilian^[16] show that the large impact effect of the oil-specific demand shock is very different from those caused by other factors.

Thirdly, the timing of these oil-specific demand shocks and the direction of their effects is in accordance with the timing of exogenous events such as the outbreak of the Persian Gulf War that would be expected to affect uncertainty about future oil supply shortfalls on a priori grounds.

Fourthly, the movements in the real price of oil induced by this shock are highly correlated with independent measures of the precautionary demand component of the real price of oil based on futures prices. Applying oil futures market data, Alquist and Kilian^[29] show that this correlation may be as high as 80 percent, notwithstanding the use of a completely different dataset and methodology.

2.2 DCC-GARCH Model

Bollerslev^[30] proposes constant conditional correlation GARCH model (CCC-GARCH) shown as follows:

Let

r_{1, t}, r_{2, t}

denote random variables with mean equal to 0 and satisfy:

\begin{array}{rcl} r_{t} = β_{0} + \sum_{i = 1}^{k} β_{i} r_{t - i} + u_{t} = μ_{t} + u_{t}, \end{array}

(3)

\begin{array}{rcl} μ_{t} = E [r_{t} | Ω_{t - 1}], \end{array}

(4)

\begin{array}{rcl} u_{t} | Ω_{t - 1} \sim N (0, H_{t}), \end{array}

(5)

where

H_{t}

denotes conditional covariance matrix given as follows:

H_{t} = D_{t} R D_{t},

(6)

where

D_{t} = d i a g {\sqrt{h_{i i, t}}}, R = (ρ_{i j})_{N \times N}

R

denotes the matrix of conditional correlation coefficients which are constants.

h_{i i, t}

denotes the conditional variance characterized by an univariate GARCH model given as follows:

h_{i i, t} = ω_{i} + \sum_{p = 1}^{P_{i}} α_{i p} ε_{i, t - p}^{2} + \sum_{q = 1}^{Q_{i}} β_{i q} h_{i i, t - q} .

(7)

Although CCC model can simplify the calculation of variance matrix, the assumption that the correlation coefficient between assets is constant does not conform to the reality. For this reason, Engle^[31] introduced the DCC-GARCH model, assuming that the correlation coefficients are time-varying, and showing this model's advantage in parameter estimation. Specifically, the mean equation of DCC model is the same as Equation (3)

\sim

(5) while the residual equation (6) is replaced by

H_{t} = D_{t} R_{t} D_{t},

(8)

where

D_{t} = d i a g {\sqrt{h_{i i, t}}}, R_{t} = (ρ_{i j}), i, j = 1, 2

R_{t}

denotes the dynamic, time-varying matrix of conditional correlation coefficient with the dynamic constraint

\begin{array}{rcl} R_{t} = Q_{t}^{* - 1} Q_{t} Q_{t}^{* - 1}, \end{array}

(9)

\begin{array}{rcl} Q_{t} = (1 - α - β) \bar{Q} + α (ε_{1, t - 1} ε_{2, t - 1}^{'}) + β Q_{t - 1}, \end{array}

(10)

where

Q_{t} = (\begin{matrix} q_{11, t} & q_{12, t} \\ q_{21, t} & q_{22, t} \end{matrix}),

Q_{t}^{*} = (\begin{matrix} \sqrt{q_{11, t}} & 0 \\ 0 & \sqrt{q_{22, t}} \end{matrix})

\bar{Q}

denotes unconditional correlation matrix of standardized residual. Obviously, the

Q

matrix to be positive definite or positive semi-definite is a sufficient condition for the

R_{t}

to be positive definite.

α

and

β

are coefficients of the DCC-GARCH model.

2.3 Conditional Expectation Model

Conditional expectation (CoE) for the macroeconomic variable is the expectation of the distribution of the macroeconomic variable, conditional on the fact that a given oil price shock occurs. Let

R_{t}^{u}

be the values of the macroeconomic variable and

R_{t}^{v}

be the values of the oil price shock. For the oil supply shock,

R_{t}^{v}

decrease means the global oil supply declines. For the aggregate demand shock and the oil-specific demand shock,

R_{t}^{v}

increase means the oil demand rises. Then, the "shock" can be defined by

R_{t}^{v}

falling below some threshold for the oil supply shock or rise above some threshold for both demand shocks. For simplified and unified computation, we use the sign-opposite value of sequences for both aggregate demand shocks and oil-specific demand shocks. These three kinds of "shocks" can be all defined as

{R_{t}^{v} \leq {V a R}_{α}}

, where is

{V a R}_{α}

the

α

-quantile as the threshold. The conditional distribution of the macroeconomic variable

R_{t}^{u}

is defined as follows:

F (r^{u} | R_{t}^{v} \leq {V a R}_{α}) = P (R_{t}^{u} \leq r^{u} | R_{t}^{v} \leq {V a R}_{α}),

(11)

and the conditional expectation (CoE) of

R_{t}^{u} | R_{t}^{v} \leq {V a R}_{α}

can be calculated by

\begin{array}{rcl} {C o E}_{t}^{u | v, α} & = & E (R_{t}^{u} | R_{t}^{v} \leq {V a R}_{α}) \\ = & \int_{a}^{b} \frac{\partial F (R_{t}^{u} | R_{t}^{v} \leq {V a R}_{α})}{\partial (R_{t}^{u} | R_{t}^{v} \leq {V a R}_{α})} (R_{t}^{u} | R_{t}^{v} \leq {V a R}_{α}) d (R_{t}^{u} | R_{t}^{v} \leq {V a R}_{α}) . \end{array}

(12)

An alternative method to calculate the conditional distribution shown as Equation (11) is copula function if we express this equation as

\begin{array}{rcl} F (r^{u} | R_{t}^{v} \leq {V a R}_{α}) & = & P (R_{t}^{u} \leq r^{u} | R_{t}^{v} \leq {V a R}_{α}) \\ = & \frac{P (R_{t}^{u} \leq r^{u}, R_{t}^{v} \leq {V a R}_{α})}{P (R_{t}^{v} \leq {V a R}_{α})} \\ = & \frac{C (U (r^{u}), V ({V a R}_{α}))}{α} \\ = & \frac{C (U (r^{u}), α)}{α}, \end{array}

(13)

where

U (\cdot)

and

V (\cdot)

are marginal distributions of

R_{t}^{u}

and

R_{t}^{v}

, respectively.

C (U, V)

is the copula function linking both marginal distributions. Given a specific copula function, the conditional expectation (CoE) in Equation (12) can be calculated.

We denote the intensity of effects of the oil shock on the macroeconomic variable by

Δ {C o E}_{t}^{u | v} = \frac{{C o E}_{t}^{u | v, α} - {C o E}_{t}^{u | v, α = 0.5}}{{C o E}_{t}^{u | v, α = 0.5}},

(14)

where

α = 0.5

means oil supply or demand are in the median state. Hence,

Δ {C o E}_{t}^{u | v}

can capture the percentage difference between the expected level of the macroeconomic variable conditional on the oil price shock

R_{t}^{v} \leq {V a R}_{α}

and the expected level of the macroeconomic variable conditional on the benchmark state of the oil price.

Considering that there could exist negative, positive, and tail dependence between the oil price shock and the macroeconomic variable, we link both two marginal distributions by the Student's

t

copula specified as:

T_{ρ, ν} (u, v) = t_{ρ, ν} (t_{ν}^{- 1} (u), t_{ν}^{- 1} (v)),

(15)

where

ρ

denotes the dependence parameter and

ν

denotes the degree of freedom. In addition, we consider possible time-varying parameter dependence by allowing the parameter to vary according to a specific evolution equation, which is the ARMA

(1, q)

type process (see [24]) as follows:

ρ_{t} = Λ (\bar{ω} + \bar{β} ρ_{t - 1} + \bar{α} \frac{1}{q} \sum_{j = 1}^{q} t_{ν}^{- 1} (u_{t}) t_{ν}^{- 1} (v_{t})),

(16)

where

Λ (x) = (1 - e^{- x}) (1 + e^{- x})^{- 1}

is the modified logistic transformation that keeps the value of

ρ_{t}

(- 1, 1)

. The dependence parameters are explained by constants, an autoregressive term, and the average product over the last

q

observations of the transformed variables.

3 Data Description

As for countries to be investigated, the selection criteria consist of that monthly macroeconomic data is available, and that the selected countries should be the top 20 oil importers or exporters in the world, and are the major economies that have an important impact on the world economy. Based on these two points, we choose China, U.S., Japan and India as the oil importers, and Brazil, Canada and Russia as the oil exporters. In importing countries, China's degree of dependence on international crude oil in 2016 was 64.5%, U.S.'s degree of dependence on international crude oil in 2015 was 42.5%, Japan's in 2015 was 99.7% and India's in 2014 was 70%. Obviously, our selected importing countries all have a high degree of dependence on international crude oil and their economies is more likely to be influenced by international crude oil price changes.

As for the macroeconomic variables, the selection criteria consist of that the chosen variable should be an important indicator of the macroeconomic situation of an economy (e.g., production, exchange rate, interest rate) and that the monthly data of these chosen variables is available. Based on this two points, we choose industrial output, economic policy uncertainty, inflation and exchange rate as macroeconomic variables.

The monthly data of selected variables consist of two parts: One is the international oil market data, the other is the macroeconomic indicator data of different countries. As for the international oil market, we refer to Kilian's selection: Global crude oil production (OPROD), global economic activity (EACT), international oil price (OP). As for macroeconomic variables, we choose industrial output index (IOI), economic policy uncertainty (EPU), inflation (Customer Price Index, CPI), exchange rate (EXR). The variables are prefixed by the name of the country to represent the macroeconomic variables of different countries.

Global crude oil production data is collected from the US Energy Information Agency (EIA, http://www.eia.gov/). We take the Global Dry Bulk Rate Index proposed by Kilian^[16] as the proxy of the variable EACT. The reason of GDP or the weighted increment of the industrial production of these countries are not chosen is that the disclosure of GDP is generally lagging, the cycle is larger, even many countries have no monthly disclosure of GDP. Based on Dry Cargo Bulk Freight Rates Kilian^[16] proposed the Global Dry Bulk Rate Index, which contains more dynamic information and has significant positive correlation with economic activity so as to reflect the situation of global economic activity. This index can be collected from Kilian's website (http://www-personal.umich.edu/~lkilian/). Furthermore, Brent crude oil price are considered as the proxy of the global international oil price (OP) because its covering about 60 percent share of the crude oil trade and can be collected from the U.S. Energy Information Agency (EIA).

As a proxy of industrial output index (IOI), we use the gross industrial production for U.S. and Canada, the industrial added values on a year-on-year basis for China (see [19]), the industrial production index for Japan, India, Russia and Brazil. These data are collected from WIND database. The economic policy uncertainty (EPU) can be measured by Economic Policy Uncertainty index proposed by Baker, et al.^[32] based on the frequency of the news reports to measure the changes of the economic uncertainty relevant with the policy. For instance, as they show, the US EPU index has peaks in the presidential election, the first and second Gulf wars, the 911 attacks, the Lehman Brothers bankruptcy, the 2011 European debt crisis and other major fiscal policy events (for the EPU index of the major countries in the world, see http://www.policyuncertainty.com/index.html). As a proxy of inflation, we use the Consumer Price Index (CPI) collected from the WIND database. As a proxy of exchange rate (EXR), we use real effective exchange rate index collected from Bank of international settlements.

Apart from Indian economic policy uncertainty index whose available data interval is relatively short (from January 2003 to December 2016), the sample interval of all other countries is from January 2000 to December 2016, including 204 observations for each variable (Indian EPU variable includes 168 observations). Many studies have focused on the three oil crisis in twentieth Century and the global financial crisis in 2008, while there are few studies on the price of oil in recent years. The sample interval of this article allows us to study the impact of oil prices on the macro-economy after 2008, filling the timeliness and completeness of the gaps and shortcoming.

Following the treatment of Kilian^[16], apart from the crude oil production (OPROD) using the returns, all other variables use the raw data. Considering the stationary of the data sequence, we use the first order difference of all series to ensure the stability of the fitted SVAR model. Table 1 reports some basic descriptive statistics about all series in our sample.

Table 1 Descriptive statistics of the sample series

	Variable	Mean	Max	Min	Std.Dev	Skew.	Kurt.	JB stat.	$p$ -value
Global	OPROP	0.14	2.62	$-$ 2.46	0.76	$-$ 0.17	4.12	11.67	0.0029
	EACT	6.85	66.36	$-$ 133.65	33.64	$-$ 0.57	3.80	16.55	0.0003
	OP	64.79	132.72	18.71	32.31	0.34	1.79	16.40	0.0003
China	CH_IOI	12.41	29.20	$-$ 2.93	4.76	0.01	3.13	0.15	0.9285
	CH_EPU	136.71	646.91	9.07	94.25	1.99	8.81	420.66	0.0000
	CH_CPI	100.20	102.60	98.40	0.66	0.22	3.50	3.76	0.1523
	CH_EXR	100.90	131.01	81.89	13.81	0.70	2.40	19.58	0.0001
U.S.	US_IOI	3488.48	3750.79	3137.32	144.74	$-$ 0.15	2.11	7.47	0.0239
	US_EPU	113.85	245.13	57.20	36.99	0.77	2.94	20.13	0.0000
	US_CPI	209.14	241.73	168.80	22.38	$-$ 0.19	1.67	16.41	0.0003
	US_EXR	108.84	128.93	93.15	9.96	0.30	2.00	11.57	0.0031
Japan	JA_IOI	101.41	125.30	73.50	9.10	0.03	3.38	1.27	0.5305
	JA_EPU	106.05	239.81	48.38	33.20	1.11	4.92	73.28	0.0000
	JA_CPI	97.71	100.40	95.70	1.28	0.64	2.22	19.10	0.0001
	JA_EXR	95.50	128.09	68.19	14.59	0.06	2.42	2.97	0.2268
India	IN_IOI	134.69	198.70	73.00	39.18	$-$ 0.24	1.55	19.84	0.0000
	IN_EPU	98.11	283.69	24.94	53.75	1.14	4.12	45.16	0.0000
	IN_CPI	509.91	878.00	299.00	193.36	0.57	1.85	22.43	0.0000
	IN_EXR	93.91	104.77	84.24	4.70	0.48	2.47	10.26	0.0059
Brazil	BR_IOI	91.11	112.60	65.33	10.95	$-$ 0.06	2.13	6.63	0.0363
	BR_EPU	139.45	473.53	22.30	77.31	1.78	6.89	236.34	0.0000
	BR_CPI	3012.79	4940.78	1598.24	908.36	0.35	2.29	8.61	0.0135
	BR_EXR	78.34	110.15	42.07	16.37	$-$ 0.16	2.06	8.36	0.0153
Canada	CA_IOI	48068.49	53668.07	38330.99	2956.46	$-$ 0.87	3.52	27.96	0.0000
	CA_EPU	135.32	442.37	39.32	76.87	1.15	4.39	61.36	0.0000
	CA_CPI	112.80	129.10	93.50	10.18	$-$ 0.13	1.83	12.23	0.0022
	CA_EXR	89.47	106.80	71.81	9.69	$-$ 0.24	1.75	15.09	0.0005
Russia	RU_IOI	95.30	127.50	64.60	14.83	$-$ 0.39	2.24	10.16	0.0062
	RU_EPU	118.96	370.51	12.40	71.61	1.15	4.17	56.60	0.0000
	RU_CPI	100.88	103.85	99.59	0.66	1.33	5.33	106.66	0.0000
	RU_EXR	85.07	110.61	47.49	16.43	$-$ 0.27	1.88	13.10	0.0014

Notes: Std.Dev, Skew., Kurt., and JB stat. denote standard deviation, skewness, kurtosis and JB statistics, respectively.

From Table 1, we can find that Jarque-Bera tests reject the null Gaussian distribution for all the series under 0.05 confidence level, except for China industrial output, China CPI, Japan industrial output, Japan real exchange rate, which shows the evidence that the distributions of most series are not normal.

The establishment of the SVAR model and the DCC model requires that all the time series are stationary and have the ARCH effects. Through ADF test and ARCH-LM test as shown in Table 2, we find that all series are stationary and that international oil market and most of the macroeconomic variables have ARCH effects. Though the economic policy uncertainty, inflation and real exchange rate in some countries do not passed by test, referring to Li, et al.^[19], we still keep these variables for consideration of the integrity of this research.

Table 2 Unit root test and ARCH effects test for all series

	Variable	ADF	$p$ -value	ARCH-LM	$p$ -value
Global	OPROP	$-$ 12.03	0.0000	3.86	0.0024
	EACT	$-$ 10.51	0.0000	9.05	0.0000
	OP	$-$ 9.49	0.0000	10.15	0.0000
China	CH_IOI	$-$ 13.23	0.0000	6.77	0.0000
	CH_EPU	$-$ 14.47	0.0000	19.61	0.0000
	CH_CPI	$-$ 12.63	0.0000	2.66	0.0238
	CH_EXR	$-$ 10.12	0.0000	0.87	0.5012
U.S.	US_IOI	$-$ 3.66	0.0271	15.42	0.0000
	US_EPU	$-$ 12.75	0.0000	0.56	0.7310
	US_CPI	$-$ 9.08	0.0000	4.11	0.0015
	US_EXR	$-$ 10.30	0.0000	0.48	0.7894
Japan	JA_IOI	$-$ 5.06	0.0002	4.15	0.0013
	JA_EPU	$-$ 12.82	0.0000	2.87	0.0159
	JA_CPI	$-$ 11.76	0.0000	0.14	0.9817
	JA_EXR	$-$ 10.87	0.0000	2.09	0.0679
India	IN_IOI	$-$ 3.52	0.0406	7.89	0.0000
	IN_EPU	$-$ 16.11	0.0000	21.45	0.0000
	IN_CPI	$-$ 7.21	0.0000	0.43	0.6120
	IN_EXR	$-$ 13.07	0.0000	0.06	0.8935
Brazil	BR_IOI	$-$ 3.89	0.0141	3.66	0.0019
	BR_EPU	$-$ 9.94	0.0000	4.38	0.0008
	BR_CPI	$-$ 6.08	0.0000	1.36	0.2431
	BR_EXR	$-$ 9.84	0.0000	5.52	0.0001
Canada	CA_IOI	$-$ 14.88	0.0000	9.49	0.0000
	CA_EPU	$-$ 11.11	0.0000	1.47	0.2017
	CA_CPI	$-$ 11.23	0.0000	0.72	0.6077
	CA_EXR	$-$ 11.09	0.0000	1.74	0.0752
Russia	RU_IOI	$-$ 3.19	0.0892	2.70	0.0221
	RU_EPU	$-$ 12.94	0.0000	5.79	0.0001
	RU_CPI	$-$ 5.76	0.0000	4.03	0.0462
	RU_EXR	$-$ 3.59	0.0335	11.46	0.0000

4 Empirical Results

In this section, first we apply impulse response function to analyze the impact of structural oil price shocks on macroeconomic variables for each country. Then we empirically investigate dynamic dependence between structural oil price shocks and macroeconomic variables using the monthly observations of the international crude oil market and the macroeconomic variables in different countries. Finally, the dynamic impact of oil price shocks on macroeconomic variables are calculated by applying the measure of the impact intensity,

Δ C o E

, described in Section 2.3.

4.1 The Impact of Structural Oil Price Shocks on Macro-Economy

The underlying causes of oil price increases are decomposed into three kinds of shocks in this work. Oil supply shocks are defined as unpredictable interruptions to global oil production; aggregate demand shocks represent innovations in the oil flow demand induced by a global economic boom; and oil-specific demand shocks are innovations in crude oil prices and mainly caused by expectation changes or speculative activities in global oil markets. It can be expected that the influence of these shocks on the macroeconomic indicators of each oil-importing and exporting country will be quite different because of their different transmission paths.

The oil supply shock is the oil price increasing caused by the shortage of global oil supply. The main source is the interruption of global oil production, such as war, weather, production limit of exporting countries etc. The impacts of one negative standard deviation oil supply shock on industrial output, exchange rates, inflation, and economic policy uncertainty for each country are shown in Figure 1. Then, the positive response means oil supply disruptions will increase the macroeconomic indicator, which is different from aggregate demand shocks and oil-specific demand shocks. The significance of impulse response function is defined as both one-standard-error bands are above or below zero on the

y

-axis.

	Importers				Exporters
	China	U.S.	Japan	India	Brazil	Canada	Russia
IOI	0.00	0.16	0.06	$-$ 0.44	0.19	0.24	$-$ 0.21
EPU	0.17	$-$ 0.27	$-$ 0.06	$-$ 0.14	$-$ 0.22	0.44	0.09
CPI	$-$ 0.70	$-$ 0.16	0.30	$-$ 0.15	$-$ 0.01	$-$ 0.73	0.06
EXR	0.04	$-$ 0.36	$-$ 0.37	$-$ 0.16	0.08	0.00	$-$ 0.10

	Importer				Exporter
	China	U.S.	Japan	India	Brazil	Canada	Russia
IOI	$-$ 0.04	$-$ 0.13	0.08	0.02	$-$ 0.11	0.08	$-$ 0.11
EPU	0.60	$-$ 0.23	$-$ 0.44	$-$ 0.03	$-$ 0.32	$-$ 0.21	$-$ 0.11
CPI	0.13	$-$ 0.28	$-$ 0.34	0.11	0.24	$-$ 0.20	$-$ 0.02
EXR	$-$ 0.03	0.04	0.26	0.02	0.01	0.18	0.06

Received	Accepted	Published
2020-10-25	2021-01-14	2021-10-25
Issue Date
2021-10-25

模态框（Modal）标题

Please choose a citation manager

Content to export

Abstract

Key words

Cite this article

1 Introduction

2 Methodologies

2.1 SVAR Model

2.2 DCC-GARCH Model

2.3 Conditional Expectation Model

3 Data Description

Table 1 Descriptive statistics of the sample series

Table 2 Unit root test and ARCH effects test for all series

4 Empirical Results

4.1 The Impact of Structural Oil Price Shocks on Macro-Economy

Figure 1 Response of industrial output, exchange rates, inflation, and economic policy uncertainty to oil supply shocks

Figure 2 Response of industrial output, exchange rates, inflation, and economic policy uncertainty to aggregate demand shocks

Figure 3 Response of industrial output, exchange rates, inflation, and economic policy uncertainty to oil-specific demand shocks

4.2 The Dynamic Dependence Between Structural Oil Price Shocks and Macro- Economy

4.2.1 Oil Supply Shocks and Macro-Economy

Table 3 Mean value of dynamic correlation sequence between the oil supply shock and the macroeconomic variables

Figure 4 Dynamic dependence between oil supply shock and IOI of importers and exporters

Figure 5 Dynamic dependence between oil supply shock and EPU of importers and exporters

Figure 6 Dynamic dependence between oil supply shock and CPI of importers and exporters

Figure 7 Dynamic dependence between oil supply shock and EXR of importers and exporters

4.2.2 Aggregate Demand Shocks and Macro-Economy

Table 4 Mean value of dynamic correlation sequence between the aggregate demand shock and the macroeconomic variables

Figure 8 Dynamic dependence between aggregate demand shock and IOI of importers and exporters

Figure 9 Dynamic dependence between aggregate demand shock and EPU of importers and exporters

Figure 10 Dynamic dependence between aggregate demand shock and CPI of importers and exporters

Figure 11 Dynamic dependence between aggregate demand shock and EXR of importers and exporters

4.2.3 Oil-Specific Demand Shocks and the Macro-Economy

Table 5 Mean value of dynamic correlation sequence between the oil-specific demand shock and the macroeconomic variables

Figure 12 Dynamic dependence between oil-specific demand shock and IOI of importers and exporters

Figure 13 Dynamic dependence between oil-specific demand shock and EPU of importers and exporters

Figure 14 Dynamic dependence between oil-specific demand shock and CPI of importers and exporters

Figure 15 Dynamic dependence between oil-specific demand shock and EXR of importers and exporters

4.3 The Dynamic Impact of Structural Oil Prices Shocks on Macro-Economy

Figure 16 Dynamic impacts of oil supply shocks on IOI of importers and exporters

Figure 17 Dynamic impacts of oil supply shocks on EXR of importers and exporters

Figure 18 Dynamic impacts of oil supply shocks on CPI of importers and exporters

Figure 19 Dynamic impacts of oil supply shocks on EPU of importers and exporters

Figure 20 Dynamic impacts of aggregate demand shocks on IOI of importers and exporters

Figure 21 Dynamic impacts of aggregate demand shocks on EXR of importers and exporters

Figure 22 Dynamic impacts of aggregate demand shocks on CPI of importers and exporters

Figure 23 Dynamic impacts of aggregate demand shocks on EPU of importers and exporters

Figure 24 Dynamic impacts of oil-specific demand shocks on IOI of importers and exporters

Figure 25 Dynamic impacts of oil-specific demand shocks on EXR of importers and exporters

Figure 26 Dynamic impacts of oil-specific demand shocks on CPI of importers and exporters

Figure 27 Dynamic impacts of oil-specific demand shocks on EPU of importers and exporters

5 Conclusions

{{custom_sec.title}}

{{custom_sec.title}}

References

{{custom_fnGroup.title_en}}

Footnotes