I. Introduction

Economic policy is an important measure of government intervention, and economic policy uncertainty (EPU) has broad and profound impacts at the macro and micro levels. It can affect household consumption and investment decisions, unemployment, and economic growth (Bernanke, 1983; Bloom, 2009). At the corporate level, EPU affects corporate behaviors (Gulen & Ion, 2015), such as risk management (Kelly et al., 2016), capital investment, and spending (Kahle & Stulz, 2013; Kim & Kung, 2016), then transmitting to the stock market. The findings from prior research show that the EPU index has a significantly negative relation with stocks (Ko & Lee, 2015) and can be used to predict future returns in the financial market (Brogaard & Detzel, 2015). Volatility in the financial market is often accompanied by changes in economic activities. An efficient and stable stock market is indispensable for industrial development and economic growth. Therefore, a study of the impacts of EPU on the stock market should be of great significance.

This study focuses on the relation between EPU and the energy stock market in China. According to previous literature, the effect of EPU depends on the country, the strength of the economy, the size of the stock market (Christou et al., 2017), and the type of industry (Boutchkova et al., 2012; Yu et al., 2017). The increasingly severe global energy problem has received unprecedented attention. Energy is the lifeline of the national economy and national security. Countries attach great importance to the formulation of an energy economy strategy. Based on the close relation between the energy market and economic growth, it is very important to understand the impact of EPU on the energy stock market.

Theoretically, the energy market is more sensitive to EPU than to other forms of uncertainty. Energy companies are characterized by their large scale and long payback period, which entail higher risks than in other industries. EPU can not only affect the behavior of enterprises mentioned above, but also directly cause fluctuations in energy prices (oil, coal, and renewable energy), further causing fluctuations in the energy market supply and demand and changing energy consumption and production activities. Shafiullah et al. (2021) show a long-term negative correlation between US EPU and renewable energy consumption. Qin et al. (2020) show that the uncertainty of US economic policy is closely related to the oil market. However, Appiah-Otoo’s (2021) findings indicate that EPU has a nonsignificant negative effect on renewable energy growth. Therefore, the effect of EPU is related to national conditions. As an important country in terms of energy, the impact of China’s EPU on the energy market cannot be ignored. This study aims to address this issue.

In this study, we use a structural vector autoregression (SVAR) model to examine the effect of EPU on the energy market. We use the CSI 300 Index and the CSI Energy Index as proxy variables for China’s stock market and energy stock market, respectively. In addition, we adopt the widely used EPU index constructed by Baker et al. (2016) to measure the degree of EPU in China.

Our study provides more empirical evidence of the impact of EPU on the stock market and supplements the research. We innovatively examine the impacts of EPU from the perspective of China’s energy market, which is an important part of the world’s energy market. Furthermore, our study provides a more accurate reference benchmark for market investors, financial regulators, and policy makers for investment portfolio decisions, risk management, and policy formulation and evaluation. In addition, in future research, the difference between the effects of EPU on traditional energy enterprises and renewable energy enterprises is worth further discussion, because the two different types of enterprises differ greatly in technology and industrial growth. Moreover, the transmission mechanism of the effect of EPU on the energy market is worthy of further discussion in future research.

This paper is organized as follows. Section II introduces the data and methodology. Section III discusses the results, and Section IV concludes the paper.

II. Data and Methodology

A. Data and Variables

We adopt the EPU index constructed by Baker et al. (2016) to measure the EPU for China. Baker et al. constructed a scaled frequency count of articles on policy-related economic uncertainty in the South China Morning Post, Hong Kong’s leading English-language newspaper. Our method follows their news-based indexes of EPU for the United States and other countries, which provides convenient conditions for research. In addition, we use the China Shanghai–Shenzhen 300 index (CSI300) and energy stock index (ESI) to measure the price levels of China’s stock market and energy stock market, respectively. The sample data in this article include monthly data from July 2007 to June 2021.


To test the relation between EPU and the energy stock market in China, we use a structural vector autoregressive (SVAR) model for estimation. Before the SVAR analysis, we first test the stationarity of the variables, and find that each variable is stable after first difference. After a unit root test, we perform first differencing on the variables to ensure that the time series variables ultimately used in the model are stable, to avoid a false regression. The SVAR model is as follows:

\[Y_{t} = \sum_{i = 1}^{p}{A_{i}Y_{t - i}} + Au_{t} \tag {1}\]

\[\text{Au}_{t} = Bԑ_{t} \tag {2}\]

where \(Y_{t}\) denotes a vector of three variables \(({EPU}_{t},\ {CSI300}_{t},\ {ESI}_{t})\); \(A_{i}\) is the unknown coefficient matrix to be estimated; and \(u_{t}\) and \(ԑ_{t}\), respectively, denote a simplified perturbation term and a structural disturbance term for \(Y_{t}\). Thus, given these restrictions on the A and B matrices, the model can be written in matrix form as follows:

\[\begin{align} \begin{bmatrix} 1 & 0 & 0 \\ a_{21} & 1 & 0 \\ a_{31} & a_{32} & 1 \\ \end{bmatrix}\left\lbrack d\begin{matrix} {dEPU} \\ CSI300 \\ t{dESI} \\ \end{matrix} \right\rbrack \\ = \begin{bmatrix} b_{11} & 0 & 0 \\ 0 & b_{22} & 0 \\ 0 & 0 & b_{33} \\ \end{bmatrix}\begin{bmatrix} Ԑ_{{dEPU}} \\ Ԑ_{dCSI300} \\ Ԑ_{{dESI}} \\ \end{bmatrix} \end{align}\tag{3}\]

\[dEPU = b_{11}Ԑ_{\text{dEPU}} \tag{4}\]

\[ dCSI300 = -a_{21}{dEPU} + b_{22}Ԑ_{dCSI300} \tag{5}\]

\[ dESI = -a_{31}{dEPU}\ - a_{32}\text{dCSI}300 + b_{33}Ԑ_{{dESI}} \tag {6}\]

Specifically, \(a_{21}\) and \(a_{31}\) reflect the impacts of EPU on the CSI 300 Index (CSI300) and the energy stock market index (ESI), respectively. Moreover, considering that the energy stock market is affected by the overall stock market level, we set the coefficient \(a_{32}\) to test the effect of CSI300 on the energy stock market.

III. Results

Table 1 reports the descriptive statistics of all the variables, where \(dEPU\), \(dCSI300\), and \(dESI\) are the first difference sequences of \(EPU\), \(CSI300\), and \(ESI\), respectively. Table 1 shows that the mean of the EPU index (\(EPU\)) is 305.513, the maximum and minimum are, respectively, 26.144 and 970.830, and the standard deviation is 247.697, indicating that strong EPU volatility, which is consistent with reality. Furthermore, the maximum, minimum, and standard deviation of \(dESI\) are larger than for \(dCSI300\), which suggests that the monthly energy stock index data fluctuate more than the CSI 300 data.

Table 1.Descriptive Statistics
Variable Mean SD Min P50 Max
\(EPU\) 305.513 247.697 26.144 213.549 970.830
\(dEPU\) 1.709 99.167 -280.348 7.662 257.773
\(CSI300\) 3346.436 869.504 1800.775 3280.889 5624.885
\(dCSI300\) 7.390 228.454 -700.865 16.048 817.780
\(ESI\) 2525.190 1188.442 1201.678 1997.384 7204.880
\(dESI\) -17.017 262.623 -817.860 -6.830 1413.660

This table presents descriptive statistics of all variables. In addition, dEPU, dCSI300 and dESI are the first-order difference sequences of EPU, CSI300 and ESI, respectively, which are stationary.

Table 2 reports the optimal lag order selected based on the final prediction error (FPE) criterion, Akaike information criterion (AIC), Hannan–Quinn information criterion (HQIC), and Bayesian information criterion (BIC). It is important to construct the SVAR model with a reasonable lag order: if the lag order is too small, the estimation results will be inconsistent, and if the lag order is too large, too many parameters will be estimated, which will affect the effectiveness of the model estimation. According to the HQIC and BIC, the optimal lag order of the model is one; however, the AIC and FPE criterion show that the optimal lag order of the model is two. Furthermore, according to likelihood ratio statistics, the optimal lag order is two and the impact of EPU cannot be fully reflected in the short term; therefore, we choose an optimal lag order of two.

Table 2.Results of the optimal lag order selection
0 -3120.77 8.9e+12 38.3284 38.3515 38.3854
1 -3089.3 62.924 9 0.000 6.7e+12 38.0528 38.1453* 38.2806*
2 -3075.67 27.269* 9 0.001 6.4e+12* 37.9959* 38.1578 38.3945
3 -3071.33 8.6763 9 0.468 6.7e+12 38.0531 38.2843 38.6225
4 -3064.36 13.932 9 0.125 6.9e+12 38.0781 38.3786 38.8183

This table reports the selection results of the optimal lag order, mainly referring to FPE, AIC, HQIC and SBIC.

Table 3 reports the estimation results of the SVAR model (Panel A) and robustness test (Panel B). The key coefficient \(a_{31}\) is significantly positive at the 5% level (\(-a_{31}\)is negative), indicating that EPU has a negative impact on the energy stock market. EPU can affect the consumption and investment behavior of micro-subjects through expected effects and then be transmitted to stock prices. Specifically, EPU has an impact on the stock market by affecting investors’ and companies’ future expectations. Investors’ future expectations will change with policy uncertainty. When the degree of uncertainty increases, investors will require more risk premiums in the future, which will affect stock market prices. In addition, the coefficient \(a_{32}\) is positive, which indicates that the overall stock market has a positive impact on the energy stock market.

Table 3.Estimation results of SVAR model
Panel A: SVAR results
Coef. St. Err. z P>z [95% Confidence Interval]
Matrix A
\(a\) 21 0.265 0.180 1.47 0.141 -0.087 -0.616
\(a\) 31 0.214 0.106 2.01 0.044 0.005 0.422
\(a\) 32 -0.729 0.046 -15.93 0.000 -0.819 -0.639
Matrix B
\(b\) 11 97.928 5.391 18.17 0.000 87.362 108.493
\(b\) 22 225.869 12.434 18.17 0.000 201.499 250.239
\(b\) 33 132.781 7.309 18.17 0.000 118.455 147.107
Panel B: Robustness test
Matrix A
\(a\) 21 -1.201 0.351 -3.42 0.001 -1.889 -0.513
\(a\) 31 0.156 0.086 1.81 0.070 -0.013 0.326
\(a\) 32 -0.314 0.019 -16.18 0.000 -0.352 -0.276
Matrix B
\(b\) 11 103.471 5.994 17.26 0.000 91.723 115.219
\(b\) 22 443.584 25.696 17.26 0.000 393.220 493.947
\(b\) 33 105.091 6.088 17.26 0.000 93.159 117.022

This table reports the estimated coefficients of A and B matrices. The statistics, Coef., St. Err., z, and P>z, denote, respectively, coefficient, standard error, z-statistic, and p-value.

Finally, we conduct a robustness test to ensure the credibility of our results. We replace the energy stock index variable \(ESI\) with \(SZESI\), an alternative proxy for the Chinese energy stock price, and we obtain results similar to those presented in Table 3.

IV. Conclusion

From the perspective of the energy market and based on China’s policy environment, this paper examines the impact of EPU on the stock market. We use the EPU index constructed by Baker et al. (2016) and data from China’s stock market to empirically test the relation between EPU and the energy stock market. Based on a SVAR model, we find that an increase in EPU will reduce energy stock prices and that the overall stock market situation has a positive effect on the energy stock market.

Future research on China’s policy environment should further explore the mechanism of the impact of EPU on the stock market.