International Studies of Economics
Early View
RESEARCH ARTICLE
Open Access

Effect of Climate Changes, Induced Risks, and Oil Price Appreciation on Energy Stock Returns in World Markets

Thomas C. Chiang, 

Corresponding Author

Thomas C. Chiang

Department of Finance, Drexel University, Philadelphia, Pennsylvania, USA

Correspondence: Thomas C. Chiang ([email protected])

Search for more papers by this author
First published: 23 April 2025
https://doi.org/10.1002/ise3.70003

ABSTRACT

This study examines the impact of climate policy uncertainty (CPU) on world energy stock returns. Evidence shows that a rise in CPU causes stocks to plummet in individual countries, regions, and the world energy stock markets. The negative effects are also exhibited in climate induced risks, the covariance between a change in CPU and equity market volatility (EMV) as well as the covariance between energy and environmental uncertainty and EMV on stock returns. Evidence shows the presence of negative relationships between oil prices and stock returns, except in Gulf Cooperation Council region and Kuwait, which are oil-exporting markets.

1 Introduction and Literature

Since the Biden Administration rejoined the Paris Agreement in 2021, substantial attention has been given to reducing greenhouse gas emissions to meet the goal of carbon neutrality by 2050. This abrupt turnaround in US policy needs to be considered not only in terms of the costs to execute the policy such as implementing a carbon tax in reducing emissions, but also in terms of its effect on climate policy-driven financial stability risk (Carattini et al. 2023). President Trump is withdrawing from the Paris Agreement again, reversing US climate policy. This back-and-forth shift in climate policy creates not only a substantial political risk but also waste of resources and creates uncertainty about future directions.

The US Environmental Protection Agency (EPA) divides climate risks into two categories: physical and transformational risks. Physical risks refer to acute climatic events, including flooding, extreme heat, and chronic climatic & environmental events, involving losses to physical property that result from climate change (Batten 2018; Boushey et al. 2021; Xu et al. 2023). In particular, Hong et al. (2019) investigated the impact of extreme weather and found that the stocks with greater exposure to drought risk tend to receive lower returns due to market participants' underreacting to climate risk. Bansal et al. (2016) examined the impact of temperature fluctuations attributable to global warming. They found evidence that temperature risks have a negative effect on stock market valuation. Barberà-Mariné et al. (2023) showed that carbon emissions have a negative effect on stock returns.

Transition risks are associated with the process of moving away from a reliance on fossil fuels and transitioning toward a low-carbon economy, including the risk of changes from climate legislation, technology, and market, investor and consumer sentiment as it pertains to a greener environment (Persefoni Report 2024). Climate policy uncertainty ( CPU $\text{CPU}$ ) stems from the lack of clarity regarding future government policies to address climate changes, new forms of energy and the uncertainty of environment policy that can cause significant energy market volatility (Salisu et al. 2023) as business firms and investors to switch from traditional fossil fuels and adapt to new forms of renewable energy. This uncertainty can even be transmitted to stock markets and spillover global markets. Diaz-Rainey et al. (2021) studied climate policy risks and find that it can upward shift in transition risks in the oil sector and renewable energy competitiveness, resulting in a decrease in profits for US non-renewable energy industries. Fried et al. (2021) found evidence that climate policy risk reduces the expected return of fossil capital relative to clean capital, shifting the economy toward cleaner production. Yao et al. (2023) emphasized that CPU's effect on stock returns can be explained by the spillover induced by the production network. In analyzing climate policy uncertainty (CPU), Guo et al. (2022) discovered that CPU has a time-varying effect on energy prices, moving from positive to negative coefficients, which are statistically significant. He and Zhang (2022) employ a news-based climate policy uncertainty (CPU) index (Gavriilidis 2021) to test the stock return predictability of the oil industry and demonstrate that CPU is a good predictor of future oil industry stock returns. Tedeschi et al. (2024) investigated the effect of climate policy uncertainty on European stock prices and find that the CPU shock produces a positive impact on clean energy stock prices. Xu et al. (2023) investigated the US CPU and China's CPU for both stock return and volatility. Their evidence from the Chinese market indicates that high CPU decreases current stock market returns and increases volatility. However, for the US market, an increase in CPU decreases stock returns in the short term but increases it in the long term. Yet, an upward shift in CPU in the US contributes to a rise in volatility in the short term but a mixed result in volatility at the latter point of time. Lasisi et al. (2024) examined the relationship between climate policy uncertainty (CPU) and stock market volatility and found that the slope of the CPU is significantly positive, suggesting that higher CPU values tend to heighten stock market volatility in both the US and UK.

Climate change risks are found to have a spillover effect on energy market uncertainty. For instance, Ye (2022) analyzed the effect of the WSJ climate news index (CNI) and crude oil volatility index (OVX) and concluded that CNI has a negative impact on energy market uncertainty. Salisu et al. (2023) employed a mixed data sampling of the Generalized Autoregressive Conditional Heteroscedasticity (GARCH-MIDAS) model to examine the predictive value of CPU for crude oil market volatility and confirmed that including the CPU in the predictive model of oil market volatility provides a better out-of-sample prediction compared with a benchmark model that excludes the CPU. Applying a similar procedure, Hoque et al. (2023) found that world energy stocks and carbon emissions futures are well connected to US climate policy uncertainty. The evidence indicates that volatility of world energy stocks and carbon emissions futures could be transmitted through the energy sector in the US market. However, no empirical study has been done to connect the impact of the change in US CPU on different countries and regions. Apparently, the issue that connects change in CPU to energy market (volatility) has not been explicitly incorporated into the traditional empirical study on the stock return and oil price relationship and this gap should be filled in.

Recent studies also intended to incorporate different forms of uncertainty when testing the stock return equation. Xiao and Liu (2023) tested the CPU effect as well as other uncertainty variables such as economic policy uncertainty (EPU) (Chiang 2019) and equity market volatility (EMV) (Dutta et al. 2021) on stock markets; their evidence shows that CPU has a more profound effect in triggering oil market fears in recent years. However, during the COVID-19 pandemic, CPU, EPU, and EMV played a significant role in triggering oil market fears and causing stock markets to plunge (Tedeschi et al. 2024), suggesting that the impact of the 2008 global financial crisis and the 2020 pandemic should be clearly identified in the empirical analysis.

The above literature provides several important insights that will guide the current research. In addition to modeling the direct effect from a change in CPU on stock returns, the literature readily argues that a rise in climate policy uncertainty can disturb oil prices and create equity market volatility. This spillover effect on stock returns is due to the anxiety and fears that spawn a negative effect. A special feature of this study is the incorporation of these induced risks expressed as the covariance between a change in CPU ( ∆ cpu t $\unicode{x02206}{{cpu}}_{t}$ ) and equity market volatility ( cov σ , ∆ cpu , t ) $({\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t})$ and the covariance between energy and environmental policy changes and equity market volatility ( cov σ , ∆ EN , t ) $({\text{cov}}_{\sigma ,\unicode{x02206}{EN},t})$ while controlling for extreme market conditions due to the high volatility associated with 2008 the global financial crisis ( σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ ) and the 2020 COVID-19 pandemic ( σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ).

Furthermore, the model will be examined using data from 15 countries across 7 different regions. Thus, this study presents a comprehensive empirical model, which encompasses uncertainty factors changes { ∆ cpu t $\unicode{x02206}{{cpu}}_{t}$ , Cov σ , ∆ cpu , t , ∆ p oil , t ${{Cov}}_{\sigma ,\unicode{x02206}{cpu},t},{\unicode{x02206}p}_{{oil},t}$ , cov σ , ∆ EN , t , ${\text{cov}}_{\sigma ,\unicode{x02206}{EN},t},$ σ GFC , t 2 , ${\sigma }_{{GFC},t}^{2},$ and σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ } consistent with the argument of Barnett et al. (2020). In this way, the model can better capture climate policy uncertainty and related risks, shedding more light on the prevailing global financial market behavior.

This study makes several empirical findings by examining monthly data from January 1997 to December 2023. First, this study finds a negative relationship between energy stock returns and changes in lagged climate policy uncertainty ( ∆ cpu t − 1 and ∆ cpu t − 2 ) ${\unicode{x02206}{cpu}}_{t-1}\text{and}{\unicode{x02206}{cpu}}_{t-2})$ for individual countries, regional data and world stock returns. Second, with some minor exceptions, evidence suggests a negative relationship between stock returns and ∆ $\unicode{x02206}$ cpu as well as induced covariance ( cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ and cov σ , ∆ cpu , t − 2 ) . ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}).$ Third, evidence supports the negative relationship between stock returns and energy and environmental-induced volatility ( cov σ , EN , t ) ${(\text{cov}}_{\sigma ,{EN},t})$ with time lags. Fourth, evidence reveals that changes in lagged oil prices and energy stock returns are negatively correlated, except in oil-exporting markets, such as GCC and Kuwait. Fifth, further tests on the control variables, including the 2018 financial crises and 2020s COVID-19, find a negative effect on stock returns. Sixth, the model is robust, whether the dependent variable is based on aggregate market stock returns or energy stock returns. An equivalent result is achieved whether we test the relationship between stock return and changes in oil prices or between stock return and changes in gasoline prices.

The remainder of this paper is organized as follows. Section 2 presents an econometric model that features a multiple-dynamic regression model in a GED-GARCH process to evaluate the underlying hypotheses. Section 3 describes the data and related variables used in the empirical estimation. Section 4 presents empirical evidence for individual countries in the global market. Section 5 conducts robustness tests to examine the relationship between aggregate stock returns and the same set of independent variables. Section 6 reports the test equation by applying the data on regional and world markets. Section 7 concludes with the empirical findings and the practical implications.

2 The Model

The discussions in the previous section provide directions for specifying a dynamic regression model to test stock returns in relation to changes in climate policy uncertainty and changes in oil prices. We begin with a simple model expressed as:
R t = α + β s ( L s ) ∆ cpu t + γ k ( L k ) ∆ p oil , t + Control + ε t ${{R}_{t}=\alpha +\beta }_{s}({L}^{s}){\unicode{x02206}{cpu}}_{t}+{\gamma }_{k}({L}^{k}){\unicode{x02206}p}_{{oil},t}+{Control}+{\varepsilon }_{t}$ ()
where R t ${R}_{t}$ is the aggregate (or energy) stock market return at time t, ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ is the change in climate policy uncertainty in the natural logarithm (expressed in a lower case) at time t. ∆ p oil , t ${\unicode{x02206}p}_{{oil},t}$ is the change in Brent crude oil price. C ontrols t ${C{ontrols}}_{t}$ is the control variable, which may consist of equity market volatility during the 2008 financial crisis, σ FC , t 2 ${\sigma }_{{FC},t}^{2}$ , and equity market volatility associated with the 2020 COVID-19 period, σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ . L s ${L}^{s}$ and L k ${L}^{k}$ are lagged operators with lagged length of s and k periods, respectively. α $\alpha $ , β $\beta $ , and γ $\gamma $ are constant parameters. Equation (1) can be viewed as a dynamic regression model with input variables ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ , ∆ p oil , t , σ GFC , t 2 ${\unicode{x02206}p}_{{oil},t},{\sigma }_{{GFC},t}^{2}$ , and σ ID , t 2 , ${\sigma }_{{ID},t}^{2},$ where the output variable, R t ${R}_{t}$ , is a function of multi-input variables, where β s ( L s ) ${\beta }_{s}({L}^{s})$ and γ k ( L K ) ${\gamma }_{k}({L}^{K})$ are linear (or rational) polynomials of L s ${L}^{s}$ and L K ${L}^{K}$ , which are of finite order (Wei 2018). The error term, ε t , ${\varepsilon }_{t},$ → $\to $ │ I t − 1 ∽ GED ( 0 , σ t − 1 2 , ν ) ${{\rm{I}}}_{t-1}\backsim \text{GED}(0,{\sigma }_{t-1}^{2},\nu )$ .

There are good reasons to include both ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and ∆ p oil , t ${\unicode{x02206}p}_{{oil},t}$ enter into an equation to explain stock returns. A rise in oil prices reflects an increase in demand for oil consumption as advancement in economic activity, which is accompanied by producing sizable greenhouse gases (GHGs) from the combustion of more fossil fuels. With existing technologies that continually burn fossil fuels that bring about increases in temperatures, which induces the likelihood of motivating government policies to reduce emissions (Batten 2018). As a result, we observe a positive correlation between ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and ∆ p oil , t ${\unicode{x02206}p}_{{oil},t}$ . However, the relationship between them may not occur contemporaneously as different reactions involving in timing for different economic agents. Complications may arise due to economic agents' concern about the changes in climate risk that lead to the possibility of tax increases on emissions, causing the development adaptation to green energy versus fossil sources of energy. In the long run, this ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ certainly would alter the relative prices between green industries and fossil use commodity prices. Thus, investors who hold stocks in industries using fossil fuel are likely to demand a risk premium due to fears of changes in climate policy and potential higher carbon taxes imposed by the government (Fried et al. 2021). The fear of higher taxes likely causes equity market volatility and financial instability, resulting in a lower stock prices.

The dynamic model in Equation (1) is perceived to have time lags based on the following justification: First, economic agents' response to news arrival may not be immediate due to market friction; second, investors may pursue a real options theory (Pindyck 1991), which posits that a rational investor may give up the option to make decision now and wait to receive better information to act (strike) (Pindyck 1991). As a result, the model specification should include time lags. Third, literature (Chen et al. 2023) suggests that as policy uncertainty in the climate domain contains useful information for predicting risk in the stock return equation.

To capture the insight into the dynamics, Equation (1) is expressed as a regression:
R t = C + β 1 ∆ cpu t − 1 + β 2 ∆ cpu t − 2 + β 3 ∆ p oil , t − 1 + β 4 ∆ p oil , t − 2 + β 5 σ GFC , t 2 + β 6 σ I D , t 2 + ε t ${R}_{t}=C+{\beta }_{1}\unicode{x02206}{{cpu}}_{t-1}+{\beta }_{2}\unicode{x02206}{{cpu}}_{t-2}+{\beta }_{3}\unicode{x02206}{p}_{{oil},t-1}+{\beta }_{4}\unicode{x02206}{p}_{{oil},t-2}+{\beta }_{5}{\sigma }_{{GFC},t}^{2}+{\beta }_{6}{\sigma }_{ID,t}^{2}+{\varepsilon }_{t}$ ()
where the lagged ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and ∆ p oil , t ${\unicode{x02206}p}_{{oil},t}$ lengths are determined by the Final Prediction Error and Akaike criterion (Hsiao 1981; Wei 2018; Tsay 2002). The restrictions of the parameters and, hence, the hypotheses can be briefly given below. First, β 1 ${\beta }_{1}$ < 0, β 2 ${\beta }_{2}$ < 0 if the net present value (NPV) of assessing ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ impact on stock price is negative. The net present value is based on an evaluation that discounts future income streams and subtracts the costs from emissions and social costs of carbon (Barnett et al. 2020). Barnett et al. (2020) suggested using the subjective discount rate to calculate the net present value. However, in practice, we may use the market rate of interest plus the expected inflation rate and risk premiums associated with different risks/uncertainties as a measure of the discount rate. In the vein of the theory of asset pricing involving multiple shocks (uncertainties) to the system, differential exposure to these shocks (uncertainties) should be discounted in different risk premiums encoded in stochastic discount factors. In a model with a broad economic perspective, Barnett et al. (2020) suggest that the uncertain social costs should also be accounted for as a discount factor.

Second, the study sets β 3 ${\beta }_{3}$ < 0 and β 4 ${\beta }_{4}$ < 0 as noted by Hamilton (1996) and Hwang and Kim (2021), who argue that a rise in oil prices has a negative effect on stocks due to oil price shocks, which causes output production to decline, precipitating the onset of an economic recession that eventually weakens stock returns. This view is consistent with the findings of Jones and Kaul (1996), who showed a significant negative relationship between oil price movements and stock returns. However, Tembelo and Ozyesil (2024) found a statistically significant positive association between stock returns and oil prices in OECD countries. Alamgir and Amin (2021) found a positive relationship between world oil prices and the stock market index in four selected South Asian countries. Empirical evidence with mixed signs provides no direction for both β 3 ${\beta }_{3}$ and β 4 ${\beta }_{4}$ . Third, β 5 ${\beta }_{5}$ < 0 and β 6 ${\beta }_{6}$ < 0 for the control variables are anticipated to present negative signs due to the impact of unusual economic conditions arising from the 2008 GFC and the 2020 COVID-19 pandemic. By using indicator variable, where I GFC , t ${I}_{{GFC},t}$ is set to unity for the event dates for the period of September 2008 – April 2009 and I ID , t   for the period ${I}_{{ID},t}\ \text{for the period}\,$ March 2020 – September 2022 is set to unity and zero otherwise (Batten et al. 2021; Cheema et al. 2022), we derive that σ GFC , t 2 = EMV GFC · I GFC , t ${\sigma }_{{GFC},t}^{2}={{EMV}}_{{GFC}}\cdot {I}_{{GFC},t}$ and σ I D , t 2 = EMV ID · I ID , t ${\sigma }_{ID{,}t}^{2}={{EMV}}_{{ID}}\cdot {I}_{{ID},t}$ . This treatment was chosen for this empirical study to control extreme behavior during the crisis period (Peña 2001). Note that evidence provided by Prabheesh et al. (2020) suggests a positive co-movement between oil price returns and stock price returns during the COVID-19 pandemic in China, India, Japan, and Korea. The evidence may result from the fact that there were no controls for the uncertainty variables σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ and σ I D , t 2 ${\sigma }_{ID{,}t}^{2}$ in their tests. At any rate, the signs of β 5 ${\beta }_{5}$ and β 6 ${\beta }_{6}$ can be mixed.

To complete the model, the variance equation is assumed to follow a GARCH(1,1) process recommended by Bollerslev et al. (1992) and the representation is given by
σ i , t 2 = ω 0 + ω 1 ε i , t − 1 2 + ω 2 σ i , t − 1 2 , ${\sigma }_{i,t}^{2}={{\rm{\omega }}}_{0}+{{\rm{\omega }}}_{1}{\varepsilon }_{i,t-1}^{2}+{{\rm{\omega }}}_{2}{\sigma }_{i,t-1}^{2},$ ()
where σ i , t 2 ${\sigma }_{i,t}^{2}$ is the conditional variance and ε i , t 2 ${\varepsilon }_{i,t}^{2}$ is the shock squared. The Equation (3) describes the conditional variance evolving by using the information of the lagged σ i , t 2 ${\sigma }_{i,t}^{2}$ and lagged ε i , t 2 , ${\varepsilon }_{i,t}^{2},$ featuring a time-varying process. This feature helps to model the volatility clustering phenomenon and relax the assumption of constant variance.
Following Nelson (1991), the error term obeys the GED distribution expressed as:
ε i , t │ Φ t − 1 ∽ GED ( 0 , σ i , t − 1 2 , ν ) ${\varepsilon }_{i,t}{\rm{\unicode{x02502}}}{{\rm{\Phi }}}_{t-1}\backsim \text{GED}(0,{\sigma }_{i,t-1}^{2},\nu )$ ()
The rationale for using this GED distribution stems from the error series, which can be smoothly transformed to deal with leptokurtotic distribution (fat tails) or a platykurtotic distribution (thin tails). It follows that the GED can work with heavy tails of a distribution, especially the case for the stock return series in facing dramatic movements. (Nelson 1991; Chiang 2023). Further, the shape parameter ( ν $\nu $ ) has high flexibility and allows the GED to adapt to different levels of kurtosis (peakiness or flatness) observed in the real data. This can be illustrated by the GED density function given by:
f ( z; ν ) = ν · exp − 1 2 | z / λ | ν λ 2 ( 1 + 1 ν ) Γ ( 1 / ν ) , $f({\rm{z;}}\nu )=\frac{\nu \,\cdot \,\text{exp}\left[-\tfrac{1}{2}{|z/\lambda |}^{\nu }\right]}{\lambda {2}^{(1+\frac{1}{\nu })}{\rm{\Gamma }}(1/\nu )},$ ()
with λ ≡ 2 ( − 2 / ν ) Γ 1 ν Γ ( 3 / ν ) 1 / 2 , $\text{with}\,\lambda \equiv {\left(\frac{{2}^{(-2/\nu )}{\rm{\Gamma }}\left(\frac{1}{\nu }\right)}{{\rm{\Gamma }}(3/\nu )}\right)}^{1/2},$ ()
where Γ ( · ) $\text{where}\,{\rm{\Gamma }}(\cdot )$ is the gamma function, λ is related $\lambda \,\text{is related}$ to the rate of decay of the tails of the distribution. When ν = 2 $\nu =2$ , the GED reduces to a standard Normal distribution, when ν = 1 , $\nu =1,$ it becomes a Laplace distribution. The property of flexibility with GED is crucial for capturing the fat tails and varying levels of kurtosis observed in the stock return series data (Nelson 1991, 352; Chiang 2023).

3 Data Description

The data in this study cover monthly observations from January 1997 through December 2023 for 15 markets. The data contain energy stock price indices and oil prices for individual countries: United States (US), Canada (CA), United Kingdom (UK), France (FR), Germany (GM), Italy (IT), and Japan (JP). These are the major industrial markets and non-G7 markets: China (CN), South Korea (KO), Kuwait (KW), Indonesia (ID), India (IN), Singapore (SG), Brazil (BR), and Mexico (MX). The data also cover country aggregate market, regional and world stock indices: Developed market (DEVL), emerging market (EMER), Pacific region (PAC), Asian excludes the Pacific area (A_XPAC), the Gulf Cooperation Council (GCC), The Economic and Monetary Union (EMU), Latin American (LATIN), and world (WORLD) indices. The selection of these groups of markets allows us to analyze the impact on different regions, which may capture various economic structures and market behaviors. The data length is subject to availability. The data were obtained from the Data Stream and economic data base of the Federal Reserve Bank of St. Louis.

The returns of various stock indices and the growth in Brent crude oil price indices are transformed by taking the first difference in natural logarithm times 100. As shown in Table 1, Panels A and B report the statistical results for energy stock returns, which range from −0.817% (GM) to 1.392% (MX) and have a spread in standard deviations from 6.152 (IT) to 20.079 (ID). To check the normality distribution of the energy stock return series, the Jarque-Bera statistics (JB) were used. The results, which range from 20.599 (UK) to 2152.427 (SG), are much greater than the critical value at the 5% level (=5.991) with 2 degrees of freedom, allowing us to reject the null of normality for all the markets. The evidence supports the use of the GED-GARCH model. In Panel C, which reports the statistics for regional and world energy stock returns, the mean values range from to 0.183% (PAC) to 0.498% (LATIN). The spread of standard deviations ranges from 6.322 (PAC) to 10.953 (LATIN). However, the WORLD standard deviation is 6.178, reflecting lower variability among the members of individual markets. The Jarque-Bera statistics (JB) range from 43.514 (EMU) to 2904.726 (GCC) and the WORLD level is 227.065, indicating the rejection of the null of normality for all the markets.

Table 1. Summary statistics for energy stock returns in individual countries and regional performance: February 1997–December 2023.
A. G7 markets
R_US R_CA R_FR R_GM R_IT R_UK R_JP
Mean 0.396 0.313 0.354 −0.817 0.145 0.239 −0.063
Median 0.464 0.615 0.438 −0.263 0.426 0.532 0.591
Maximum 27.858 22.463 31.558 32.509 24.747 26.180 21.706
Minimum −46.248 −39.579 −23.411 −48.219 −21.946 −20.281 −35.123
Std. Dev. 6.925 6.399 6.265 10.429 6.152 6.264 7.660
Skewness −0.804 −0.916 0.302 −0.462 −0.035 −0.013 −0.740
Kurtosis 9.654 7.911 5.388 4.855 4.239 4.237 4.493
J-Bera 630.712 369.743 81.685 37.754 20.727 20.599 58.180
Obs 323 323 323 211 323 323 316
B. Non-G7 markets
R_CN R_KO R_KW R_ID R_IN R_SG R_BR R_MX
Mean 0.393 0.542 −0.072 0.762 0.805 0.125 0.923 1.392
Median 0.244 0.234 0.000 0.000 0.833 0.113 1.313 1.155
Maximum 62.641 48.393 32.022 109.909 30.895 51.032 39.401 34.067
Minimum −41.965 −36.224 −26.634 −69.315 −38.626 −80.374 −57.319 −90.142
Std. Dev. 11.111 10.986 7.581 20.079 8.731 10.722 11.191 17.484
Skewness 0.528 0.135 0.070 1.037 −0.059 −0.561 −0.551 −2.111
Kurtosis 7.280 4.291 4.927 9.077 6.276 15.597 6.054 12.310
J-Bera 261.568 23.397 37.338 554.914 144.632 2152.427 138.812 330.918
Obs 323 323 240 323 323 323 316 76
C. Regional and world markets
R_DEVEL R _EMER R _GCC R_ A_XPAC R_PAC R_EMU R_ LATIN R_World
Mean 0.312 0.497 0.457 0.450 0.341 0.183 0.498 0.310
Median 0.433 1.505 0.025 0.883 0.839 0.585 0.948 0.480
Maximum 23.336 22.349 78.977 21.767 17.878 30.101 29.700 15.527
Minimum −35.960 −41.516 −51.313 −35.996 −31.924 −21.289 −64.194 −34.685
Std. Dev. 6.422 7.799 10.937 6.897 6.915 6.322 10.953 6.178
Skewness −0.642 −1.147 1.800 −0.563 −0.530 −0.092 −0.845 −0.878
Kurtosis 6.439 7.802 20.433 5.986 4.674 4.789 6.849 6.713
Jarque-Bera 181.392 381.195 2904.726 137.084 52.866 43.514 237.843 227.065
Obs 323 323 220 323 323 323 323 323
  • Note: Jarque–Bera statistics is a joint test for normality using x 2 ( 2 ) ${x}^{2}(2)$ distribution, which is strongly rejected. OBS denotes the number of observations.

This study employs a newspaper-based Equity Market Volatility (EMV) tracker (Baker et al. 2016, 2019) calibrated to specific state variables such as energy policy regulation, financial crisis, and infectious diseases. The EMV refers to three words that are denoted by E: {economic, economy, financial}, M: market, equity, equities, Standard and Poors (and variants)}, V: {volatility, volatile, uncertain, uncertainty, risk, risky}. The category specific EMV trackers are calculated by the share of EMV articles in each category and multiplied by the contemporaneous EMV tracker value (Baker et al. 2019; Terry et al. 2020) with a multiple factor. Categorical EMV trackers in this study include the Energy and Environmental Regulation EMV tracker ( EMV EN , t ) $({{EMV}}_{{EN},t})$ , the financial crisis $\text{the financial crisis}$ EMV Tracker ( EMV FC , t ) ${{EMV}}_{{FC},t})$ and the Infectious Disease EMV tracker ( EMV ID , t ${{EMV}}_{{ID},t}$ ). These categorical EMV t trackers and the Climate Policy Uncertainty (CPU t ) index proposed by Gavriilidis (2021) are download from https://www.policyuncertainty.com/EMV_monthly.html. For more information on Baker's methods (2016), term lists, and a discussion of term selection and related papers, please refer to Economic Policy Uncertainty (Baker et al. 2019 and updated). The time series of covariance between ∆ cpu t $\unicode{x02206}{{cpu}}_{t}$ and overall EMV t ${{EMV}}_{t}$ ( cov σ , Δ cpu , t ) ${(\text{cov}}_{\sigma ,{\rm{\Delta }}\text{cpu},t})$ was derived from the dynamic conditional correlation (DCC) procedure using the method proposed by Engle (2002) and the applications of Engle (2009) and Chiang et al. (2007) in analyzing financial time series.

To gain a brief understanding of the relationships among the different uncertainty variables, we report the correlation matrix for the variables of {CPU t , cov σ , Δ cpu , t ${\text{cov}}_{\sigma ,{\rm{\Delta }}\text{cpu},t}$ , EMV EN , t ${{EMV}}_{{EN},t}$ , EMV FC . t ${{EMV}}_{{FC}.t}$ , and EMV ID , t } ${{EMV}}_{{ID},t}\}$ in the upper panel of Table 2. The statistics indicate that the relationships between CPU t , and EMV EN , t $\text{and}\,{{EMV}}_{{EN},t}$ are positive and highly significant. A log-difference was taken from these two variables to remove a possible common time trend, and the following group was obtained and is expressed as { ∆ cpu t , cov σ , Δ cpu , t $\unicode{x02206}{{cpu}}_{t},{\text{cov}}_{\sigma ,{\rm{\Delta }}\text{cpu},t}$ , ∆ emu EN , t $\unicode{x02206}{{emu}}_{{EN},t}$ , ∆ emu FC , t , ∆ emu ID , t } $\unicode{x02206}{{emu}}_{{FC},t},\unicode{x02206}{{emu}}_{{ID},t}\}$ . The resulting correlation matrix is presented in the lower panel, and the statistics exhibit the absence of significant correlations. These variables are used in empirical analyses.

Table 2. Statistics of correlations for uncertainty measures of (change) climate policy uncertainty and its covariance with (change) equity market volatility, energy uncertainty, variance in financial crisis, and variance in infectious disease.
Corr/t-Statistic CPU t cov σ , Δ cpu , t ${\text{cov}}_{\sigma ,{\rm{\Delta }}{cpu},t}$ EMV EN , t ${{EMV}}_{{EN},t}$ EMV FC . t ${{EMV}}_{{FC}.t}$ EMV ID , t ${{EMV}}_{{ID},t}$
CPU t 1
—
cov σ , Δ cpu , t ${\text{cov}}_{\sigma ,{\rm{\Delta }}{cpu},t}$ −0.174*** 1
−3.13 —
EMV EN , t ${{EMV}}_{{EN},t}$ 0.252*** −0.029 1
4.62 −0.51 —
EMV FC , t ${{EMV}}_{{FC},t}$ −0.049 −0.106 0.216 1
−0.87 −1.90 3.92 —
EMV ID , t ${{EMV}}_{{ID},t}$ 0.526*** −0.029 0.178*** 0.049 1
10.99 −0.52 3.22 0.86 —
Corr/t-Statistic ∆ cpu t $\unicode{x02206}{{\boldsymbol{cpu}}}_{t}$ cov σ , Δ cpu , t ${{\bf{\text{cov}}}}_{\sigma ,{\rm{\Delta }}\text{cpu},t}$ ∆ emu EN , t $\unicode{x02206}{{\boldsymbol{emu}}}_{{EN},t}$ ∆ emu FC , t $\unicode{x02206}{{emu}}_{{\boldsymbol{FC}},t}$ ∆ emu ID , t $\unicode{x02206}{{\boldsymbol{emu}}}_{{ID},t}$
∆ cpu t $\unicode{x02206}{{cpu}}_{t}$ 1
—
cov σ , Δ cpu , t ${\text{cov}}_{\sigma ,{\rm{\Delta }}\text{cpu},t}$ 0.082 1
1.46 —
∆ emu EN , t $\unicode{x02206}{{emu}}_{{EN},t}$ 0.056 0.043 1
1.00 0.77 —
∆ emu FC , t $\unicode{x02206}{{emu}}_{{FC},t}$ 0.017 −0.106* 0.021 1
0.30 −1.89 0.37 —
∆ emu ID , t $\unicode{x02206}{{emu}}_{{ID},t}$ 0.030 −0.029 −0.038 0.049 1
0.53 −0.52 −0.68 0.87 —
  • Note: Values in the first row are the estimated correlation coefficients, and the second row are the t-statistics. CPU t denotes climate policy uncertainty, ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ is log-difference of climate policy uncertainty. cov σ , Δ cpu , t ${\text{cov}}_{\sigma ,{\rm{\Delta }}{cpu},t}$ is the covariance between ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and equity market volatility, EMV ENU , t ${{EMV}}_{{ENU},t}$ is the energy and environment policy induced volatility $\text{volatility}$ ; the EMV FC , t ${{EMV}}_{{FC},t}$ is the equity market volatility calibrated to financial crisis index, EMV ID , t ${{EMV}}_{{ID},t}$ is the equity market volatility calibrated to infectious disease index. ∆ emu t $\unicode{x02206}{{emu}}_{t}$ is log-different of EMV t . ${{EMV}}_{t}.$ *** and * are significant at the 1% and 10% levels, respectively.

4 Empirical Evidence

4.1 Estimates From Baseline Modes

Estimates of the equation system from (2) to (4) are reported in Table 3. Equation (2) can be viewed as a traditional energy stock return and oil price equation with incremental variables of ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ and ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ while controlling for the σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ and σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ . Several comments are worth noting. First, the coefficients of ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ and ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ are negative and statistically significant at the 5% level or better, except in the cases in Germany and India, which have opposite signs for either the ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ or ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ . The evidence of energy stocks for countries with a negative sign reflects the fact that the NPV < 0, which may be attributed to a spillover of climate uncertainty to business production or household consumption as ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ causes uncertainty and financial instability (Raza et al. 2024; Kayani et al. 2024) that negatively affect economic activity (Barnett et al. 2022). Under this perspective, a rise in climate uncertainty tends to induce governments to impose taxes on carbon emissions, which raise production costs and, in turn, reduce profits, especially for companies that heavily rely on fossil fuel energy to produce goods. This evidence is consistent with the findings of Lasisi et al. (2024) and Bouri et al. (2022).

Table 3. Estimates of energy stock returns to lagged climate policy uncertainty, change in oil price, controlling for unusual observations using variance variables in 2008 and 2020.
Market C ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ ∆ p oil , t − 1 ${\unicode{x02206}p}_{{oil},t-1}$ ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ω i , 0 ${\omega }_{i,0}$ ε i , t − 1 2 ${\varepsilon }_{i,t-1}^{2}$ σ i , t − 1 2 ${\sigma }_{i,t-1}^{2}$ R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$
US 0.834*** −0.002*** −0.010*** −0.121*** −0.015** −4.301*** −0.630*** 29.792 0.223 0.794* 0.10
16.48 −4.51 −9.69 −25.70 −2.48 −21.53 −5.73 0.40 0.40 1.67
CA 0.783*** −0.012*** −0.022*** −0.019*** −0.024*** −4.188*** −0.437*** 23.972 0.188 0.845** 0.12
8.93 −9.13 −19.31 −9.09 −6.19 −8.69 −4.54 0.36 0.37 2.20
FR 0.760*** −0.002*** −0.010*** −0.051*** −0.015*** −1.209*** −0.196* 26.525 0.512 0.756** 0.01
12.70 −15.17 −4.88 −6.37 −5.00 −4.51 −1.77 0.49 0.64 1.98
GM −0.039 0.018*** −0.012** 0.032 −0.103*** −8.132*** −0.378*** 69.328 0.036 0.655 0.09
−0.90 2.99 −2.05 1.12 −4.32 −6.81 −9.61 0.22 0.11 0.42
IT 0.846*** 0.005 −0.011** −0.089*** −0.029*** −3.330*** −0.365*** 31.275 0.117 0.488 0.05
5.52 1.42 −2.24 −7.12 −4.52 −7.29 −7.22 0.32 0.34 0.32
UK 0.450*** −0.006** −0.013*** −0.053*** −0.001** −0.044 −0.373*** 25.627 0.156 0.884*** 0.03
5.38 −2.17 −3.98 −5.84 −1.92 -0.48 −11.32 0.38 0.21 3.27
JP 0.474*** −0.003** −0.025*** −0.021*** −0.040*** 5.511*** −0.411** 37.783 0.320 0.792* 0.07
8.10 −2.06 −10.52 −5.67 −3.52 −5.2 −2.59 0.36 0.42 1.69
CN 0.392*** −0.008*** −0.006*** −0.031*** 0.039*** −6.428*** −0.082 25.946 0.700 0.703*** 0.01
3.39 −3.30 −4.61 −3.80 3.05 −50.87 −1.41 0.80 1.08 3.20
KO 0.967*** −0.016*** −0.001 −0.135*** −0.062*** 1.015 −0.402** 10.031 0.349 0.847*** 0.02
17.69 −24.01 −1.63 −21.12 −9.85 0.87 −2.09 0.48 1.05 5.72
KW 0.097 −0.008*** −0.003 −0.009 0.072*** −0.100 −0.182** 37.551 0.292 0.828* 0.03
0.91 −2.91 −1.25 −0.64 7.27 −0.38 −2.06 0.36 0.34 1.94
ID 1.060*** −0.051*** −0.022*** 0.017*** 0.235*** −12.06*** −0.100*** 141.81 1.771 0.761*** 0.03
12.54 −23.75 −9.54 34.99 42.42 −14.56 −6.50 0.55 0.79 3.03
IN 0.998*** −0.018*** 0.005*** −0.036** 0.037*** −3.720*** −0.374* 45.971 0.244 0.814 0.01
11.91 −10.53 4.54 −2.44 3.28 −3.17 −1.82 0.31 0.38 1.52
SG 0.521*** −0.009*** 0.002 0.067*** −0.050*** −6.946*** −0.284*** 137.89 1.684 0.351 0.07
10.25 −3.38 1.34 7.68 −10.99 −4.09 −7.37 1.17 1.01 1.02
BR 1.550*** −0.024*** −0.019*** −0.020*** −0.055*** 3.833*** −0.595*** 282.75 0.927 0.187 0.02
10.77 −11.36 −12.74 −8.96 −4.22 3.13 −3.71 0.88 0.67 0.28
MX 0.818*** −0.005*** −0.035*** −0.061*** −0.020*** −5.275*** −0.477*** 34.112 0.300 0.697 0.10
21.30 −5.88 −38.18 −20.28 −5.02 −10.02 −5.24 0.35 0.44 0.93
  • Note: The dependent variable is the aggregate stock at time t. ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ is log-difference of climate policy uncertainty. ∆ p oil , t $\unicode{x02206}{p}_{{oil},t}$ denotes change in Brent oil price. σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ = EMV GFC , t ${{EMV}}_{{GFC},t}$ * I 2008 , EMV FC , t ${{EMV}}_{{FC},t}$ is equity market volatility calibrated to financial crisis, I 2008 is an indicator variable that sets to unity in 2008-global financial crisis and zero otherwise; σ ID , t 2 = ${\sigma }_{{ID},t}^{2}=$ EMV ID , t ${{EMV}}_{{ID},t}$ * I 2020 , where EMV ID , t ${{EMV}}_{{ID},t}$ is the equity market volatility calibrated to infected disease, where I 2020 is an indicator variable that sets to unity in 2020 COVID-19 period and zero otherwise. The numbers in the first row are the estimated coefficients, the second row contains the z-statistics. The critical values of z-distribution at the 1%, 5%, and 10% levels of significance are 2.63, 1.98, and 1.66, respectively. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. R ¯ 2 ${\bar{R}}^{2}$ is the adjusted R-squared.

However, the positive signs of ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ for GM (IT is insignificant) and ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ for IN (SG is insignificant) are consistent with the projection of NPV > 0. The rationale behind this result is complex. It can be argued that climate change policy can improve the environment or economic activity; alternatively, it could be attributable to a spurious correlation due to missing variables. It should be noted that the literature also shows that investors can exploit the information content of CPU to gain high returns, leading to a positive coefficient (He and Zhang 2022; Salisu et al. 2023). As indicated earlier, the levels of CPU t − 1 ${{CPU}}_{t-1}$ and CPU t − 2 ${{CPU}}_{{\rm{t}}-2}$ are positively correlated with the other uncertainty variables, producing biased estimators. As a result, a form of change expressed by ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ and ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ are employed in the test equation.

Second, the coefficients of ∆ p oil , t − 1 ${\unicode{x02206}p}_{\text{oil},{\rm{t}}-1}$ and ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ are negative and highly significant; except for ID, which exhibits a positive sign, and CN, SG, and GM, which present mixed signs, however, in the case of GM, the sign is insignificant. The negative signs are consistent with different interpretations. As described by Kilian and Park (2009), higher oil prices cause lower returns stemming from a change in precautionary demand for crude oil, which reflects fears about the availability of future oil supplies. In contrast, positive shocks to the global aggregate demand for oil commodities due to economic growth (Bernanke 2016) can drive both oil prices and higher stock prices higher, showing a positive correlation. Further, a rise in oil prices causes higher costs of production in manufacturing commodities, which can be passed onto consumers in the form of higher prices that in turn can result in a decline in demand as consumers tighten their belts and consequentially a decline in demand for businesses' products and services and lower profits and stock returns. The evidence is consistent with the studies by Hamilton (1996), Hwang and Kim (2021), Nasir et al. (2018), and Abiad and Qureshi (2023), who argue that a rise in oil prices has a negative effect on stocks due to oil price shocks, which then cause output production to decline and the onset of an economic recession and eventual drop in stock returns. Moreover, the negative relationship between a change in oil prices and stock returns contrasts to Prabheesh et al. (2020), who found a positive relationship between oil price returns and stock price returns during the COVID-19 period in CN, IN, JP, and KO. Their results were found in a unusual economic conditions that did not control for the impact from σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ and σ ID , t 2 . ${\sigma }_{{ID},t}^{2}.$ However, our evidence of a change in oil prices in UK, ID, and KW indicates positive and significant effects on stock returns that may be attributable to this study's inclusion of oil-abundant countries that can benefit from oil-exporting to boost profits and stock returns (Atif et al. 2022). The remaining countries of CN, IN, and SG present mixed signs, which may be the result of market reversal, government manipulation, or spurious correlations due to missing variables.

Third, the coefficients of σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ and σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ have negative signs and are highly significant. Evidence suggests that energy stock returns were adversely affected by dramatic economic conditions stemming from the 2008 GFC and 2020–21 COVID-19 pandemic. The results are consistent with the findings in the literature (Terry et al. 2020; Batten et al. 2021; Cheema et al. 2022; Chiang and Tang 2023). Moreover, the inclusion of these extreme volatilities is necessary since these influential observations can produce biased estimators (Wei 2018; Peña 2001) if without including them.

4.2 Estimates With Climate Induced Volatility

It is recognized that climate policy alleviates some of the harms of climate change by restraining the consumption of fossil fuels, which alters the supply and demand of oil and gas and leads to high price volatility of fossil fuels and, in turn, stock volatility (Salisu et al. 2024). This effect stems from a rise for ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ , which can cause a rise in costs as changes in production structure enter new, unknown terrain or expectations of heighten emission tax rise (Diaz-Rainey et al. 2021), both of which jeopardize business profits. Moreover, these changes can be reinforced by unexpected climate-influenced moods, which can have an impact on decision-making and hence stock returns (Schulte-Huermann 2020). The argument for the above-mentioned indirect effect involves: ∆ cpu t → σ t ${\unicode{x02206}{cpu}}_{t}\to {\sigma }_{t}$ → $\to $ R t ${R}_{t}$ → Global contagion in R t $\to {Global\; contagion}\,\text{in}\,{R}_{t}$ . In other words, the covariance between ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and overall EMV t denoted by cov σ , ∆ cpu , t $\text{denoted by}\,{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t}$ with two-period lags can be incorporated into the test equation. The estimates, which are reported in Table 4, show the coefficients of cov σ , ∆ cpu , t − 1 and ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}\text{and}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ are mainly negative and statistically significant, although a few countries exhibit mixed signs depending on the time period. The negative sign is consistent with our expectation that a rise in climate policy uncertainty has a significant impact on energy volatility, and hence the equity market volatility, which can produce fears and lower stock returns. This effect is implicitly embodied in the study of Isah et al. (2023), which identifies a unique channel and suggests an indirect impact to explain stock returns in addition to the direct effect from the lagged ∆ cpu t . ${\unicode{x02206}{cpu}}_{t}.$ With respect to the performance of other regressors, including { ∆ cpu t − 1 , ∆ cpu t − 2 , ∆ p oil , t − 1 $\unicode{x02206}{{cpu}}_{t-1},\unicode{x02206}{{cpu}}_{t-2},\unicode{x02206}{p}_{\text{oil},t-1}$ , ∆ p oil , t − 2 $\unicode{x02206}{p}_{\text{oil},t-2}$ , · σ GFC , t 2 $\cdot \,{\sigma }_{{GFC},t}^{2}$ and σ ID , t 2 } ${\sigma }_{{ID},t}^{2}\}$ , the estimated coefficients in this table reveal very comparable and qualitative results.

Table 4. Estimates of energy returns response to dynamic changes in climate policy uncertainty ( Δ ${\rm{\Delta }}$ cput ), covariance between Δ ${\rm{\Delta }}$ cput and stock return volatility, and appreciation of oil price controlling for unusual variance observations in 2008 and 2020 crises.
Market C ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p oil , t − 1 ${\unicode{x02206}p}_{{oil},t-1}$ ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ω i , 0 ${\omega }_{i,0}$ ε i , t − 1 2 ${\varepsilon }_{i,t-1}^{2}$ σ i , t − 1 2 ${\sigma }_{i,t-1}^{2}$ R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$
US 0.850*** −0.002** −0.011*** −0.003*** −0.007*** −0.104*** −0.009* −4.327*** −0.844*** 28.261 0.197 0.872*** 0.06
14.66 −2.31 −8.33 −5.40 −4.61 −11.03 −1.69 −8.05 −4.15 0.41 0.36 2.95
CA 0.781*** −0.010** −0.015*** 0.010** −0.006* −0.031*** −0.024*** −4.777*** −0.445*** 61.086 0.748 0.183 0.12
6.21 −2.35 −3.26 2.56 −1.96 −3.01 −3.27 −6.35 −5.14 0.81 0.86 0.23
FR 0.596*** -0.008** -0.007*** 0.007*** -0.018*** -0.047*** -0.007*** -2.348*** −0.199* 28.161 0.483 0.760* 0.03
5.71 −2.48 −5.23 4.21 −6.23 −5.26 −3.10 −3.78 −1.90 0.45 0.61 1.83
GM 0.068 0.020*** −0.020*** −0.016* −0.011* 0.034 −0.074*** −8.376*** −0.380* 69.907 0.004 0.729 0.08
1.43 2.72 −3.09 −1.95 −1.72 1.29 −2.82 −6.57 −1.84 0.23 0.01 0.62
IT 0.727*** 0.002 −0.017** 0.007 −0.016*** −0.085*** 0.027 −3.888*** −0.329*** 19.837 0.043 0.578 0.04
3.65 0.32 −2.45 1.18 −3.41 −4.47 1.45 −5.25 −6.07 0.19 0.22 0.27
UK 0.414*** −0.008*** −0.014*** −0.005*** −0.004*** −0.052*** 0.009*** −0.042 −0.375*** 26.050 0.089 0.893*** 0.03
3.82 −3.17 −4.05 −11.28 −3.62 −12.50 11.98 −1.16 −51.92 0.41 0.13 3.60
JP 0.507*** −0.006*** −0.026*** −0.009*** −0.011** −0.013*** −0.052*** −6.969*** −0.418*** 37.297 0.224 0.823* 0.06
5.46 −20.62 −9.03 −27.77 −10.39 −7.37 −13.31 −12.01 −2.67 0.37 0.37 1.98
CN 0.340*** −0.013*** −0.011** −0.001*** −0.016*** −0.012*** 0.054*** −6.314*** −0.052 31.888 0.876 0.719*** 0.02
3.78 −2.78 −2.66 −5.85 −5.87 −2.72 4.53 −13.51 −1.14 0.68 0.94 2.94
KO 0.919*** −0.017*** −0.012*** −0.023*** −0.033*** −0.130*** −0.021** −2.747* −0.410* 33.931 0.673 0.828*** 0.03
9.23 −4.20 −2.84 −4.90 −6.59 −12.64 −2.12 −1.89 −1.79 0.43 0.74 3.49
KW −0.271* −0.007*** −0.007** −0.006** −0.004*** 0.001* 0.051*** −0.205 −0.182** 37.301 0.189 0.728 0.03
−1.82 −3.15 −2.43 −2.35 −3.34 1.67 3.49 −0.33 −2.02 0.33 0.34 0.97
ID 0.962*** −0.052*** −0.026*** −0.020*** 0.005*** −0.021*** 0.209*** −12.49*** −0.106**** 99.735 1.381 0.805*** 0.02
15.78 −16.19 −17.08 −6.63 2.73 −4.23 17.41 −13.48 −6.68 0.55 0.85 4.15
IN 1.036*** −0.025*** −0.001** −0.014*** 0.007*** −0.019** 0.039*** −3.615*** −0.368* 53.069 0.133 0.737 0.01
15.67 −11.08 −2.17 −3.18 5.61 −2.00 5.99 −3.47 −1.81 0.26 0.30 0.79
SG 0.519*** −0.015*** −0.013*** −0.036*** −0.018*** 0.090*** −0.020*** −5.407*** −0.329*** 42.420 0.615 0.607** 0.07
5.82 −5.69 −7.63 −7.71 −5.14 6.14 −4.81 −3.37 -4.31 1.02 1.10 2.27
BR 1.244*** −0.024*** −0.031** −0.021*** −0.002* −0.013*** −0.017** −1.189 −0.599*** 76.718 0.471 0.735 0.04
5.90 −4.58 −5.56 −3.98 −1.89 −3.78 −2.46 −1.18 −3.88 0.48 0.53 1.65
MX 0.068 0.020*** −0.020*** −0.016* −0.011* 0.034 −0.074*** −8.376*** −0.380* 69.907 0.004 0.729 0.08
1.43 2.72 −3.09 −1.95 −1.72 1.29 −2.82 −6.57 −1.84 0.23 0.01 0.62
  • Note: The equation is R t = C + β 1 ∆ cpu t − 1 + β 2 ∆ cpu t − 2 + β 3 cov σ , ∆ cpu , t − 1 + β 4 cov σ , ∆ cpu , t − 2 ${R}_{t}=C+{\beta }_{1}{\unicode{x02206}{cpu}}_{t-1}+{\beta }_{2}{\unicode{x02206}{cpu}}_{t-2}+{\beta }_{3}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}+{\beta }_{4}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ + β 5 ∆ p oil , t − 1 + β 6 ∆ p oil , t − 2 + β 7 σ GFC , t 2 + β 8 σ GFC , t 2 + $+\,{{\beta }_{5}{\unicode{x02206}p}_{{oil},t-1}+{\beta }_{6}{\unicode{x02206}p}_{{oil},t-2}+\beta }_{7}{\sigma }_{{GFC},t}^{2}+{\beta }_{8}{\sigma }_{{GFC},t}^{2}\,+$ ε t . ${\varepsilon }_{t}.$ The numbers in the first row are the estimated coefficients, the second row contains the z-statistics. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$ is the adjusted R-squared.

4.3 Estimates Based on Uncertainty in Energy and Environmental Policy

The implementation of climate policy tends to provoke an effect that governments implement energy & environmental policy because energy is a primary factor affecting business production and household consumption. Energy policy uncertainty most likely will have a dramatic impact across the economy, causing market anxiety and volatility in equity markets. This covariance between change in energy and environmental policy and equity market volatility can be captured by the measure of ∆ emv EN , t ${\unicode{x02206}{emv}}_{{EN},t}$ (Terry et al. 2020, Economic Policy Uncertainty Index). Following this line of reasoning, this study incorporates the lagged energy policy & environment policy changes that affect market volatility into the equation. The estimates of the model, which include the addition of ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ and ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ , are reported in Table 5.

Table 5. Estimates of energy returns response to dynamic changes in CPU and its induced equity volatility, oil price change and change in EMV calibrated to energy policy change controlling for unusual observation volatility in 2008 and 2020 crises.
Market C ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p oil , t − 1 ${\unicode{x02206}p}_{{oil},t-1}$ ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ω i , 0 ${\omega }_{i,0}$ R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$
US 1.013*** −0.002*** −0.013*** −0.002*** −0.007*** −0.097*** −0.010* −0.004*** −0.001*** −4.402*** −0.842*** 29.936 0.06
13.90 −11.38 −7.55 −4.87 −4.89 −10.63 −1.68 −4.25 −3.28 −8.08 −4.18 0.43
CA 0.977*** −0.013*** −0.019*** 0.016*** −0.007*** −0.008*** −0.015*** −0.003*** −0.000*** −4.651*** −0.521*** 25.584 0.11
16.63 −8.18 −11.09 6.35 −4.21 −2.92 −3.14 −27.84 −4.62 −6.02 −6.15 0.50
FR 0.548*** −0.009*** −0.006** 0.007*** −0.017*** −0.049*** −0.001*** −0.001*** 0.001*** −2.182*** −0.197** 22.466 0.03
7.30 −4.40 −2.31 4.73 −6.88 −10.78 −4.11 −3.09 3.43 −3.72 −1.93 0.34
GM 0.102*** 0.021*** −0.022*** −0.014** −0.011*** 0.038*** −0.077*** −0.005*** 0.004*** −8.436*** −0.382 ** 67.480 0.08
11.17 4.38 −9.71 −2.33 −16.16 6.83 −5.78 −10.16 23.69 −7.11 −1.94 0.31
IT 0.493*** 0.003** −0.008*** 0.003 −0.016*** −0.078*** 0.009** −0.002** −0.004*** −3.460*** −0.309*** 26.103 0.04
5.64 2.00 −3.90 1.10 −9.28 −9.25 2.62 −2.49 −6.85 −6.52 −16.60 0.29
UK 0.445*** −0.007*** −0.013*** −0.007** −0.002*** −0.053*** 0.001*** −0.001** −0.109** −0.001 −0.372*** 26.515 0.02
5.20 −3.47 −3.67 −2.12 −2.87 −8.68 4.64 −2.37 −2.37 −0.12 −5.00 0.35
JP 0.329*** −0.006** −0.028*** −0.007** −0.011*** −0.019** −0.035*** −0.001*** 0.002*** −7.303*** −0.414** 33.088 0.05
2.73 −2.27 −11.10 −2.17 −4.40 −2.01 −3.75 −2.70 7.15 −9.69 −2.62 0.35
CN 0.784*** −0.009*** −0.015*** −0.009**** −0.013*** 0.018** 0.048*** −0.004*** −0.000 −6.503*** −0.080 28.937 0.02
9.07 −8.60 −5.93 −4.60 −4.46 1.94 8.96 −21.94 −1.60 −18.11 −0.89 0.66
KO 0.650*** −0.007 −0.005*** 0.021*** −0.012*** −0.040*** 0.033*** −0.001 0.001*** −2.420*** −0.186* 20.563 0.01
8.74 −3.90 −8.20 8.06 −5.15 −4.38 3.87 −1.49 3.54 −4.87 −1.85 0.81
KW −0.149 −0.006*** −0.005*** −0.007*** −0.003* −0.006*** 0.084*** −0.003*** −0.008*** 0.071 −0.165* 43.552 0.01
−1.40 −3.68 −2.79 −3.25 −1.67 −2.91 12.09 −11.34 −11.21 1.26 −1.81 0.30
ID 1.596*** −0.050*** −0.027*** −0.012*** 0.003 0.013*** 0.235*** −0.006*** −0.009*** −12.41*** −0.094 93.091 0.02
12.87 −9.75 −11.38 −11.01 4.76 7.76 14.00 −5.61 −5.22 −13.50 −1.29 0.54
IN 0.973*** −0.020*** −0.002*** −0.013*** 0.007 −0.019** 0.042*** −0.001 0.002* −3.712*** −0.368* 48.670 0.00
8.49 −7.50 −3.23 −5.35 2.36 −2.25 4.97 −1.50 1.69 −3.29 −1.77 0.28
SG 0.686*** −0.013*** −0.013*** −0.031*** −0.011 0.088*** −0.020** −0.006*** 0.001** −7.277*** −0.326*** 55.108 0.08
10.90 −4.62 −4.32 −10.46 −17.64 9.95 −2.22 −11.22 2.36 −3.38 −4.26 0.74
BR 1.581*** −0.026*** −0.031*** −0.028*** −0.003 −0.033*** −0.001 −0.006*** 0.001 3.782*** −0.613*** 93.630 0.01
7.55 −7.08 −7.08 −4.01 −2.04 −2.92 −0.13 −4.01 0.11 5.03 −3.86 0.42
MX 0.885*** −0.007*** −0.018*** −0.004*** −0.005 −0.05*** −0.021** −0.001*** 0.001 −6.328*** −0.637*** 37.003 0.08
9.33 −2.75 −7.89 −3.39 −3.50 −7.44 −2.24 −2.90 0.21 −14.05 −3.79 0.49
  • Note: The equation is R t = C + β 1 ∆ cpu t − 1 + β 2 ∆ cpu t − 2 + β 3 cov σ . ∆ cpu , t − 1 + β 4 cov σ , ∆ cpu , t − 2 ${R}_{t}=C+{\beta }_{1}{\unicode{x02206}{cpu}}_{t-1}+{\beta }_{2}{\unicode{x02206}{cpu}}_{t-2}+{\beta }_{3}{\text{cov}}_{\sigma .\unicode{x02206}{cpu},t-1}+{\beta }_{4}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ + β 5 ∆ p oil , t − 1 + β 6 ∆ p oil , t − 2 + ∆ β 7 emv EN , t − 1 + β 8 ∆ emv EN , t − 2 + β 9 σ GFC , t 2 + β 10 σ GFC , t 2 + $+\,{{\beta }_{5}{\unicode{x02206}p}_{{oil},t-1}+{\beta }_{6}{\unicode{x02206}p}_{{oil},t-2}+{\unicode{x02206}{\beta }_{7}{emv}}_{{EN},t-1}+{\beta }_{8}{\unicode{x02206}{emv}}_{{EN},t-2}+\beta }_{9}{\sigma }_{{GFC},t}^{2}+{\beta }_{10}{\sigma }_{{GFC},t}^{2}+$ ε t . ${\varepsilon }_{t}.$ The numbers in the first row are the estimated coefficients, the second row contains the z-statistics. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$ is the adjusted R-squared.

Focusing on the estimated coefficients of ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ and ∆ emv EN , t − 2 , ${\unicode{x02206}{emv}}_{{EN},t-2},$ we see that the signs are negative and statistically significant for ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ ; however, some of the estimates for ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ are weaker or exhibit reversed signs. The negative coefficients reveal the dynamic nature of energy & environmental uncertainty, which triggers anxiety in investors and motivates them to engage in panic selloff that causes stocks to plunge. In a highly integrated global market, fears in US markets are most likely to be transmitted to the global market via a contagion effect (Chiang et al. 2007; Georgiadis 2016). This evidence is consistent with the results of Joo and Park (2021) and Salisu et al. (2024). The results also indicate that parametric relationships are rather stable, and that the performance of the other variables exhibit comparable results. In summary, the evidence indicates that the model is consistent with a broad uncertainty specification (Barnett et al. 2022).

5 Robustness Tests

5.1 Test on the Aggregate Market

To examine the validity of the model arguments on stock returns, the dependent variable energy stock return, is now replaced by aggregate market return, which is used in estimating each country's equation. Estimates are reported in Table 6. Evidence shows that the statistical results are very comparable except for some minor changes. Let us briefly go through each variable in the test equation. First, it can be seen that all of coefficients of ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ and ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ are significantly negative except the coefficients in IT and SG in lag one and KU in lag 2 to exhibit a significantly positive sign. Thus, the estimated results with negative coefficients are consistent with the previous table and agree with those the literature (Barnett et al. 2022; Raza et al. 2024; Kayani et al. 2024).

Table 6. Robust test of investigating market returns in response to dynamic changes in CPU and its induced equity volatility, oil price change and change in EMV calibrated to energy policy controlling for unusual variance volatility in 2008 and 2020 crises.
Market C ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p oil , t − 1 ${\unicode{x02206}p}_{{oil},t-1}$ ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ω i , 0 ${\omega }_{i,0}$ R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$
US 1.961*** −0.015*** −0.024 −0.035*** −0.017*** −0.062*** 0.003*** −0.003*** 0.001*** −6.617*** −0.542* 44.934 0.09
28.18 −10.18 −6.41 −8.05 −4.91 −8.70 16.52 −13.13 12.18 −14.81 −1.71 0.31
CA 1.085*** −0.006*** −0.007 −0.005*** −0.011*** −0.028*** 0.007*** −0.003*** −0.001*** −4.531*** −0.470*** 8.993 0.07
29.99 −6.13 −4.22 −3.58 −10.66 −6.36 3.95 −6.67 −10.35 −20.11 −8.64 0.55
FR 1.043*** −0.005*** −0.005 −0.011*** −0.006*** −0.072*** 0.014*** −0.002*** 0.003*** −3.617*** −0.301*** 11.810 0.07
22.49 −5.11 −4.23 −20.39 −5.28 −15.17 11.86 −9.27 12.70 −11.26 −6.45 0.30
GM 0.784*** −0.001*** −0.002 −0.016*** −0.011*** −0.064*** 0.013*** −0.002*** 0.001*** −3.135*** −0.319*** 22.374 0.05
18.580 −6.875 −1.821 −10.935 −14.145 −22.48 6.338 −3.554 3.915 −11.735 −20.385 0.370
IT 0.852*** 0.005** 0.001 −0.009*** −0.011*** −0.102*** 0.008* −0.000 0.001 −4.072*** −0.447*** 33.390 0.09
11.56 2.20 0.33 −6.01 −6.98 −11.46 1.82 −0.10 0.31 −11.00 −5.24 0.77
UK 0.963*** −0.004** −0.010*** −0.014*** −0.005*** −0.036*** −0.007** −0.001** 0.002*** −3.098*** −0.311*** 10.391 0.10
17.27 -2.04 −6.54 −8.72 −3.97 −10.16 −2.40 −2.55 5.73 −37.06 −19.50 0.41
JP 0.472*** −0.011*** −0.014*** −0.006*** −0.003** −0.020** −0.002*** 0.004*** −0.001*** −3.708*** −0.143 14.422 0.07
6.24 −5.13 −8.37 −7.53 −2.08 −2.43 −3.08 7.96 −3.02 −7.99 −1.44 0.33
CN 0.865*** −0.012** −0.012*** −0.014*** −0.011*** 0.018*** 0.026*** −0.003*** −0.001*** −6.995*** −0.113 30.095 0.01
4.45 −2.41 −2.57 −2.84 −3.71 2.91 3.44 −3.20 −5.90 −16.07 −0.92 0.64
KO 0.772*** −0.004** −0.028*** −0.021*** −0.027*** −0.049*** −0.037*** −0.003*** 0.001*** −4.838*** −0.254*** 0.952 0.03
8.02 −2.43 −8.76 −9.61 −9.65 −3.70 −3.47 −3.94 4.87 −6.40 −5.65 0.46
KW 0.648*** 0.002 0.009*** 0.003* −0.001** −0.016*** 0.090*** 0.002*** 0.002*** −2.339*** −0.370*** 11.360 0.13
7.41 0.75 2.77 1.91 −2.23 −2.73 8.71 5.64 2.96 −4.45 −18.14 0.55
ID 1.263*** −0.006*** −0.024*** −0.013*** −0.024*** −0.044*** 0.012*** −0.006*** 0.001*** −4.879*** −0.174*** 24.760 0.06
23.88 −2.94 −9.68 −7.18 −16.30 −6.10 2.83 −18.62 2.88 −13.51 −6.56 0.40
IN 1.367*** −0.020*** −0.006*** −0.008*** 0.002*** −0.079*** 0.023*** −0.002*** 0.005*** −4.396*** −0.464*** 30.043 0.06
16.19 −24.40 −2.83 −3.62 15.02 −7.29 2.67 −3.52 6.10 −7.95 −5.00 0.40
SG 0.091*** 0.001*** −0.001** −0.001*** −0.001*** 0.002*** 0.004*** −0.001*** −0.001*** −0.085 −0.006 0.010 0.02
9.78 3.09 −2.25 −3.41 −3.87 2.87 5.46 −7.15 −2.92 −1.07 −0.80 0.54
BR 1.439*** −0.009*** −0.015*** −0.014*** −0.016*** 0.031*** −0.014* −0.006*** −0.003*** −3.156*** −0.109*** 31.523 0.06
19.40 −3.12 −3.78 −3.66 −9.77 4.28 −1.89 −5.06 −3.75 −4.68 −3.25 0.37
MX 1.307*** −0.002*** −0.012*** −0.015*** −0.003*** 0.043*** −0.014*** −0.002*** −0.004*** −1.968*** −0.171*** 2.472 0.09
20.19 −3.05 −7.09 −10.83 −4.56 17.59 −5.77 −9.02 −13.61 −4.00 −8.29 0.34
  • Note: The equation is: R t = C + β 1 ∆ cpu t − 1 + β 2 ∆ cpu t − 2 + β 3 cov σ . ∆ cpu , t − 1 + β 4 cov σ , ∆ cpu , t − 2 ${R}_{t}=C+{\beta }_{1}{\unicode{x02206}{cpu}}_{t-1}+{\beta }_{2}{\unicode{x02206}{cpu}}_{t-2}+{\beta }_{3}{\text{cov}}_{\sigma .\unicode{x02206}{cpu},t-1}+{\beta }_{4}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ + β 5 ∆ p oil , t − 1 + β 6 ∆ p oil , t − 2 + β 7 ∆ emv EN , t − 1 + β 8 ∆ emv EN , t − 2 + β 9 σ GFC , t 2 + β 10 σ GFC , t 2 + ε t . $+\,{{\beta }_{5}{\unicode{x02206}p}_{{oil},t-1}+{\beta }_{6}{\unicode{x02206}p}_{{oil},t-2}+{{\beta }_{7}\unicode{x02206}{emv}}_{{EN},t-1}+{\beta }_{8}{\unicode{x02206}{emv}}_{{EN},t-2}+\beta }_{9}{\sigma }_{{GFC},t}^{2}+{\beta }_{10}{\sigma }_{{GFC},t}^{2}+{\varepsilon }_{t}.$ The numbers in the first row are the estimated coefficients, the second row contains the z-statistics. ***,**, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$ is the adjusted R-squared.

Second, evidence from the induced risks, cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ and cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ , indicates a significantly negative result, except for one instance in IN. This outcome is consistent with the market behavior in which rational investors observe a rise in cov σ , ∆ cpu , t ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t}$ driven by government's announcement of a higher ∆ cpu t − 1 $\unicode{x02206}{{cpu}}_{t-1}$ and engage selloff stocks accompanied by market volatility, resulting in a stock market plumet. The evidence derived from our test validates the significance of the inclusion of the argument.

Third, the responses of stock returns to ∆ p oil , t − 1 ${\unicode{x02206}p}_{\text{oil},{\rm{t}}-1}$ and ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ exhibit very similar behavior, showing negative signs and statistically significant; except for CN and SG and partly in BR and MX, which exhibits reveal a positive sign. The majority of countries that exhibit significantly negative coefficients is consistent with the notion of the recession hypothesis as noted by Hamilton (1983, 1996) and the subsequent studies by Hwang and Kim (2021), Nasir et al. (2018) and Abiad and Qureshi (2023), who argue that a rise in oil prices lead to higher production costs and bring about an adverse effect on economic activities that jeopardizes stock returns.

Fourth, the estimated results of the coefficients of ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ are negative and statistically significant except JP and KU, indicating that uncertainty in the energy & environmental policy tends to create the sentiment of fears to economic agents, frustrating stock market performance. However, in the longer lag, the coefficients of ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ exhibit mixed signs with market reverse in some countries. But this is consistent with the case of Table 5, where the energy returns are served as dependent variables. The negative coefficients reveal the dynamic nature of energy and environmental uncertainty, which triggers anxiety in investors and motivates them to engage in panic selloff that causes stocks to plunge. The evidence is consistent with the findings of Joo and Park (2021) and Salisu et al. (2024).

Fifth, the estimated coefficients for the variance series, σ ̃ GFC , t 2 ${\widetilde{\sigma }}_{{GFC},t}^{2}$ and σ ̃ ID , t 2 ${\widetilde{\sigma }}_{{ID},t}^{2}$ , all exhibit a negative sign and are statically significant with minor exceptions which are insignificant. The negative signs simply reveal the damaging impact due to the shocks from global financial crises or damaging effect from the pandemic that cause global selloff. The findings are consistent with the reports by Terry et al. (2020) and Batten et al. (2021). In sum, the evidence concludes that the model is robust whether the test is conducted by using energy stock return or aggregate stock returns.

5.2 Test the Effect of Change in Gasoline Prices

This section tests the model's performance using an alternative measure of the price variable, i.e., replacing the change in oil prices, ∆ p Oil , t $\unicode{x02206}{p}_{{Oil},t}$ , with the change in gasoline prices, ∆ p Gas , t . $\unicode{x02206}{p}_{{Gas},t}.$ The estimates are reported in Table 7. Apparently, the coefficients of ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ , ∆ cpu t − 2 , ${\unicode{x02206}{cpu}}_{t-2},$ cov σ , ∆ cpu , t − 1 , ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1},$ and cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ continually have the expected signs and statistically significant, the estimated results are comparable to those presented in Tables 5 and 6. While checking with the new variable, the coefficients of ∆ p Gas , t − 1 $\unicode{x02206}{p}_{{Gas},t-1}$ in most G7 countries (except JP) and KO, ID and IN mainly present negative signs and statistically significant; the remining countries exhibit positive signs. Yet, some countries with positive signs, including JP, MX SG, BR, and MX in time t–1 then turn to negative at later points of time, as shown in the coefficients of ∆ p Gas , t − 2 . $\unicode{x02206}{p}_{{Gas},t-2}.$ The diverse signs for the coefficients of ∆ p Gas , t − 1 $\unicode{x02206}{p}_{{Gas},t-1}$ and ∆ p Gas , t − 2 $\unicode{x02206}{p}_{{Gas},t-2}$ may be subject to the local government interfering or affected by local country's market conditions. The negative sign can be interpreted based on the rationale of Hamilton's argument (2011) in that a rise in gasoline prices can raise production costs in business operation as well as in the households' consumption spending, both forces can reduce the profits of business firms and eventually reflects in profits and lower stock prices. However, the demand for gasoline consumption is very inelastic, and a higher price in gasoline can be easily shifted forward to consumer's prices, which raise company's profits and lifts in stock price. Thus, we can observe a positive relation between stock returns and gasoline prices. Finally, with respect to the coefficients of ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ , ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ , σ GFC , t 2 , ${\sigma }_{{GFC},t}^{2},$ and σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ , the signs and significance level are stable and comparable with previous findings. We can conclude that the model is robust whether oil prices or gasoline prices are used as an argument in the test equation.

Table 7. Robust test of examining aggregate stock returns in response to dynamic changes in CPU and its induced equity volatility, gasoline price change and change in EMV calibrated to energy policy change controlling for unusual variance volatility in 2008 and 2020 crises.
Market C ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p Gas , t − 1 ${\unicode{x02206}p}_{{Gas},t-1}$ ∆ p Gas , t − 2 ${\unicode{x02206}p}_{{Gas},t-2}$ ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ω i , 0 ${\omega }_{i,0}$ R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$
US 1.961*** −0.015*** −0.024*** −0.035*** −0.017*** −0.062*** 0.003*** −0.003*** 0.001*** −6.617*** −0.542** 44.934 0.09
28.18 −10.18 −6.41 −8.05 −4.91 −8.70 16.52 −13.13 12.18 −14.81 −1.71 0.31
CA 1.075 −0.005*** −0.008*** −0.006*** −0.011*** −0.010*** −0.011*** −0.003*** −0.001*** −3.502*** −0.436*** 10.396 0.08
38.96*** −5.55 −9.12 −7.51 −14.82 −3.50 −4.58 −13.16 −3.90 −12.75 −34.38 0.40
FR 0.970*** −0.005*** −0.005*** −0.011*** −0.005** −0.041*** −0.008** −0.004*** −0.001*** −3.357*** −0.281*** 14.338 0.07
29.26 −3.87 −5.64 −7.36 −2.58 −9.66 −2.40 −11.54 −20.74 −38.26 −10.40 0.36
GM 0.836*** −0.003*** −0.002** −0.013*** −0.010*** −0.041*** −0.012* −0.002*** 0.002** −3.199*** −0.312*** 18.134 0.04
9.92 −2.82 −2.38 −4.67 −5.12 −6.86 −1.77 −6.75 2.26 −11.17 −4.55 0.48
IT 0.809*** 0.005** 0.002 −0.009*** −0.011*** −0.053*** −0.021*** −0.001*** 0.000*** −4.057*** −0.469*** 26.031 0.07
12.68 2.47 1.10 −4.23 −7.40 −8.38 −4.97 −3.88 11.04 −22.33 −12.31 0.32
UK 0.963*** −0.004** −0.010*** −0.014*** −0.005*** −0.036*** −0.007** −0.001** 0.002*** −3.098*** −0.311*** 10.391 0.10
17.27 −2.04 −6.54 -8.72 −3.97 −10.16 −2.40 −2.55 5.73 −37.06 −198.50 0.41
JP 0.447*** −0.011*** −0.015*** −0.008*** −0.003*** 0.005*** −0.009** 0.003*** −0.001 *** −4.135*** −0.139 16.561 0.07
8.46 −8.21 −6.45 −3.81 −4.77 3.29 −2.29 4.91 −9.77 −10.49 −1.28 0.34
CN 0.783*** −0.013*** −0.012*** −0.025*** −0.011*** 0.006** 0.002 −0.003*** −0.002 *** −7.079*** −0.119*** 57.159 −0.01
11.94 −5.22 −6.06 −8.62 −15.73 2.54 0.69 −8.85 −5.56 −45.99 −5.43 0.45
KO 0.953*** −0.011*** −0.025*** −0.031*** −0.028 −0.027*** −0.045*** −0.003*** 0.002*** −2.502*** −0.180*** 35.688 0.05
15.37 −5.64 −8.77 −14.40 −22.67*** −6.87 −9.26 −21.03 3.21 −5.41 −3.47 0.44
KW 0.766*** −0.003 −0.001 −0.004** 0.003 0.034*** 0.005 0.003*** 0.002** −3.102*** −0.398*** 7.000 0.07
10.07 −1.61 −0.06 −2.16 1.24 3.81 1.06 3.78 2.32 −4.39 −4.00 0.71
ID 1.192*** 0.005** −0.021*** −0.014*** −0.001* −0.034*** 0.013*** −0.004*** 0.002 −5.322*** −0.349*** −0.216 0.04
11.49 2.25 −6.42 −3.88 −1.83 −3.40 2.71 −5.11 1.61 −4.73 −3.84 −0.15
IN 1.352*** −0.020*** −0.006** −0.007** 0.001*** −0.075*** −0.015** −0.001** 0.005*** −4.367*** −0.461*** 20.718 0.07
16.61 −9.28 −2.17 2.36 2.82 −8.59 −2.56 −2.09 6.27 −4.53 −4.32 0.53
SG 0.73*** −0.01*** −0.02*** −0.03*** −0.01*** 0.05*** −0.02*** −0.01*** 0.001*** −7.50*** −0.35*** 57.88 0.08
6.36 −3.73 −3.91 −12.80 −4.17 6.83 −4.57 5.19 9.21 −3.65 −5.53 0.72
BR 1.166*** −0.011** −0.013*** −0.011*** −0.011*** 0.015*** −0.003*** −0.020*** −0.002*** −3.503*** −0.291*** 31.347 0.10
18.37 −6.34 −4.48 −12.92 −7.86 14.31 7.51 −13.74 −11.52 −8.26 −2.95 0.37
MX 1.291*** −0.001*** −0.010*** −0.016*** −0.015*** 0. 042*** −0.014*** −0.004*** −0.003*** −2.252*** −0.167*** 2.359 0.11
22.70 −3.62 −6.52 −12.16 −7.48 11.35 −4.58 −8.00 −7.20 −7.48 13.19 0.20
  • Note: The equation is: R t = C + β 1 ∆ cpu t − 1 + β 2 ∆ cpu t − 2 + β 3 cov σ . ∆ cpu , t − 1 + β 4 cov σ , ∆ cpu , t − 2 ${R}_{t}=C+{\beta }_{1}{\unicode{x02206}{cpu}}_{t-1}+{\beta }_{2}{\unicode{x02206}{cpu}}_{t-2}+{\beta }_{3}{\text{cov}}_{\sigma .\unicode{x02206}{cpu},t-1}+{\beta }_{4}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ + β 5 ∆ p Gas , t − 1 + β 6 ∆ p Gas , t − 2 + ∆ β 7 emv PT , t − 1 + β 8 ∆ emv PT , t − 2 + β 9 σ GFC , t 2 + β 10 σ GFC , t 2 + $+\,{{\beta }_{5}{\unicode{x02206}p}_{{Gas},t-1}+{\beta }_{6}{\unicode{x02206}p}_{{Gas},t-2}+{\unicode{x02206}{\beta }_{7}{emv}}_{{PT},t-1}+{\beta }_{8}{\unicode{x02206}{emv}}_{{PT},t-2}+\beta }_{9}{\sigma }_{{GFC},t}^{2}+{\beta }_{10}{\sigma }_{{GFC},t}^{2}+$ ε t . ${\varepsilon }_{t}.$ The numbers in the first row are the estimated coefficients, the second row contains the z-statistics. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$ is the adjusted R-squared.

6 Evidence From Regional and World Performance

6.1 Evidence From Regional and World Models

Having shown the individual country's performance, this section extends the study to examine regional and world markets. The estimated results, which are reported in Table 8, merit several comments. First, the slopes of ∆ cpu t − 1 and ∆ cpu t − 2 $\unicode{x02206}{{cpu}}_{t-1}\text{and}\,\unicode{x02206}{{cpu}}_{t-2}$ are negative and highly significant. This evidence is consistent with anticipation that an upward shift in climate policy uncertainty produces a negative effect on energy stock returns for both regional and world levels. This result can be attributed to the transitional risk from climate risks (Boushey et al. 2021; Xu et al. 2023), which leads to switching from a dependence on fossil fuels toward industries with a low-carbon economy. Further, the risks may involve a change in the use of technology and carbon taxes, and an adjustment of investor and consumer sentiment as an adaptation to a greener environment is likely to cause financial instability (Persefoni Report 2024).

Table 8. Estimates of regional and world energy stock returns in response to dynamic changes in climate policy uncertainty ( Δ ${\rm{\Delta }}$ cput), covariance between Δ ${\rm{\Delta }}$ cput induced EMV, appreciation of oil price, and energy policy induced EMV change controlling for volatility in 2008 and 2020 crises.
Market C ∆ cpu t − 1 ${\unicode{x02206}{cpu}}_{t-1}$ ∆ cpu t − 2 ${\unicode{x02206}{cpu}}_{t-2}$ cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p oil , t − 1 ${\unicode{x02206}p}_{{oil},t-1}$ ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ ω i , 0 ${\omega }_{i,0}$ R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$
DEVL 0.863*** −0.006*** −0.021*** −0.003*** −0.001 −0.066*** −0.010*** −0.006*** −0.000*** −5.390*** −0.653*** 24.475 0.08
38.53 −19.33 −10.06 −3.06 −0.70 −6.81 −3.11 −25.12 −9.73 −22.28 −3.59 0.30
EMER 1.091*** −0.016*** −0.021*** −0.010*** −0.006** −0.001 0.012 −0.007*** 0.001 −5.575*** −0.562*** 40.039 0.07
12.60 −4.36 −8.48 −4.51 −2.39 −0.13 1.37 −7.48 0.71 −27.24 −3.20 0.39
GCC 0.960*** −0.010*** −0.025*** 0.023*** −0.031** 0.069*** 0.001*** −0.001*** −0.001*** −3.790** −0.305** 24.291 0.03
11.35 −6.74 −8.23 5.06 −12.27 4.67 6.31 −3.95 −4.27 −2.07 −2.45 0.74
A_XPAC 0.896*** −0.014*** −0.012*** −0.013*** −0.001** −0.031*** 0.002 −0.006*** 0.001 −5.464*** −0.447** 30.547 0.05
7.01 −7.13 −5.21 −5.98 −2.08 −5.42 0.73 −7.49 0.99 −8.64 −2.57 0.39
PAC 0.846*** −0.012*** −0.021*** −0.012*** −0.011*** −0.032*** −0.027*** −0.005*** 0.001*** −6.423*** −0.442*** 41.151 0.06
12.48 −4.89 −8.97*** −10.41 −3.28 −3.24 −6.92 −7.88 3.57 −7.54 −2.87 0.50
EMU 0.575*** −0.004** −0.012** −0.010*** −0.008** −0.042*** −0.03*** 0.001* 0.001 −5.250*** −0.291*** 45.696 0.07
4.49 −2.12 2.47 −3.88 −2.17 −4.17 −4.26 1.74 1.28 −6.97 −11.64 1.52
LATIN 1.341*** −0.016*** −0.020*** −0.024*** −0.003*** −0.062*** −0.044*** −0.009*** −0.003*** −5.652*** −1.149*** 75.218 0.07
15.53 −12.43 −9.04 −8.90 −9.29 −5.89 −8.24 −8.84 −4.69 −3.53 −22.88 0.35
WORLD 0.744*** −0.007*** −0.020*** −0.004*** −0.006*** −0.056*** −0.005*** −5.649*** −0.370*** 15.104 1.843 −0.147 0.11
8.38 −3.92 −6.68 −3.54 −4.77 −8.41 −5.14 −16.72 −3.07 1.32 0.81 −0.30
  • Note: The equation is: R t = C + β 1 ∆ cpu t − 1 + β 2 ∆ cpu t − 2 + β 3 cov σ , ∆ cpu , t − 1 + β 4 cov σ , ∆ cpu , t − 2 ${R}_{t}=C+{\beta }_{1}{\unicode{x02206}{cpu}}_{t-1}+{\beta }_{2}{\unicode{x02206}{cpu}}_{t-2}+{\beta }_{3}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}+{\beta }_{4}{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ + β 5 ∆ p oil , t − 1 + β 6 ∆ p oil , t − 2 + ∆ β 7 emv EN , t − 1 + β 8 ∆ emv EN , t − 2 + β 9 σ GFC , t 2 + β 10 σ ID , t 2 + $+{{\beta }_{5}{\unicode{x02206}p}_{{oil},t-1}+{\beta }_{6}{\unicode{x02206}p}_{{oil},t-2}+{\unicode{x02206}{\beta }_{7}{emv}}_{{EN},t-1}+{\beta }_{8}{\unicode{x02206}{emv}}_{{EN},t-2}+\beta }_{9}{\sigma }_{{GFC},t}^{2}+{\beta }_{10}{\sigma }_{{ID},t}^{2}+$ ε t . ${\varepsilon }_{t}.$ The dependent variable is the stock return R t ${R}_{t}$ in different regions and world at time t. The ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ is change in the climate policy uncertainty (CPU) expressed in natural logarithm . cov σ , ∆ cpu , t $.{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t}$ is the covariance between ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and equity market volatility ( σ $\sigma $ ); ∆ emv EN , t = cov σ , ∆ EN , t ${{\unicode{x02206}{emv}}_{{EN},t}=\text{cov}}_{\sigma ,\unicode{x02206}{EN},t}$ is the change in equity market volatility calibrated to energy and environment policy. ∆ emv EN , t ${\unicode{x02206}{emv}}_{{EN},t}$ is the change in equity market volatility calibrated to energy and environment policy. ∆ p oil , t $\unicode{x02206}{p}_{{oil},t}$ denotes change in Brent crude oil price. The σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ = EMV FC , t ${{EMV}}_{{FC},t}$ * I 2008 , EMV GFC , t ${{EMV}}_{{GFC},t}$ is equity market volatility calibrated to financial crisis, I 2008 is an indicator variable that sets to unity in 2008-global financial crisis and zero otherwise; σ ID , t 2 = ${\sigma }_{{ID},t}^{2}=$ EMV ID , t ${{EMV}}_{{ID},t}$ * I 2020 , where EMV ID , t ${{EMV}}_{{ID},t}$ is the equity market volatility calibrated to infectious disease, where I 2020 is an indicator variable that sets to unity in 2020 COVID-19 period and zero otherwise. The numbers in the first row are the estimated coefficients, the second row contains the z-statistics. ***, **, and * are significant at the 1%, 5% and 10% levels, respectively. The critical values of Z-distribution at the 1%, 5%, and 10% levels of significance are 2.63, 1.98, and 1.66, respectively. R ̅ 2 ${\mathop{R}\limits^{&#773;}}^{2}$ is the adjusted Rsquared.

Second, the coefficients of cov σ , ∆ cpu , t − 1 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ and cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ display negative signs and are statistically significant for all regions and worldwide. These indirect spillover effects from changes in climate policy tend to induce market fears and, in turn, stock volatility that causes stock returns to plummet and spread to different regions through a contagion effect (Chiang 2020). This effect is present not only in individual countries but also in regional performance.

Third, the estimates of ∆ emv EN , t − 1 ${\unicode{x02206}{emv}}_{{EN},t-1}$ are negative and statistically significant for all regions and world markets, indicating that the energy and environmental policy uncertainty tends to be globally prevalent. This result essentially reflects a business behavior that an increase in environmental policy uncertainty lowers firm-level investment in industries sensitive to environmental policies (Palikhe et al. 2024). The evidence is consistent with the finding reported by Wang and Kong (2022) and Alqahtani et al. (2019). However, the evidence of ∆ emv EN , t − 2 ${\unicode{x02206}{emv}}_{{EN},t-2}$ is less impressive and some of these estimates even turn to positive, reflecting a market reversal.

Fourth, the results for ∆ p oil , t − 1 ${\unicode{x02206}p}_{\text{oil},{\rm{t}}-1}$ and ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ are negative and significant at the 1% level. The exception is EMER region, where the coefficients are insignificant. The result of negative sign is consistent with the market phenomenon that a higher oil price adds costs in production, lowering the growth rate of economic activities (Hamilton 1983, 1996), and weakening the future cash flows from oil revenues (Jones and Kaul 1996; Salisu and Oloko 2015). However, for the oil-exporting region, such as GCC, oil price appreciation can bring about gains in stock markets. This is shown in the positive values of ∆ p oil , t − 1 ${\unicode{x02206}p}_{\text{oil},{\rm{t}}-1}$ and ∆ p oil , t − 2 ${\unicode{x02206}p}_{{oil},t-2}$ . Behind this is the fact that demand for the oil is inelastic, and prices of nonfossil fuels are not lower enough to attract users to make a substitute. Hamilton (1996), Hwang and Kim (2021), Nasir et al. (2018), and Abiad and Qureshi (2023), who argued that rising oil prices negatively affect stocks due to oil price shocks having a negative effect on economic activity.

Fifth, coefficients of financial market crisis, σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ , and COVID-19 pandemic, σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ , both show a negative impact and are strongly significant, indicating that during the periods of global financial crisis and COVID-19 pandemic, fears of market uncertainty increase stock market volatility, leading to dramatic downturn of stock returns worldwide. In summary, the world data, which aggregate countries' observations, eliminate individual heterogeneity of country factors, and therefore reach a more consistent result. While testing the data for regional and individual countries, some minor variations arise, reflecting local factors, as shown in higher oil production areas, such as in the GCC region and Kuwait, which exhibit stock price appreciation as a benefit of oil export.

6.2 Do the Parametric Estimates for Individual Countries (Regions) Deviate From the World Level?

Given the estimated coefficients reported in Tables 5–8, it is of interest to inquiry which specific countries (regions) may serve as the leading ones to explain the global performance or are there any countries' behavior that significantly deviated from the world level? To this end, we examine the restriction of no difference between individual country-and world-level coefficients relative to the respective standard errors. In formula we write:
β i , j − β w , j sd ( β i , j − β w , j ) $\frac{{{\beta }_{i,j}-\beta }_{w,j}}{{sd}({{\beta }_{i,j}-\beta }_{w,j})}$ ()
where Var ( β i , j − β w , j ) ${{\beta }_{i,j}-\beta }_{w,j})$ = Var ( β i , j ${\beta }_{i,j}$ ) + Var ( β w , j ${\beta }_{w,j}$ ) − 2 Cov ( β i , j − β w , j ) . $-2{Cov}({{\beta }_{i,j}-\beta }_{w,j}).$ β i , j ${\beta }_{i,j}$ is the jth coefficient in country $\text{country}$ ith stock return equation; β w , j ${\beta }_{w,j}$ is the jth coefficient in the world stock return equation, sd stands for standard deviation, where sd ( β i , j − β w , j ) ${sd}({{\beta }_{i,j}-\beta }_{w,j})$ = Var ( β i , j − β w , j ) . $\sqrt{\text{Var}({{\beta }_{i,j}-\beta }_{w,j})}.$

The test results are contained in Table 9, where coefficient with “ ~ ” denotes $\unicode{x0007E}\mbox{''}\,\text{denotes}$ coefficient of country i deviating from that of world's level. For instance, coefficient of ∆ cpu ̃ iw , t − 1 ${\widetilde{\unicode{x02206}{cpu}}}_{{iw},t-1}$ = coefficient of ∆ cpu i , t − 1 − coefficent of ∆ cpu w , t − 1 ${\text{coefficient of}\unicode{x02206}{cpu}}_{i,t-1}{-\text{coefficent of}\unicode{x02206}{cpu}}_{w,t-1}$ . The null hypothesis is that the coefficient of ∆ cpu ̃ iw , t − 1 = 0 ${\widetilde{\unicode{x02206}{cpu}}}_{{iw},t-1}=0$ , indicating no difference in the coefficients between ∆ cpu i , t − 1 and ∆ cpu w , t − 1 . ${\unicode{x02206}{cpu}}_{i,t-1}{\text{and}\unicode{x02206}{cpu}}_{w,t-1}.$ That is, β i , ∆ Cpu t − β w , ∆ Cpu t ${{\beta }_{i,{\unicode{x02206}{Cpu}}_{t}}-\beta }_{w,{\unicode{x02206}{Cpu}}_{t}}$ = 0. As anticipated, the country's coefficients that deviate from the world level vary from country (region) to country (region). However, there are some parallel behaviors due to associated with similar economic states, geographic locations, the establishment of international agreement/organization, and the sources of uncertainty, among others, although some countries appear far from the world level. First, we check the deviation of each country's estimated coefficient from the world level. Evidence shows that coefficients from US and CA act very similarly to their deviation from the world level. A similar trend was observed in the three European countries – FR, GM, and IT. However, for the UK and MX, the coefficients are more consistent with those at the world level. Grouping behavior is also exhibited in regions for the DEV and EMU countries. Second, there are countries where the coefficients are significantly deviated from that of world level, including GM, IT, ID, and IN. However, the UK is the most consistent market with the world's behavior. Third, from a regional perspective, the null hypothesis for EMER, GCC, A_XPAC, and LATIN is mostly rejected, as shown in significantly different from the world level. The testing results suggest that world-level behavior is mainly driven by the developed region. Fourth, among different variables, 10 out of 15 countries the null is rejected for the coefficient of ∆ ̃ p oil , t − 1 ${\widetilde{\unicode{x02206}}p}_{{oil},t-1}$ , reflecting that oil price behavior for the majority of countries significantly deviates from world level. Further, evidence shows that none of the nulls is rejected for σ ̃ ID , t 2 ${\widetilde{\sigma }}_{{ID},t}^{2}$ , confirming that COVID-19 appears to be a worldwide phenomenon. While testing the regional performance, none of the null for σ ̃ GFC , t 2 ${\widetilde{\sigma }}_{{GFC},t}^{2}$ and σ ̃ ID , t 2 ${\widetilde{\sigma }}_{{ID},t}^{2}$ are statistically significant, except the measure of σ ̃ ID , t 2 in LATIN region , ${\widetilde{\sigma }}_{{ID},t}^{2}\text{in LATIN region},$ indicating the regional performance consistently moves along with the world level when global financial crisis and COVID-19 hit the market, although a heterogeneous effect is present in different countries.

Table 9. Estimates of regional and world energy stock returns in response to dynamic changes in climate policy uncertainty ( Δ ${\rm{\Delta }}$ cput), covariance.
Coeff. Diff C ∆ cpu ̃ t − 1 ${\unicode{x02206}\widetilde{{cpu}}}_{t-1}$ ∆ cpu ̃ t − 2 ${\unicode{x02206}\widetilde{{cpu}}}_{t-2}$ cov ̃ σ , ∆ cpu , t − 1 ${\widetilde{\text{cov}}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov ̃ σ , ∆ cpu , t − 2 ${\widetilde{\text{cov}}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p ̃ oil , t − 1 ${\widetilde{\unicode{x02206}p}}_{{oil},t-1}$ ∆ p ̃ oil , t − 2 ${\widetilde{\unicode{x02206}p}}_{{oil},t-2}$ ∆ emv ̃ EN , t − 1 ${\widetilde{\unicode{x02206}{emv}}}_{{EN},t-1}$ ∆ emv ̃ EN , t − 2 ${\widetilde{\unicode{x02206}{emv}}}_{{EN},t-2}$ σ ̃ GFC , t 2 ${\widetilde{\sigma }}_{{GFC},t}^{2}$ σ ̃ ID , t 2 ${\widetilde{\sigma }}_{{ID},t}^{2}$
US-World 0.137 0.006*** 0.003 −0.000 −0.002 −0.044*** 0.012 0.000 0.000** 1.164 −0.462
z-statistic 1.34 3.91 1.20 −0.59 −0.96 −3.67 1.30 −0.01 −2.10 1.64 −1.94
CA-World 0.101 −0.005** 0.004 0.018*** −0.002 0.045*** 0.006 0.002 0.000 0.915 −0.141
z-statistic 1.09 −2.44 −1.20 7.12 −0.92 5.43 0.71 1.46 −0.72 1.02 −0.93
FR-World −0.328** −0.001 0.010** 0.009*** −0.012*** 0.005 0.021*** 0.004*** 0.001*** 3.384*** 0.183
z-statistic −2.15 −0.46 2.65 5.97 −4.39 0.52 2.82 3.37 4.04 4.55 1.13
GM-World −0.77*** 0.028*** −0.006* −0.012** −0.006*** 0.091*** −0.055*** −0.001 0.004*** −2.870 ** −0.002
z-statistic −10.64 5.71 −1.95 −2.00 −4.73 9.53 −3.63 −0.81 24.06 −2.26 −0.01
IT-World −0.383*** 0.011*** 0.008*** 0.005* −0.010*** −0.024** 0.030*** 0.002 −0.004*** 2.106*** 0.071
z-statistic −3.38 4.94 2.76 1.74 −5.24 -2.11 3.78 1.34 −6.58 3.01 0.56
UK-World 0.451*** 0.002 0.000 −0.003 0.002 0.007 0.021 0.004*** −0.003** 5.493*** 0.028
z-statistic −3.04 0.50 0.09 −0.97 0.90 0.49 1.63 3.40 −2.19 7.12 0.21
JP-World −0.546*** 0.002 −0.012*** −0.005 −0.005** 0.034*** −0.013 0.004*** 0.002*** −1.737** −0.034
z-statistic −3.89 0.64 −3.43 1.58 −2.06 2.76 −1.13 3.09 7.79 −1.98 −0.17
CN-World −0.092 −0.002 0.001 −0.007*** −0.008** 0.071*** 0.069*** 0.000 0.000 −0.937 0.300
z-statistic −0.82 −0.88 0.26 −3.54 −2.52 5.93 7.66 0.20 0.93 −1.61 1.94
KO-World −0.225** 0.000 0.011*** 0.023*** −0.006** 0.014 0.055*** 0.004*** 0.001*** 3.146*** 0.194
z-statistic −2.17 0.19 4.34 8.82 −2.58 1.16 4.88 3.45 4.25 4.66 1.20
KW-World −1.025*** 0.002 0.011*** −0.005** 0.003 0.048*** 0.106*** 0.001 −0.007*** 5.637*** 0.215
z-statistic −7.97 1.00 3.68 −2.29 1.35 5.96 10.47 1.04 −10.97 12.24 1.38
ID-World 0.72*** −0.043*** −0.011*** −0.010*** 0.008*** 0.067*** 0.257*** −0.002 −0.009*** −6.846*** 0.286
z-statistic 5.02 −7.91 −3.27 −9.10 6.86 8.36 14.02 −1.24 −5.13 −6.67 1.97
IN-World 0.097 −0.012*** 0.014*** −0.011*** 0.012*** 0.034*** 0.064*** 0.003** 0.002* 1.854 0.012
z-statistic 0.71 −3.96 5.68 −4.50 3.99 2.96 5.70 2.07 1.82 1.52 0.05
SG-World −0.190** −0.005 0.003 −0.029*** −0.006*** 0.141*** 0.001 −0.002 0.001** −1.711 0.054
z-statistic −1.98 −1.59 0.67 −9.77 −4.95 11.99 0.11 −1.26 2.62 −0.78 0.36
BR-World 0.705*** −0.019*** −0.015*** −0.026*** 0.002 0.020 0.021*** −0.002 0.000 9.348*** −0.233
z-statistic 3.18 −4.62 −3.02 −3.72 0.99 1.45 2.76 −0.98 0.23 10.62 −1.15
MX-World 0.009 0.000 −0.002 −0.002 0.000 −0.003 0.001 0.003** 0.000 −0.762 −0.257
z-statistic 0.07 0.12 −0.71 −1.60 0.16 −0.25 0.08 2.46 0.41 −1.19 −1.22
Coeff. Diff C ∆ cpu ̃ t − 1 ${\unicode{x02206}\widetilde{{cpu}}}_{t-1}$ ∆ cpu ̃ t − 2 ${\unicode{x02206}\widetilde{{cpu}}}_{t-2}$ cov ̃ σ , ∆ cpu , t − 1 ${\widetilde{\text{cov}}}_{\sigma ,\unicode{x02206}{cpu},t-1}$ cov ̃ σ , ∆ cpu , t − 2 ${\widetilde{\text{cov}}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ ∆ p ̃ oil , t − 1 ${\widetilde{\unicode{x02206}p}}_{{oil},t-1}$ ∆ p ̃ oil , t − 2 ${\widetilde{\unicode{x02206}p}}_{{oil},t-2}$ ∆ emv ̃ EN , t − 1 ${\widetilde{\unicode{x02206}{emv}}}_{{EN},t-1}$ ∆ emv ̃ EN , t − 2 ${\widetilde{\unicode{x02206}{emv}}}_{{EN},t-2}$ σ ̃ GFC , t 2 ${\widetilde{\sigma }}_{{GFC},t}^{2}$ σ ̃ ID , t 2 ${\widetilde{\sigma }}_{{ID},t}^{2}$
DEV-World −0.013 0.002 −0.005 −0.001 0.004** -0.012 0.012 −0.002 −0.001 0.176 −0.273
−0.17 0.96 −1.54 −0.90 2.50 -0.99 1.47 −1.31 −0.62 0.34 −1.23
EMER-World 0.215* −0.009** −0.005 −0.008*** −0.001 0.052*** 0.033*** −0.002 0.000*** −0.009 −0.182
1.91 −2.12 −1.41 −3.58 −0.34 4.09 2.97 −1.65 2.89 −0.02 −0.84
GCC-World 0.080 −0.005* −0.010** 0.019*** −0.023*** 0.128*** 0.026*** 0.004*** 0.000 1.553 −0.205
0.80 −1.69 −2.43 4.21 −15.78 8.46 3.42 3.04 −0.23 0.85 −1.19
A_XPAC-World 0.020 −0.007** 0.004 −0.011*** 0.004*** 0.022** 0.024*** −0.002 0.001 0.102 −0.067
0.14 −2.61 1.28 −5.04 4.01 2.28 3.03 −1.21 1.11 0.13 −0.31
PAC-World −0.028 −0.003 −0.005 −0.010*** 0.853*** 0.019 −0.005 0.000 0.001*** −0.855 −0.063
−0.20 −0.88 −1.12 −3.22 3.92 1.56 −0.52 −0.23 5.16 −0.91 −0.32
EMU-World −0.301** 0.003 0.004 −0.008*** −0.003 0.011 -0.008 0.006*** 0.001 0.316 0.089
−2.05 1.30 0.64 −3.12 −0.78 0.89 −0.79 4.03 1.45 0.36 0.69
LATIN-World 0.465*** −0.008*** −0.004 −0.022*** 0.003** −0.009 −0.022** −0.004*** −0.003*** −0.086 −0.769***
4.13 −4.14 −1.13 −8.13 2.39 −0.67 −2.44 −2.86 −4.45 −0.05 −5.66
  • Note: Coefficients with “ ~ ” denote $\unicode{x0007E}\mbox{''}\,\text{denote}$ the deviation from that of world's coefficients. For instance, ∆ cpu ̃ t − 1 ${\unicode{x02206}\widetilde{{cpu}}}_{t-1}$ = ∆ cpu i , t − 1 , − ∆ cpu w , t − 1 ${\unicode{x02206}{cpu}}_{i,t-1},{-\unicode{x02206}{cpu}}_{w,t-1}$ , so does the other variables. The null hypothesis is : ∆ cpu ̃ t − 1 = 0 $:\,{\unicode{x02206}\widetilde{{cpu}}}_{t-1}=0$ , implying that the coefficients for country (region) i is not statistically significant from that of the world level. ***, **, and * are signficiant at the 1%, 5% and 1% levels. The bold face values in Table 7 indicate the rejection of the null at the 5% or better level.

7 Conclusions and Implications

This paper utilizes the GED-GARCH model to examine global stock returns in response to the effects of dynamic changes in climate policy uncertainty, appreciation of oil price returns, and other uncertainty factors. Evidence from this study consistently shows that the coefficients of lagged ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ present negative values; this result holds true for individual countries, regions, and the world level as investors perceive that climate policy uncertainty could damage economic activity and carbon costs that jeopardize their future investment performance. However, the exceptions are Germany and Italy (in some test equations) that exhibit positive signs, which can be caused by spurious correlations or investors belief that ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ can lead to a positive NPV on their stocks.

This study also introduces the lagged values of induced risks arising from the covariance between ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ and equity market volatility, as well as covariance of change in equity market volatility calibrated to energy & environmental policy. The evidence of these two factors, which have a negative influence on the stock returns, represents the spillover effect from a change in climate policy uncertainty and energy & environmental policy uncertainty that create economic damages, which in turn raise both private costs and social costs, leading to an adverse effect on stock returns.

We controlled for both 2008s global financial crisis and 2020s infectious disease due to COVIV-19 by using the variance variables, σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ and σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ . The coefficients of these two variables are negative and highly significant, indicating the damaging effects of these events on global stocks. The evidence is consistent with Salisu and Vo (2020), Salisu and Adediran (2020), Liu et al. (2020), Terry et al. (2020), and Cheema et al. (2022). The inclusion of these variables also helps mitigate the estimated bias that would occur otherwise.

The evidence in this study also contends that heightened oil prices produce a negative effect on economic growth (Hamilton 1983; Hamilton and Herrera 2004) and depress stock returns. This holds true at the individual country, region and world levels. However, for the oil-exporting region, GCC and country of Kuwait, the higher oil prices increase the flow of funds into these areas, which has a positive impact on the stock prices (Osah and Mollick 2022). Note that in examining the oil price change and its impact on stock returns, the risk factors, including ∆ cpu t , cov σ , ∆ cpu , t , ∆ emv EN , t $\unicode{x02206}{{cpu}}_{t},{\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t},{\unicode{x02206}{emv}}_{{EN},t}$ , σ GFC , t 2 ${\sigma }_{{GFC},t}^{2}$ , and σ ID , t 2 ${\sigma }_{{ID},t}^{2}$ , or at least some of them, were excluded from the empirical test in the standard model. Further, the evidence, which suggests the explanatory variables are significant at the lagged values, is consistent with real options behavior that posits investors should delay their decisions and wait for better timing to make a choice for managing portfolio, a strategy that would reduce costly consequences (Pindyck 1991).

While testing the parametric impact from different countries (region) against world level, evidence shows that advanced countries appear to have close agreement with the world level because the null of no difference in the estimated coefficients cannot be rejected in most cases. Evidence indicates that only the variance coefficients of pandemic for each country deviated from that of world level cannot be rejected uniformly, indicating the pandemic is a worldwide phenomenon affecting stock returns.

The implications of this study can help investors and policy makers identify both the direct effect and induced effect of ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ , whose spillover carries a negative impact on stock returns. These indirect effects were ignored in previous studies that examined stock returns in relation to ∆ cpu t ${\unicode{x02206}{cpu}}_{t}$ . This approach uses a broad uncertainty model specification that includes climate policy spillover risk, energy & environmental policy risk, the turbulence of financial crises, and the impact of the pandemic. This approach is consistent with the notion by Barnett et al. (2022). These uncertainties should be priced into stocks by rewarding different risk premiums, and the evidence also provides advice regarding the heterogenous effects on different countries' stock returns. This study is limited by using total change of CPU. The extension of the research can be done by dividing climate risk into physical climate change and transition climate change risks and examine their effects on stock returns. This will be the future reserach.

Acknowledgments

This paper has benefited from presentation to seminar at the National Yang Ming Chao Tung University. Professors Jin-Li Hu, Shuh-Chyi Doog, Sheng-Yung Yang, Fu-Lai Lin and seminar participants provided constructive comments and suggestions. There was no funding support from individuals or institutions.

    Disclosure

    The idea and empirical work are done and responsible by the author. The author has nothing to report.

    Ethics Statement

    The author has nothing to report.

    Conflicts of Interest

    The author declares no conflicts of interest.

    Endnotes

  1. 1 The Environmental Protection Agency (EPA) provides a clear definition and offers statements about climate-related financial risks and opportunities. https://www.epa.gov/climateleadership/climate-risks-and-opportunities-defined; https://www.epa.gov/climateleadership/climate-risks-and-opportunities-defined.
  2. 2 The relationship between stock returns and oil prices has been extensively investigated in the literature. Studies by Sadorsky (1999; 2001), Boyer and Filion (2004), El-Sharif et al. (2005), Lescaroux and Mignon (2008), and Wang et al. (2013) found a positive relationship between stock returns and oil prices using various data. However, a negative relationship between stock returns and oil prices has been reported in studies by Jones and Kaul (1996), Filis (2010), Hamilton (2011), and Diaz et al. (2016). Smyth and Narayan (2018) provided a survey, and Degiannakis et al. (2018) reviewed the relationship between stock returns and oil prices.
  3. 3 The term of cov σ , ∆ EN , t used in this study ${\text{cov}}_{\sigma ,\unicode{x02206}{EN},t}\text{used in this study}$ is equivalent to ∆ emv EN , t ${\unicode{x02206}{emv}}_{{EN},t}$ (Terry et al. 2020) in the empirical study. The measure of this variable can be found in Economic Policy Uncertainty Index.
  4. 4 The geopolitical risk (Caldara and Iacoviello 2022; Chiang 2021), downside risk (Bali et al. 2009; Chen et al. 2018), and other categorical uncertainties in Baker et al. (2016, 2019). However, this study only includes uncertainty related to climate and energy that reduces the possibility of over parametrization. Campiglio et al. (2023) provided a good survey on the climate-related risks to stock markets.
  5. 5 Literature suggests that “damages” as a function of temperature and temperature increases are attributable to carbon concentration (Morgan et al. 2017; Pezzey. 2019; Barnett et al. 2020).
  6. 6 Here, the study only refers a few articles on the relationship between stock returns and oil prices to provide a rationale. A detailed review is available from Smyth and Narayan (2018), Degiannakis et al. (2018), and Bagirov and Mateus (2024).
  7. 7 On May 5, 2023, the World Health Organization (WHO) declared an end to the global Public Health Emergency (PHE) for COVID-19. The US Department of Health and Human Services (HHS) declared the same for the United States on May 11. Because vaccines and treatments were provided much earlier, this study sets March 2020 – September 2022 as the pandemic spread time.
  8. 8 The GARCH (1,1) model was popularized by Bollerslev et al. (1992) to obtain a better fit of the stock return equation. Bollerslev (2010) summarized the specifications of ARCH-type models. Note that Equation (3) contains no asymmetric terms for the lagged shock because it is statistically insignificant.
  9. 9 The original specification by Nelson (1991) and the elaboration by Purczynski (2014) provided a detailed illustration of the GED density function.
  10. 10 Following He and Zhang (2022), CPU t − 1 ${{CPU}}_{t-1}$ and CPU t − 2 ${{CPU}}_{{\rm{t}}-2}$ , which are used in the test equation and the estimated results for the US market, are positive. Our evidence suggests that coefficients of CPU t − 1 ${{CPU}}_{t-1}$ and C P U t - 2 $CP{U}_{{\rm{t}} \mbox{-} 2}$ are positive and statistically significant in the long run.
  11. 11 Note that cov σ , ∆ cpu , t − 1 and ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-1}\text{and}$ cov σ , ∆ cpu , t − 2 ${\text{cov}}_{\sigma ,\unicode{x02206}{cpu},t-2}$ should be expressed as cov σ t , ∆ cpu , t − 1 and ${\text{cov}}_{{\sigma }_{t},\unicode{x02206}{cpu},t-1}\text{and}$ cov σ t , ∆ cpu , t − 2 ${\text{cov}}_{{\sigma }_{t},\unicode{x02206}{cpu},t-2}$ , respectively. We have dropped subscript t of σ $\sigma $ to save space and to fit the table width through the entire article.
  12. 12 The energy uncertainty index, EUI t (Dang et al. 2023) was also used to estimate the test equation in literature. Due to the EUI t series, which includes relatively shorter observations and does not cover all countries and regions under investigation, we did not report the results.

      13. The full text of this article hosted at iucr.org is unavailable due to technical difficulties.