The Impact and Consequences of Economic Policy Uncertainty

Uncertainty about economic policy can be traced back as far as 1921. Knight strictly distinguished between uncertainty and risk and proposed that uncertainty refers to the risk that individuals or enterprises cannot accurately predict future gains and losses in economic activities [1]. Bloom et al. pointed out that, on the one hand, uncertainty is a vague concept. It reflects the uncertainty in the minds of consumers, managers, and decision-makers about the possible future. On the other hand, it is also a broad concept, which includes the uncertainty of macroeconomic indicators such as GDP growth and micro-level fluctuations such as corporate growth rate [2]. At the same time, non-economic events such as war and climate change are also considered uncertain. Gulen and Ion defined it as the inability of economic agents to accurately predict whether, when, and by what means the government will change existing economic policies [3]. Rao et al. proposed that the uncertainty of economic policies is caused by the uncertainty of the government’s future direction and implementation intensity, which are closely related to economic and social development and the production and operation of enterprises [4]. Based on the above research results, this paper defines the uncertainty of economic policies as the uncertainty of the time when the economic policies issued by the government will be issued, the direction of the policies and the intensity of their implementation, etc., which results in the economic subjects not being able to judge precisely the direction and trend of the future policy development, which is a hidden factor affecting the macroeconomic development.

From a macro perspective, most studies show that the uncertainty of economic policies will negatively affect macroeconomic indicators. Jin et al. conducted research using the factor-augmented vector autoregressive model. The empirical test shows that by reducing the expectation of economic entities for the future market, the uncertainty will harm macroeconomic indicators, including GDP, exchange rate, inflation, consumption, investment, exports, and house prices. Among them, the exchange rate and inflation were the most affected [5]. Jin and Zhang regressed the macroeconomic variables with the uncertainty index of four policies: finance, currency, trade, exchange rate, and capital account. Empirical research shows that the impact of uncertainty on economic output is significantly harmful [6].

From the micro perspective, enterprises’ investment behavior, cash holdings, and innovation are the main research objects in the existing literature. Bloom et al. proposed that when the investment is irreversible, the increase in uncertainty will lead to a significant decrease in the response of enterprises’ investment behavior to any stimulus policies [2]. Chen and Wang believe that the increase in capital cost and the reduction of the marginal rate of return are the two main ways to reduce the investment enthusiasm of enterprises under uncertainty [7]. Xu and Dong, based on the double difference propensity score matching model test, found that the negative impact of policy uncertainty on the investment decisions of private enterprises is more prominent [8]. Gulen and Ion pointed out that economic policy uncertainty hinders capital investment at this stage by increasing the return on deferred investment [9]. Rao et al. Considering that uncertainty will lead to a decrease in investment willingness on the one hand, and a more cautious investment attitude towards market expectations, on the other hand, will also improve investment efficiency [4].

The Influencing Factors of Carbon Emission Intensity

Current studies on the factors influencing carbon emissions are broadly categorized into the following areas: economic development and population growth, energy structure, industrial structure, technological level, level of financial development, policies, and regulations. Economic activities and industrial development are the primary sources of carbon emissions. Major economic sectors such as industry, transportation, energy production, and construction usually involve large amounts of energy consumption and combustion processes, thus releasing large amounts of greenhouse gases such as carbon dioxide. Population growth and changes in consumption behavior also affect carbon emission levels. A growing population increases the demand for energy and goods, which may lead to more carbon emissions. Moreover, changes in people’s propensity to consume, their mode of transportation, and their lifestyles all impact carbon emissions.

Energy structure plays a decisive role in carbon emission intensity. Different energy sources have different carbon emission factors; fossil fuels usually have high carbon emission factors, while renewable energy sources usually have low carbon emission. Therefore, transforming the energy mix is crucial for reducing carbon emissions. Optimization of energy consumption structure suppresses carbon emissions. Soytas and Sari suggest that increasing the share of clean energy in the energy consumption structure is a critical way to reduce carbon emissions from the manufacturing sector in Turkey [10]. Wang et al. used the scenario prediction method and Markov chain model to verify the carbon emission reduction effect of energy consumption structure optimization under various combination scenarios and found that the carbon emission intensity decreases to different degrees under all scenarios, which indicates that the optimization of the energy consumption structure has a particular potential to contribute to the achievement of carbon emission reduction targets [11].

Optimization of industrial structure is a critical way to reduce carbon emissions, and the results of Hammond and Brännlund et al. show that industrial upgrading of the manufacturing industry can significantly reduce carbon dioxide emissions and can play an essential role in achieving carbon emission reduction [12, 13]. Liu et al. constructed a PVAR model, and the analysis results show that the two-way interaction between industrial structure upgrading and carbon emission efficiency has basically been formed in the eastern region of China, while the coordinated development between the three variables has not yet been realized in the central and western regions [14].

Advanced technology and efficient energy utilization can reduce carbon emission intensity. Technological innovation can improve the efficiency of production and energy utilization and reduce carbon emissions per unit of output or consumption. For example, adopting energy-saving technologies, energy management systems, and cleaner production methods can reduce carbon dioxide emissions. Yang Lisha et al. pointed out from both theoretical and empirical analysis that technological progress is a significant contributing factor to achieving carbon emission reduction in China [15]. Yin He et al. argued that low-carbon technological progress ultimately promotes the realization of carbon emission reduction by promoting new energy research and development, optimizing energy structure, and reducing energy consumption intensity. In addition, by introducing economic level, energy structure, urbanization level, and other factors as threshold variables into the model, it can be found that there is also a specific threshold effect on the carbon emission reduction effect of low-carbon technological progress [16].

Financial development will affect carbon emissions. Katircioglu and Taspinar suggest that the government set long-term financial development and energy conservation and emission reduction goals and exert the carbon mitigation effect of financial development in the long term [17]. Based on the enterprise level, have pointed out that financial development reduces the cost of enterprise financing, and the expansion of reproduction by enterprises will increase energy consumption and raise carbon emissions [18-20].

Government policies and regulations have a significant impact on carbon emissions. Governments can influence the level of carbon emissions by formulating and implementing carbon pricing mechanisms, energy efficiency standards, greenhouse gas emission limitation policies, and measures to support the development of clean energy. The strength and effectiveness of policies and regulations play a crucial role in reducing carbon emissions. Deng investigates through spatial Durbin modeling that the strength of environmental regulations reduces carbon emissions, and the development of green finance similarly reduces carbon emissions and improves environmental sustainability [21].

Research on the Impact of Economic Policy Uncertainty on Carbon Emission Intensity

The research on the impact of economic policy uncertainty on carbon emissions is a little extensive at present. Jiang et al. estimated the relationship between EPU and carbon emissions for the first time using industry-level carbon emissions data in the United States. They believe that EPU can affect emissions through two channels: changes in environmental governance and damage to corporate performance. Through the first channel, enterprises will reduce their efforts to reduce emissions; through the second channel, a company’s emissions performance may be ambiguous: low emissions due to poor performance or high emissions due to a shift to cheaper but more polluting fuels. They found that the US EPU affects the uncertainty of carbon emissions and has a positive impact on the growth of carbon emissions at the tail of the emission growth distribution [9]. Adedoyin and Zakari used annual data from 1985 to 2017 to study the UK’s energy consumption and emissions framework. Research shows that the EKC hypothesis holds: low EPU will reduce the growth of carbon dioxide emissions in the short term, while the long-term effect is positive. They believe a higher EPU will hinder the company’s investment and economic growth, thus reducing environmental concerns. In a depressed environment, the industry will compensate for the low turnover rate by switching to cheaper energy for production. However, as time passes, as these industries increase their revenues, they may switch to cleaner energy for production, thereby reducing emissions [22].

Pirgaip and Dinçergök investigated the causal relationship between economic policy uncertainty (EPU) and energy consumption and carbon emissions in G7 countries and also explored the one-way causal relationship between energy consumption to carbon dioxide emissions. Based on the overall survey results, it is strongly recommended that the Group of Seven countries formulate policies to reduce energy consumption and carbon dioxide emissions according to the recent climate mission, considering the negative impact of EPU on energy conservation [23]. The study by Adams et al. the World Uncertainty Index is used to analyze the long-term relationship between economic policy uncertainty and energy consumption in countries with high geopolitical risk from 1996-2017. The results show that energy consumption and economic growth contribute to carbon dioxide emissions [24]. In addition, there is a significant correlation between economic uncertainty and carbon dioxide emissions in the long run. Based on the analysis of the panel noncausal test method by Dumitrescu and Hurlin, it is shown that there is a two-way relationship between carbon dioxide emissions and energy consumption, economic policy uncertainty and carbon dioxide emissions, economic growth, and carbon dioxide emissions [25]. Yu et al. estimated the impact of EPU on the carbon emission intensity of manufacturing enterprises using unbalanced panel data of enterprises and the newly-built provincial EPU index in China. They further tested channels through which EPU could affect the company’s emission intensity, including innovative channels, the share of fossil fuels in the total energy consumption channel, and energy intensity channels. The results show that manufacturers prefer to use cheap and dirty fossil fuels to cope with the growing EPU [26].

Based on social and political theory, Gray et al. mentioned in the article corporate social reporting CSR the data in the CSR report indicate the significant changes in social disclosure behavior during the whole period. The article’s explanation of these trends is inevitably speculative, and further research is needed on this vertical basis. In particular, corporate social responsibility is, at best, a marginal activity in corporate practice [27]. Social and environmental performance remains a relatively low priority for companies. Researchers have to ask the following questions: firstly, can the importance of corporate social responsibility be raised by paying more attention to such marginal activities? Secondly, whether this will provide an opportunity for the development of “anti-hegemony.” Social and political theory points out that enterprises will display appropriate information when facing external pressure from those interested in them. From this theory, we know that releasing information about carbon emissions will enable enterprises to show the outside world that they have good carbon emission performance. At the same time, it also meets the transparency requirements of other relevant parties.

Based on the fundamental option theory, Stewart Myers pointed out that the benefits generated by the cash flows are closely related to the use of current assets and the choice of future investment opportunities for an investment plan. This means that the company can obtain a new right in the future: to buy or sell its own physical goods or investment opportunities at a fixed price. Therefore, it can evaluate the investment in physical goods by evaluating general options. As the actual commodity is the object of the rights and obligations of both parties, the option is a real option [28]. Abel and Eberly used a news-based policy uncertainty index to demonstrate a strong negative correlation between corporate-level capital investment and the overall level of uncertainty, which is related to future policy and regulatory outcomes [29]. Policy uncertainty can inhibit corporate investment by causing a precautionary delay caused by the irreversibility of investment. When the investment cannot be reversed, the investment plan owned by the economic entity is regarded as the resources it owns, which is the view of fundamental option theory. When the EPU rises, the value of the options owned will be increased first, and the net benefits generated by “waiting” will be increased simultaneously. However, with a further increase in the value of the options, the economic entities will have higher marginal investment costs, and the net benefits of investment will then decrease.

Based on the signaling theory, Stiglitz observed that some people want to convey information. In contrast, others do not want to convey information, but “in either case, the fact that actions convey information will lead people to change their behavior, which is why information defects have such far-reaching impact.” The signal transmission theory provides a unique, practical, and verifiable point of view for social choice under imperfect information. If companies publish more information about carbon emissions, stakeholders will put less pressure on the environment [30]. For example, if companies or factories release information about carbon emissions and performance in advance, the pressure on the environmental control department will significantly reduce. Signal transmission theory points out that to send the signal of “environmental protection in low-carbon life” to the stakeholders, the senior management of the enterprise actively publishes relevant information, such as carbon emissions and carbon performance, so that investors can have a better assessment of the enterprise and have some influence on the decision-making of investors. Enterprises adopt high-energy consumption and low-cost production methods to reverse the expected downward net income trend due to EPU. At the same time, due to the lack of information disclosure, investors will maintain confidence in the investment in energy-intensive production. According to these theories, EPU will affect the economic entity’s external business environment, affecting the economic entity’s external business environment and the economic entity’s decision-making. At the same time, carbon emissions are closely related to the production decisions of microeconomic entities.

With the increase of EPU, the government’s attention on environmental governance will be shifted and reduced, and the implementation of some environmental protection policies will be adversely affected. For example, withdrawing the United States from the Paris Agreement will increase the EPU, which may affect the state government’s determination to reduce carbon emissions. Besides, EPU may damage the enterprise’s overall economic condition and performance. On the one hand, reduce economic demand for energy consumption and then reduce carbon emissions. On the other hand, due to the poor economic situation, businesses and residents may choose to use traditional and cheaper energy sources, such as coal and oil, which may produce more carbon emissions. In addition, facing the high EPU, enterprises can expect the government to relax the requirements on environmental governance, which may cause enterprises to make less effort to control carbon emissions.

Based on the above literature analysis, we put forward hypothesis H1: the increase of economic policy uncertainty will increase the regional carbon emission intensity, and the impact of economic policy uncertainty on carbon emission intensity has a spatial spillover effect.

Data and Variables

Due to the availability and completeness of the data, we selected the data of 30 provincial-level regions except for Hong Kong, Macao, Taiwan, and Tibet from 2004-2017. All variables are derived from the EPS data platform, the WIND data platform, China Statistical Yearbook, China Environment Statistical Yearbook, and provincial statistical yearbooks. All continuous variables were winsorized (up and down 1%).

(1) Dependent variable: carbon emission intensity

The per capita CO₂ emissions (metric tons per capita) proxied to carbon emission intensity. Carbon dioxide emission is an indicator of environmental pollution in the present research. CO₂ emissions were calculated using the following formula:

where CE represents CO₂ emissions; i denotes the different types of energy; Q_i means the stand coal consumption of energy i; and C_i is the carbon emissions coeffcient of energy i. 44/12 represents the molar ratio of CO₂ to C. The standard coal consumption (SCE) conversion coeffcient and CO₂ emission coeffcient for different energy types are displayed in Table 1. The unit of heat conversion to SCE is tSCE/109 kJ; the unit of electricity conversion to CO₂ emissions is 104 t C/107 kWh. Source: IPCC (2006).

(2) Core independent variable: economic policy uncertainty

The core explanatory variable is economic policy uncertainty (EPU). With reference to Yu et al.'s practice, the index selects the daily newspapers of 30 provinces (municipalities and autonomous regions) in China as the sources of news media reports, determines the basic entries about the uncertainty of economic policies by using the method of manual selection of high-frequency words and manual reading, calculates the proportion of annual target articles in 30 provinces (municipalities and autonomous regions) in economic articles, and then standardizes them to obtain EPU indexes of different provinces (municipalities and autonomous regions) [26].

(3) Control variables

In addition to core variables, we refer to previous relevant studies on the macro level, taking into account various characteristics of the region, including economic development, electricity level, investment level, openness level, consumption level, tax level, and fiscal level [31-33]. So we take a series of control variables into account in the empirical regression (Table 2).

The descriptive statistics of the variables involved in this study are shown in (Table 3). In terms of carbon emission intensity, the average COI of China is 3.153, the maximum value is 11.93, and the minimum value is 0.610, which shows that, on the whole, the carbon dioxide emission intensity varies greatly among provinces in China. In addition, from the perspective of economic policy uncertainty, the average value of EPU is 4.473, the maximum value is 5.763, and the minimum value is 3.095, which shows that, on the whole, the economic policy uncertainty of different provinces in China varies greatly.

From the control variables, the average, maximum and minimum values of the economic development level (LNGDP) are 9.203, 11.22 and 6.372, respectively. It can be seen that the overall difference is small and the average level is high, indicating that the economies of various regions in China are still developing rapidly. The average value, maximum value and minimum value of ELEC are 1441, 5430 and 121.7, respectively, and there are some differences among different regions. The average, maximum and minimum value of investment level (INV) are 8.796, 10.79 and 5.930, respectively, and there are still some differences among regions. The average, maximum and minimum value of import and export level (OPEN) are 0.242, 2.087 and -1.765, respectively, which means that there are still great differences in the openness of provinces. The average, maximum and minimum value of total regional consumption (TCONS) are 8.816, 18.37 and 5.215, respectively, and the average, maximum and minimum value of RPI are 102, 107.6 and 97.86, respectively. It can be seen that there are some differences in consumption levels among provinces. The mean value, maximum value and minimum value of fiscal decentralization (PFD) are 0.515, 0.955 and 0.148 respectively, which means that there are great differences in fiscal levels among regions. In general, there is no abnormal value affecting the results.

It can also be seen from the scatter chart that from 2004 to 2017, the uncertainty of economic policies in most provinces is on the rise (Fig. 1).

Model Structure

The spatial econometric model has a remarkable effect on the econometric analysis. We apply it to the impact of economic policy uncertainty on carbon emission intensity, and we can better analyze the neighborhood effect and mutual influence of its role. Our research framework of this paper in the materials and methods section is as follows (Fig 2).

First, we construct a standard econometric regression model without considering the spatial effect as a comparison. Its econometric regression model is as shown below:

Fig. 1.

3D scatter diagram of economic policy uncertainty and carbon emission intensity in China.

Fig. 2.

Research framework.

Where the subscript i is the province with i = 1, 2, …, 30; t is the year with t = 1, 2, …, 16, which are the years from 2004 to 2017; EPU_i,t represents economic policy uncertainty in the t year of i province, and Controls is the combination of control variables; λ_i, ν_t and ε_i,t are individual effects, time effects, and random error terms.

Then, the spatial Durbin model (SDM) is represented by the construction formula (4). Among them, EPU_i,t represents economic policy uncertainty in the t year of i province, and Controls is the combination of control variables; ρ is the spatial autoregressive coefficient; λ_i, ν_t and ε_i,t are individual effects, time effects, and random error terms; W represents the spatial weight matrix. In this paper, the spatial adjacency matrix is selected to describe the spatial association between regions.

As shown in formula (5), W is a weight matrix expressed by adjacent binary. If the space is adjacent, it takes the value of 1. Otherwise, it takes the value of 0.

Benchmark Regression Result

This preliminary regression without considering the spatial effects is presented in Table 4. Columns 1-4 show the results of least square regression and fixed effect regression before and after adding control variables. It can be seen that the coefficient of economic policy uncertainty (EPU), the core explanatory variable, is significantly positive, indicating that economic policy uncertainty will significantly increase regional carbon emission intensity. The role of economic policy uncertainty in promoting carbon emissions has been initially confirmed, and it is valuable to continue to use the spatial measurement for further research.

Spatial Autocorrelation of Carbon Emission Intensity

The Moran’I index of carbon emission intensity space from 2004 to 2017 is shown (Table 5). The global Moran’I carbon emission intensity index is all significantly positive, with a significance level of 1%, indicating that carbon emission intensity will be affected by neighboring areas. Areas with intensity are adjacent to those with high intensity in space. It can also be seen from Moran’s scatter diagram that intensity presents a spatial distribution pattern of high-low-low aggregation (Fig. 3). From the value of the Moran’I index, it is kept between 0.122 and 0.245, which indicates that the positive spatial correlation of carbon emission intensity is relatively stable.

Spatial Econometric Model Selection and Regression Results

It can be seen that the P value rejects the original hypothesis, indicating that the spatial Durbin model (SDM) cannot be simplified into the spatial lag model (SAR). The spatial Durbin model (SDM) cannot be simplified into the spatial error model (SEM), so the spatial Durbin model (SDM) is selected. Furthermore, the p-value of the Hausman result of the spatial lag model is less than 0.1, which rejects the random effect. Therefore, we select the spatial lag model with a fixed product.

As the coefficients estimated by the SDM model in Table 7 shows, most variables pass the significance test under the geographical proximity weight matrix. λ is the spatial autoregressive coefficient rho, and the statistical results of β coefficient are in the Main and Wx items in the table. After adding control variables, the spatial autoregressive coefficient is 0.400, which is significant at 1% and positive, indicating that the explained variable COI has a positive spatial spillover effect on itself. According to the statistical β value in Main, the coefficients of EPU are both significant, and the Wx term can better explain the spatial conduction effect than the coefficient of Main. Under Wx, the coefficients of EPU are both significantly positive.

The core explanatory variable, economic policy uncertainty (EPU), has a positive impact on carbon emission intensity of 1% after adding the control variable, indicating that economic policy uncertainty can significantly improve carbon dioxide emission intensity. When the EPU rises, the government will pay more attention to stabilizing economic development and overcoming the adverse effects brought by interest rate and exchange rate fluctuations, which will relax the intensity of environmental supervision. Enterprises tend to pay environmental fines and reduce investment in environmental research and development. The positive impact of EPU will lead to the relaxation of government regulation on the environment, which will ultimately have a positive impact on carbon emissions.

In order to further analyze the impact of economic policy uncertainty on carbon emission intensity, the direct effect, indirect effect and total effect under the spatial dobbin model are analyzed. Direct effect represents the degree of influence of variable X in the region on the interpreted variable Y in the region, total effect represents the degree of influence of variable change of one unit in all regions on the interpreted variable Y in the region, indirect effect equals to total effect minus direct effect, meaning the degree of influence of variable X change of one unit in the surrounding region on the interpreted variable Y in the region. The results show that EPU is significantly positive in direct effect, indirect effect and total effect. It indicates that each unit increase of EPU in this region will lead to an increase of 0.131 units of the interpreted variable COI in this region in direct effect and 0.078 units of the interpreted variable COI in this region in indirect effect. In the total effect, an increase of one unit of EPU in all regions can cause an increase of 0.209 units of the interpreted variable COI in this region. Once again, it is proved that the uncertainty of economic policy has a positive impact on carbon emission intensity, and its impact has spatial spillover effect.

Fig. 3.

Moran’s I scatter plots of carbon emission intensity in some years.

Regional Heterogeneity Test

Due to China’s vast territory, the development of the eastern, northeastern, central, and western regions is not coordinated, and there are significant differences in economy, culture, science, and technology. The analysis results are not representative of the situation in each region. To analyze the impact of economic policy uncertainty on carbon emission intensity in different regions, this part uses the above methods to study different regions of China. The 30 provinces except Hong Kong, Macao, Taiwan, and Tibet, are divided into the eastern, central, western, and northeastern regions.

From the horizontal regression results, the influence coefficient of economic policy uncertainty on carbon dioxide emission intensity in the western region is 0.264, which is significant at the level of 5%, and the coefficient in the northeast is also significantly positive, while the coefficients in the eastern and central regions are not significant. It can be seen that the uncertainty of economic policy plays a great role in improving the carbon emission intensity in the western region, because on the one hand, with the uncertainty of economic policy, the environmental maintenance cost that society needs to bear will be transformed into the environmental protection cost that enterprises bear themselves, which makes enterprises reduce their expenditure on carbon emissions within the prescribed limits, and the relaxation of supervision also makes the carbon emission intensity increase; The eastern region of China has developed economy, a good economic environment and relatively stable regional development, and the impact of economic policy uncertainty is not great. The central region’s industrial structure system and economic development level are slightly lower than those in the eastern region, but it still has its own adjustment ability, so the impact of economic policy uncertainty on carbon emission intensity is not significant. Northeast China, as an old industrial area, has a large heavy industry. There are many state-owned enterprises in the heavy industry base in Northeast China, which are large in scale and shoulder great social responsibilities. It is greatly affected by the uncertainty of economic policies, and it is easy to increase regional carbon emissions under the condition of relaxed supervision.

Robustness Test

For the robustness of the results, we then change the spatial matrix and introduce the geographic distance weight matrix. From the results (Table 10), we can see that the coefficient of EPU is significantly positive before and after adding control variables, that is, the increase of economic policy uncertainty will increase the intensity of carbon dioxide emissions. Consistent with the above, after adding the control variable, the spatial autoregressive coefficient is 1.771, which is significant at the level of 1% and is positive, indicating that the explained variable COI has a positive spatial spillover effect on itself.

Next, we change the spatial econometric regression method and use spatial lag model (SAR) and spatial error model (SEM) to verify the relationship between economic policy uncertainty and carbon emission intensity again. From Table 10 (3) and (4), the coefficients of EPU are all significantly positive, that is, the increase of economic policy uncertainty will increase carbon dioxide emission intensity.

For the robustness of the results, we then changed the explained variable, taking the regional per capita carbon dioxide emissions (PERCO2) as the explained variable. From the results (Table 11), we can see that the coefficient of EPU is significantly positive before and after adding the control variable, that is, the increase of economic policy uncertainty will enhance the intensity of carbon dioxide emissions. Consistent with the previous article, after adding the control variable, the spatial autoregressive coefficient is 0.586, which is significant at the level of 1%.

Since the introduction of the environmental protection law in 2015 may have an impact on our results, we shortened the sample life and chose the sample from 2004 to 2014 for regression based on the spatial Dobbin model. From Table 11 (3) and (4), the coefficients of EPU are all significantly positive, that is, the increase of economic policy uncertainty will increase the intensity of carbon dioxide emission.