Policy uncertainty, economic activity, and carbon emissions: a nonlinear autoregressive distributed lag approach

Over the last few years, economic uncertainty has become a global concern. Not only has its impact on economic activities, but there are pieces of evidence that show uncertainty can be the reason for CO2 emissions. It is also expected that the economic policy uncertainty may decrease or delay economic production, which may lead to a reduction in carbon emissions. Furthermore, uncertainty may decrease friendly environment policies and budgets, which cause an increase in carbon emissions. Thus, there may be an asymmetric relationship between economic uncertainty and the amount of CO2 emissions. This study investigates the effects of economic policy uncertainty and economic activity on carbon emission applying a nonlinear autoregressive distributed lag (NARDL) cointegration approach in Iran between 1971 and 2018. Findings show that both policy uncertainty and economic growth contribute to CO2 emissions. The negative and positive shocks of GDP and uncertainty index on CO2 emissions in both the short run and long run are significant. Based on the results, there is an asymmetric effect of economic production on CO2 emissions in Iran. The results of analyzing asymmetric effects of economic uncertainty show a symmetric relationship between uncertainty index and CO2 emissions, in a way that a shock in the uncertainty index lowers carbon emission. To sum up, since uncertainty may affect the analysis of carbon emissions incorrectly, some environmental policies such as allocating a budget for R&D on clean energy and environmental taxes must be implemented.


Introduction
Climate change, global warming, and environmental problems are the most critical issues in the recent decade (Anser et al. 2021a, b). These factors can easily lead to a rise in aerosols and air pollution (Anser et al. 2021a, b). Air pollution can also have noticeable side effects on the national economy (Chang et al. 2020). Since climate change and environmental degradation can result in several human diseases (Shahpari et al. 2021), they attracted the attention of international agencies to mitigate greenhouse gas (GHG). The effective strategy to overcome worldwide global warming is to reduce CO 2 emissions (Chaudhry and Shafiullah 2021;Kompas et al. 2018). Carbon dioxide (CO 2 ) as a primary greenhouse gas consists of around 80% of GHG emissions, which is emitted through human activities (IPCC 2013). Climate change is an important challenge that may threaten to attaining sustainable development through economic and environmental aspects. Therefore, to achieve sustainable development, a decrease in CO 2 emission is a critical fact. The process of how GHG emissions can impact the environment and the importance of CO 2 emissions are illustrated in Fig. 1.
Moreover, climate change can affect uncertainty, and uncertainty can have effects on climate change (Chaudhry et al. 2020). Based on IMF reports, economic policy uncertainty (EPU) is one of the main factors resulting in the GDP plunge over the past few decades. Therefore, concerns about economic policy uncertainty (EPU) are another global issue (Anser et al. 2021a, b). Many studies show that uncertainty has adverse effects on GDP or economic growth (Barrero et al. 2017;Sahinoz and Erdogan 2018;Ghosh 2019;Altig et al. 2020;Alam and Istiak 2020). However, the critical point about economic uncertainty is that besides its economic consequences, it also has some environmental impacts implicitly (Atsu and Adams 2021;Syed and Bouri 2021;Anser et al. 2021a, b).
It is expected that policy uncertainty increases the cost of production, while decreasing investment. In fact, uncertainty affects production decisions. In the other words, economic uncertainty is a helpful sign to forecast a recession because as the uncertainty increases, firms decrease or delay their consumption and investment (Ercolani and Natoli 2020;Adedoyin and Zakari 2020). Therefore, it can be concluded that economic uncertainty will affect the production plans, which in turn will result in changing CO 2 emissions.
Since uncertainty restricts the budgets available to firms, firms will be forced to use outdated production lines to cover the production costs. Using such old production lines will lead to applying unfriendly environmental technologies in the production process. In addition, the budget allocated to R&D for clean energies and innovation methods may reduce in uncertain circumstances. This could also result in a reduction in using renewable energy resources. These changes undoubtedly lead to an increase in the amount of CO 2 emissions. Adams et al. (2020) stated that economic policy uncertainty might increase CO 2 emissions, especially in resource-rich countries, because policy uncertainty may limit innovations to reduce energy consumption and carbon emissions. In resource-rich countries, the price of fossil fuel is affordable enough to prompt producers to consume more fossil fuels, especially during economic uncertainty situations.
On the other hand, policy uncertainty may decrease or delay economic production, which in turn will cause a reduction in carbon emissions (Chen et al. 2021).
Based on the above description of how uncertainty may have a relationship with the amount of CO 2 emissions, it can be concluded that economic uncertainty can either lead to a decrease or even an increase in CO 2 emissions. In other words, there can be an asymmetric relationship between economic uncertainty and the amount of CO 2 emissions. As the EPU increases, there is the possibility of reducing CO 2 emission, since economic uncertainty can decrease the production level. Therefore, the energy consumption gets lower, which can lead to less environmental degradation. On the other hand, EPU may cause producers and consumers to use cheap and dirty fossil fuels to decrease their costs. Therefore, as Yu et al. (2020) express, EPU may increase CO 2 emissions.
However, in developing countries like Iran, which use old technologies, implementing new technologies to substitute the old ones needs a noticeable investment. Moreover, as economic policy uncertainty increases, investment risk gets higher. Therefore, investors are unlikely to invest in green technologies with better environmental impacts in unstable conditions. Firms and producers will increase fossil fuel energy consumption with lower costs since fossil fuel prices are low in oil-rich countries, such as Iran (Yu et al. 2020). In this view, EPU will increase carbon emissions. Thus, as is illustrated in Fig. 2, the relationship between EPU and carbon emission is ambiguous.
According to the above explanations, the relationship between economic uncertainty and carbon emissions should be analyzed to guide the policies associated with environmental degradation. In other words, studying policy uncertainty is essential to evaluate environmental effects and to provide policymakers with more robust information for reducing CO 2 emissions.
Iran is a developing country and one of the top 10 countries of CO 2 emitters Anser et al. 2021a, b). In the 2015 United Nations Climate Change Conference in Paris, Iran made an international commitment in order to reduce CO 2 emissions. According to this commitment, Iran has to reduce 8-12% of the CO 2 emissions from its level in 2005 by 2030 as its long-run development plan (Hosseini et al. 2019;Ashena et al. 2020). In Fig. 3, the Iran CO 2 emissions rate of growth from 2001 to 2019 has been illustrated. From Fig. 3, it is evident that during this time interval, the rate of growth Several kinds of economic fluctuations have shrunk the size of the economy and, at the same time, decreased the economic stability dramatically. These features, along with environmental problems, have made Iran a suitable case study for investigating EPU impacts. This study contributes to the literature into two ways. First, to the best of our knowledge, this is the first study that considers the impacts of uncertainty on CO 2 emissions in Iran. Second, a nonlinear ARDL approach was operated, yielding reliable results as it encounters asymmetric effects.
Furthermore, this essay tries to add new insights to the existing studies by exploring an index that incorporates the uncertainty condition of economics, which affects carbon emissions via economic activity and political changes. Hence, the World Uncertainty Index (WUI) was used for more reliable estimations, which seems to be an appropriate index with the desirable features. Specifically, the current research applies the asymmetric approach of the ARDL model to determine the effects of uncertainty index and GDP on CO 2 emissions.
The rest of this article is organized as follows. In the "Literature review" section, the background of study and study area has been expanded. The "Methodology and model specification" section explains the methodology. Data and results are provided in the "Data and estimation results" section, while this article is concluded in the "Conclusion and policy implications" section.

Previous studies
This section provides a brief explanation of the previous studies on the carbon emission determinants. Economic growth has been considered one of the initial economic 10% determinants of CO 2 emissions in many prior studies over the past decades (Apergis and Payne 2010;Nejat et al. 2015;Xu et al. 2018). Environmental Kuznets curve (EKC) shows the relationship between income and environmental degradation (Wang et al. 2020;Anser et al. 2021a, b). The validity of the EKC hypothesis indicates that economic growth can produce solutions to environmental problems in the long run. Moreover, other factors such as education and development indices can affect the EKC hypothesis and contribute to environmental improvements (Shafiullah et al. 2021a, b). To sum up, most empirical assessment of the nexus between carbon emissions and GDP shows mixed outcomes.
Energy consumption is another determinant widely used in different studies, either in renewable energy consumption or nonrenewable energy consumption (Shafiullah et al. 2021a, b;Sohail et al. 2021;Zakari et al. 2021;Anser et al. 2021a, b). In general, studies showed that an increase in renewable energy consumption could lead to better environmental situations. Yet, increasing fossil consumption can increase environmental degradation ).
In addition, some researchers figured out that research and development is one of the prime determinants. They showed that R&D spending in energy to explore better methods plays a crucial role in reducing GHG emissions (Li and Jiang 2020; Guzowska et al. 2021). Finally, some studies investigated the role of EPU on the reduction of investment which can be an indirect determinant of CO 2 emissions (Jurado et al. 2015;Ercolani and Natoli 2020). Table 1 summarizes studies about the relationship between CO 2 emissions, economic growth, and EPU.
Based on the findings of empirical studies, the nexus of CO 2 emissions and EPU is not clear. For instance, some studies confirm EPU has adverse effects on the environment (Wang et al. 2022;Atsu and Adams 2021;Yu et al. 2020). On the other hand, some studies find that EPU can decrease CO 2 emissions (Lui and Zhang 2022; Anser et al. 2021a, b). Moreover, some essays found different effects of EPU and carbon emissions in the longrun and short-run (Reza Syed and Bouri 2021; Adedoyin and Zakari 2020; Adams et al. 2020).
In order to be able to broaden the horizons of this field, in this study, our contribution is to investigate asymmetric impacts of economic policy uncertainty, GDP, and energy intensity on CO 2 emissions. To fulfill this goal, we considered the world uncertainty index (WUI) as an index to measure both economic and political kinds of uncertainty. Moreover, since environmental issues should mainly be investigated in the long run, using a NARDL enables us to study the nexus in the short run and long run.

Theoretical background
Studying the factors affecting carbon emissions is essential for policymakers and scientists. Environmental degradation's leading social and economic drivers include population growth, economic expansion, and technological changes. The STIRPAT model (STochastic Impacts by Regression on Population, Affluence, and Technology) is an extended stochastic computation of IPAT, which is applicable in sociological modeling theory for drivers of environmental impacts. (Rosa and Dietz 1998;Shi 2003;York, et al. 2003). Economic expansion is related to environmental quality through ecological Kuznets curve (EKC) hypothesis, which shows an inverted U shape (Grossman and Krueger 1995;Dinda 2004).
Energy is a factor of production and is necessary during economic growth; however, more energy consumption during economic activities leads to higher energy intensity and environmental pollution. Energy intensity per unit of GDP can be used as a measure of technology (Wang et al. 2011).
Given that many countries experience economic uncertainty at different periods, the role of economic policy uncertainty should be considered as a factor affecting pollution through energy intensity. Economic policy uncertainty affects the business environment and the decision-making of economic agents. Therefore, CO 2 emissions are related to microeconomic agents' production and investment decisions (Jiang et al. 2019).
In recent years, economic uncertainty has become one of the issues of special importance. Uncertainty shocks affect the demand and supply sides of the economy through different transmission channels. However, there is not a single definition for economic uncertainty, according to the literature. In other words, there is no agreed-upon unique definition for the concept of economic uncertainty (Al-Thaqeb and Algharabali 2019). However, Jin and Wu (2021) explained EPU as the uncertainty associated with signs in monetary and fiscal policies and government regulations that influence how people and firms have their economic activities. In general, economic uncertainty can be categorized into different groups: 1. Uncertainty contributes to market volatility such as regulatory or monetary policies 2. Unexpected changes contribute to the economic ecosystems (Al-Thaqeb and Algharabali 2019) The other important point is that many factors can be the reason for uncertainty. Moreover, some of these factors have both short-run and long-run effects. Therefore, to study the effects of economic uncertainty, it is essential to consider the time horizon. There are some indices to measure uncertainty; for instance, the Chicago Board Options Exchange has been using the volatility index (VIX) for many years as an accepted proxy for firm uncertainties in the equity market (Al-Thaqeb and Algharabali 2019). However, VIX works best for mature markets and is inappropriate for all countries. Methods of text-mining and word counting are usual ways to measure economic uncertainty: keyword-based methods. Some keywords identify the index regarding these methods, such as "uncertain" and "uncertainty." Thus, the uncertainty index is measured by news containing these keywords (Baker et al. 2016). Economic policy uncertainty (EPU) is an uncertainty index calculated based on text-mining in newspaper articles of leading newspapers. The world uncertainty index (WUI) is another measure of uncertainty, which uses a single source for all countries to compare the level of uncertainty across countries. This index captures uncertainty related to economic and political events.
In order to elaborate the relationship between EPU and CO 2 emissions, which was illustrated in Fig. 2, applying a category seems helpful. EPU impacts on carbon emission can be categorized into different groups: i) Consumption effect: EPU declines energy consumption and mitigates carbon emissions. Moreover, high EPU can also plunge the consumption of pollution-intensive products. This will also lead to carbon emission mitigation. ii) Investment effect: high EPU diverts policymakers' attention from the environment to economic stabilizations (Jiang et al. 2019). Furthermore, EPU can have an adverse impact on the investment in renewable energy, innovations, and R&D. This investment effect results in carbon emission (Wang et al. 2020).
To sum up, EPU can simultaneously have negative and positive effects on emissions (Anser et al. 2021a, b).

Methodology and model specification
Scientific approaches such as Dickey and Fuller (1979) and Perron (1990) unit root tests and nonlinear with asymmetric cointegration version of autoregressive distributed lag (ARDL) are applied in this study. Since ARDL is a strong approach for short-run and long-run analysis, it has become a popular and widely used approach, especially for time series analysis (for example, for Iran's economy, Shahpari and Davoudi (2014) and Shahpari et al. (2020) used ARDL).
Based on Shin et al. (2014) and Hatemi-j (2012), the nonlinear and asymmetric cointegration tests are applied in investigating the cointegration and long-run relationship between the dependent variable (CO 2 ) and the explanatory variables (GDP, energy intensity, and economic uncertainty). The nonlinear ARDL is considered in this study because of its advantage over other approaches, such as the vector error correction model (VECM). The nonlinear ARDL approach of cointegration does not need a particular order of integration for cointegration analysis, and a mixed order of integration I(0) and I(1) could be applied.
Before representing the NARDL model, asymmetric longrun regression should be considered to obtain the partial sum processes of positive and negative changes of the independent variable as follows: y t and x t are the scalar variables, and the partial sum of x + t and x − t can be obtained by: Focusing on developing a fully dynamic model, the ARDL approach can be extended (Pesaran and Shin 1998;Pesaran et al. 2001a, b) to model relationships including long-and short-run asymmetries. So, the following nonlinear ARDL(p; q) model may be considered: where β + j and β − j are the asymmetric distributed lag parameters, and ε t is the error correction form.
Carbon emissions may increase by economic policy uncertainty as it mitigates investment in clean energy. So, policymakers tend to control uncertainty to decrease the effects on carbon emissions (Jiang et al. 2019). Danish Ulucak and Khan (2020) also concluded that EPU increases energy consumption and CO 2 emissions in the short and long run. This positive effect of EPU on CO 2 emissions is obtained in both developed and developing countries (Wang et al. 2020;Adams et al. 2020;Anser et al. 2021a, b).
On the other hand, EPU may result in a decrease of production, energy consumption, and CO 2 emissions in some cases (Adedoyin and Zakari 2020;Syed and Bouri 2021). Based on the mentioned background, two hypotheses are proposed: (H1), the EPU affects CO 2 emissions asymmetrically and may decrease or increase CO 2 emissions. As well as (H2), the economic growth-pollution relationship may be positive or negative. Moreover, energy intensity has been considered as another driver of CO 2 emissions in investigation of the two hypotheses. Based on the research variables, the generalized form of the study model can be represented as follows: (1) 4) LCO 2t = β 0 + β 1 LGDP t + β 3 WUI t + β 2 LEI t + ε t where LCO 2 , LGDP, WUI, and LEI represent the natural logarithm of carbon emission (million-ton carbon equivalent), the natural logarithm of GDP (million $ US), economic uncertainty index, and the natural logarithm of energy intensity and ε is the error correction term.
Then, the nonlinear bound test approach is applied to investigate the cointegration relationship. This bound test is developed by Shin et al. (2014) as an extended version of Pesaran et al. (2001a, b). Decomposing selected independent variables can extend the linear version of the unrestricted error correction model (UECM) into positive and negative shocks (Shin et al. 2014). The ARDL (p, q) is converted to the NARDL (p, q) by considering the positive and negative components as follows: So that p and q are the optimal number of lags, φ j are the coefficients of the lags of dependent variable, and β ij are the coefficients of the lags of independent variables. According to the following relations, the selected independent variables are decomposed into positive and negative components: where LGDP + t , WUI + t are partial sum processes of positive changes and LGDP − t , WUI − t are the partial sum processes negative changes.
The model of NARDL(p,q) with the asymmetric error correction is presented as follows: where α, γ represent the short-run and long-run effects of variables. The short-run and long-run analyses investigate the effect of independent variables shock on CO 2 emissions and assess the adjustment speed.
The bound test is done on all the lagged levels of the independent variables. The null hypothesis, no cointegration against the existing cointegration, is investigated by F-statistics. The null hypothesis is rejected where estimated F-statistics are more significant than the upper bound and vice versa (Pesaran et al. 2001a, b). If the LGDP values of the F-statistics lie between the upper and lower bounds, no precise decision can be made. Based on the results of the cointegration, the null hypotheses of symmetric coefficients in the long run or short run can be tested using the Wald statistic following an asymptotic χ 2 distribution.

Data
In this study, CO 2 emissions is considered as the dependent variable, and the influence of uncertainty index and GDP is investigated considering some control variables such as energy intensity. Likewise, although there are several uncertainty indices, this study uses the world uncertainty index. It should be noted that the research variables are transformed into logarithmic forms. This study uses yearly data for Iran in the 1971-2018 period. Data are obtained from a database of World Development Indicators (WDI). The uncertainty index is obtained from a database of policy uncertainty of WUI based on Ahir et al. (2018). Figure 4 shows the time trend of the carbon emissions, GDP, and economic uncertainty during the research period. As it is shown, the natural logarithms of GDP and CO 2 have been increasing, while economic uncertainty fluctuates during the research period. Meanwhile, WUI has been experiencing a positive upward trend since 2011. Approaches such as Dickey and Fuller (1979) and Perron (1990) methods were applied in this study for the test of a unit root. The unit root test results are presented in Table 2, in which both tests indicate that variables including LCO 2 and LGDP have a unit root in the level form and are stationary in the first difference form, while WUI and LEI variables are stationary in the level form. As stated before, the NARDL approach could be used for variables with mixed order of integration I(0) and I(1).
Considering that economic uncertainty may affect production, and thus have an impact on carbon emissions, an endogeneity test is performed to show that all regressors are exogenous. Testing endogeneity based on the Durbin-Wu-Hausman test shows that LGDP is exogenous, as the statistic cannot reject the null hypothesis (χ 2 = 0/82(prob : 0.36)). All other independent variables are also exogenous. Table 3 shows the results of the asymmetric cointegration test. Based on the bounds test approach, the long-run cointegration is confirmed, as F-statistic is greater than the critical value of the upper bound.

Estimation results
These results established a long-run relationship among the variables. The estimate of the long-run coefficients is reported in Table 4. The optimal lag length is selected based on the Akaike information criterion (AIC). Furthermore, the error correction form is estimated to distinguish the short-run effects of the descriptive variables from their long-run effects ( Table 5).
The long-run and short-run NARDL results show a significant relationship between CO 2 emissions and descriptive variables. According to Table 4, positive and negative . 5 1975 1980 1985 1990 1995 2000 2005  LGDP shocks in the partial sum of LGDP increases CO 2 emissions, and the relationship is not symmetric. This conclusion is in line with theoretical foundations and expected results. According to previous studies (Lotfalipour et al. 2010;Ghorashi and Alavi Rad 2017;Solaymani 2020), there is a positive relationship between increasing production and CO 2 emissions in Iran. Furthermore, the effect of adverse shocks of GDP on carbon emissions in the long run and short run is also positive. Although positive and negative relationship in the context of EKC is confirmed in most developed countries (Roberts and Grimes 1997), it is not approved by empirical evidence in developing countries (Harbaugh et al. 2002). Based on the results of this study, LGDP has the most significant effect on CO 2 emission. So that a 1% positive shock to LGDP causes a 1.1% increase, and a 1% negative shock causes a 0.71% increase in CO 2 emissions. In justifying this result, it can be mentioned that due to the low energy price in Iran, energy efficiency is low and is not used optimally. The results of positive and negative shocks of WUI on CO 2 emissions show a negative relationship. In other words, a positive or negative shock in WUI will decrease CO 2 emissions. This implies that higher levels of economic policy uncertainties adversely affect CO 2 emissions in Iran. This finding is consistent with the results of Syed and Bouri (2021) that conclude EPU reduces CO 2 emissions in the long run. Wang et al. (2020) described the consumption effect of EPU on CO 2 emissions reduction; therefore uncertainty downturns the energy consumption and consumption of pollution-intensive goods. As Jiang et al. (2019) stated, EPU and CO 2 emissions can be related through two channels: direct policy adjustment effect and indirect economic demand effect. Direct policy adjustment channel implies that policymakers' focus shifts from environmental policies to economic stability ones, which result in rising of CO 2 emissions, while indirect economic demand channel explains that uncertainty changes the decision-making environment, which in turn affects CO 2 emissions. In contrast, Pirgaip and Dinçergök (2020) and Danish Ulucak and Khan (2020) show the opposite result for developed countries, so that uncertainty leads to an increase in carbon.
On the other hand, a positive relationship is found in the short run. This result is similar to Syed and Bouri (2021) and Adedoyin and Zakari (2020) express that EPU increases CO 2 emissions. Therefore, it can be interpreted that EPU may raise energy consumption and CO 2 emissions in the short run. Economic uncertainty causes supply and demand shocks, possible price changes, and changing expectations that affect production and consumption decisions and change carbon emissions (Jiang et al. 2019;Chen et al. 2021). Therefore, policymakers must decide whether to reduce economic uncertainty or reduce carbon emissions. Mohmmed et al. (2019) explain that energy intensity as well as income are the most important driving factors of CO 2 emissions in top ten emitter countries. Moreover, LEI, which is regarded as a control variable in the model, shows a positive relationship with CO 2 emissions. So that increasing energy intensity will raise CO 2 emissions.
Finally, the asymmetric impact, in the long run, is examined by the Wald test. The results presented in Table 6 show the significance of asymmetry in the long-run parameter of LGDP. Moreover, the null hypothesis of a symmetric longrun relationship for WUI is not rejected and suggests symmetric effects of WUI on carbon emission.
Although the cointegration bounds test of NARDL shows the existence of long-and short-run relationships between dependent and independent variables, it does not determine the direction of causality. Therefore, as the robustness check of the results, directional causal relationships are investigated for both symmetric and asymmetric effects of independent variables on CO 2 emissions. First, the linear causality relationship is performed through the non-Granger causality test, following Toda and Yamamoto (1995). Then, an asymmetric causal relationship is investigated from positive and negative shocks of explanatory variables following Hatemi-J (2012). Following the Toda-Yamamoto (1995) test and considering positive and negative shocks, the nonlinear causality test is performed. Then, as it is proposed by Hatemi-J (2012), a random walk process for integrated variables can be considered as follows: where Y 0 and X 0 are initial values of variables and μ it is the error term.
Given error terms, positive shocks and negative shocks c a n b e d e n o t e d b y + ji = max ji , 0 a n d − ji = min ji , 0 , for j = 1, 2, i = 1, 2, … ., t , respectively. Then, the above random walk can be rewritten as follows: For each variable, cumulative terms can be used to calculate the positive and negative shocks as follow: Granger and Yoon (2002) proposed that the relationship between variables can differ from negative and positive shocks. It is indicated by Hatemi-J (2012) that asymmetric causality of variables can be explored by positive and negative shocks, using the vector autoregressive (VAR(p)) model. The optimal lag length of the model can be selected based on the criteria suggested by Hatemi-J (2003) as a function of the total number of observations and related statistics of error terms. Finally, the Wald test can be applied for testing the null hypothesis of non-causality.
The results for the Granger causality test are presented in Table 7. In addition, given negative and positive shocks of variables, the long-term relationship between the positive and negative shocks of variables can be tested (Hatemi-J 2012).
The estimation results show one-way causality running from GDP, WUI, and EI, to CO 2 emissions. Moreover, the results highlight the presence of asymmetric causal relations between CO 2 emissions and economic uncertainty as well as economic growth in Iran. This affirms the magnitude of changes in policies, regulations, and plans for the environment-growth relationship. On the other hand, a lack of causality from energy intensity and CO 2 emissions to EPU can raise the fact that energy and environmental policies do not lead to economic uncertainty.
Finally, positive and negative changes of the explanatory variables on the asymmetric responses of dependent variable can be traced by the positive and negative dynamic multipliers (m). A 1% change of WUI + , WUI − , GDP + , and GDP − can be calculated as follows: while i → ∞, multipliers of m + , m − tend to the long-run coefficients.
(11)   Figure 5 shows NARDL multipliers, representing dynamic adjustments of variables to a new equilibrium after the positive and negative shocks. The figure include asymmetry adjustments of CO 2 to negative and positive shocks (black lines), as well as asymmetric pattern implying the difference between negative and positive shocks Table 8 reports the model residual diagnostic tests, including autocorrelation, heteroscedasticity, Ramsey RESET, and normality tests. The results of these residual diagnostic tests indicate that the null hypothesis of autocorrelation, heteroscedasticity, model stability, and normality cannot be rejected. Furthermore, Fig. 6 represents CUSUM and CUSUMSQ tests of the model. The figures show stability in the model, as the significant lines lie between the critical lines.

Conclusion and policy implications
In recent decades, economic uncertainty has increased rapidly. Economic uncertainty can delay the investment plans directly and, therefore, decreases production. Results of previous research confirmed that uncertainty is a reason for sharp recessions (Bloom 2017; Kaveh-Yazdy and Zarifzadeh 2021). In addition to these economic consequences, economic uncertainty can also have environmental impacts (Anser et al. 2021a, b). For countries such as Iran, which use fossil fuel as their energy resource in the production process, this decreases fossil fuel energy consumption. CO 2 emission declines as a result of the decrease in the amount of fossil fuel usage.
This study analyzed policy uncertainty and economic growth on carbon dioxide emissions in one of the most carbon dioxide emitter countries, Iran, for the first time. Applying NARDL approach, the cointegration tests revealed a long-run relationship for all variables. The findings of the study suggest that policy uncertainty and economic growth contribute to CO 2 emissions. The negative and positive shocks of LGDP and WUI on CO 2 emissions in both the short run and long run are statistically significant. In summary, the estimated results indicate an asymmetric effect of economic production on carbon emission in Iran. These findings are consistent with those of Lotfalipour et al. (2010), Ghorashi and Alavi Rad (2017), and Solaymani (2020) as economic growth is associated with more energy consumption and CO 2 emissions. Therefore, to decouple economic growth from CO 2 emissions, it is required to invest in R&D applying low-carbon technologies and energy efficiency targets. Based on Shafiullah et al. (2021a, b), education achievements have positive effects on environmental quality in the long run.
The results of analyzing asymmetric effects of WUI show a symmetric relationship between WUI and CO 2 emissions, in a manner that a shock in policy uncertainty lowers the carbon emission. Meanwhile, Kaveh-Yazdy and Zarifzadeh (2021) also found that EPU can be the reason for higher unemployment rates which is also a signal of a decrease in production levels. Ercolani and Natoli (2020) also used  economic uncertainty to forecast recession periods, another definition for the decrease in production levels.
The implication of the relationship between policy uncertainty and CO 2 emission is that uncertainty makes it challenging to decide on economic activity, and it cannot be ignored in the GDP-CO 2 emissions relationship. In other words, policymakers should consider that economic growth will result in more CO 2 , while uncertainty will result in less CO 2 . Therefore, although policy uncertainty will reduce CO 2 emissions in Iran, it should be noted that this result may cause locking into the existing fossil fuel-based economy structure. Therefore, it is reasonable for the countries to promote economic policy that encourages innovation and stimulates capital investment in energy efficiency equipment or appliances. R&D budget should be raised to find new methods of clean production. Allocating subsidies for clean production methods is another practical policy to reduce fossil fuel consumption and encourage firms to apply modern technologies. Finally, political uproar and unrest should be adequately addressed to reduce its effect on emissions.
Policymakers should be aware of country dynamics in developing economic policies regarding energy consumption and CO 2 emissions. Due to this evidence that EPU could affect consumption, it can reduce CO 2 emissions. Moreover, decreases in investment in innovations and clean energy due to high EPU could increase CO 2 emissions. Therefore, the effects of EPU on CO 2 emissions should be considered in energy and environmental policies to avoid misspecification of relationships between variables, which could provide more accurate information for decision-makers. Future studies can investigate the role of EPU in CO 2 emissions by considering alternative types of energy sources.
Acknowledgements This article is taken from a research project entitled "Investigation of the asymmetric effect of economic policy uncertainty on pollution emissions using the NARDL approach" approved by the Research Council of Bozorgmehr University of Qaenat, notification number 39215. The authors would like to thank Bozorgmehr University of Qaenat for the financial support of this research.
Author contribution MA analyzed data and developed the model regarding the relationship between economic uncertainty and CO2 emissions. GS prepared the historical literature and introduction part of the manuscript. Both authors read and approved the final manuscript Funding Bozorgmehr University of Qaenat supported the funding of this research, notification number 39215.
Data Availability All data generated or analyzed during this study are included in the reference list of this article. Data will be available upon request.

Consent for publication Not applicable
Competing interests The authors declare no competing interests.