The Effect of Economic Policy Uncertainty on Green Technology Innovation: Evidence from China’s Enterprises

Abstract

In some cases, enterprise fears uncertainty more than the policy itself, and this fear can impede innovation. However, this study finds that uncertainty is the source of enterprise’s innovation. We took Chinese A-share listed companies from 2010 to 2018 as the research sample and studied the impact of economic policy uncertainty (EPU) on green technology innovation (GTI). Based on China’s EPU index and green patent-application data of Chinese-listed enterprises, this study adopted a panel fixed regression model and found that EPU has a slightly promoting effect on GTI of Chinese-listed enterprises. This effect is also affected by enterprise ownership and industry characteristics. Among them, EPU has a stronger promoting effect on GTI activities of state-owned enterprises and high-tech enterprises than common enterprises. In particular, if EPU increases, the GTI of high-polluting enterprises is not as great as the incentive effect of ordinary enterprises. The reasons may be that state-owned enterprises have more implicit capital guarantees, high-tech enterprises have higher innovation motivation, and high-polluting enterprises have stronger dependence on traditional production equipment. These results can provide a reference for the debate on “uncertainty.” The conclusions of this paper contain unique policy implications.

1. Introduction

The Chinese government has actively implemented economic reform measures to promote the transformation of the economic development pattern, while the government’s frequent policy adjustments have also increased the uncertainty of economic policy [1]. This uncertainty not only changes the macroenvironment, it also changes enterprises’ operating microenvironment and impacts the decisions they make. Economic policy uncertainty (EPU) exerts enormous implications. Now, academic circles have many indicators to measure the uncertainty of economic policy. For example, Julio and Yook (2012) [2] explored the uncertainty index in election years; Jurado, Ludvigson and Ng (2015) [3] developed a new macroeconomic uncertainty index using econometric techniques to measure uncertainty. These indicators have some limitations, such as being hard to replicate in other countries, or identifying only certain types of uncertainty, rather than measuring the uncertainty of a country’s entire economy. Baker et al. (2016) [4] constructed an index reflecting EPU, which is a relatively comprehensive and reliable indicator, by extracting words related to EPU from newspapers to construct uncertainty. Figure 1 shows the trend of this EPU index in China.

We can find that the index fluctuated greatly, reaching peaks in 2008 and 2012. The reason is that in 2008, the global financial crisis broke out, financial systems of various countries were severely damaged, and China could not stay out of the problem. To stabilize the economic environment, the Chinese government issued the “stimulus plan of 4 trillion yuan”, “interest cut”, and other relevant policies. In 2012, the leadership of the Chinese government changed and new economic policies were introduced. For example, the new model of “Internet +” development was proposed and the construction of “the Belt and Road” was initiated. Moreover, the overall economic uncertainty in China showed a zigzagging upward trend. Especially after 2015, the EPU index increased sharply. The government has implemented “supply-side structural reform”, and proposed to “three go, one drop, one supplement” (capacity reduction, destocking, deleveraging, cost reduction, improving underdeveloped areas), and “reform tax policy.” The introduction of economic policies is conducive to the stable transformation and development of China’s economy, but it also leads to the high uncertainty of economic policies to a certain extent.

As the main part of economic activities, enterprises’ behavior is affected by national macro- and micropolicies [5]. Moreover, enterprises’ production, operation mode, research and development (R&D) investment, and innovation behavior will also be affected by economic policy uncertainty [6], such as investment in enterprises [7,8,9] and cash-holding strategy [10,11,12,13]. In particular, the innovation behavior of enterprises has a strong policy orientation and will be affected by the policy environment. As we all know, the policy itself has a nonnegligible effect on technological innovation [14,15,16,17], and in some cases, policy uncertainty has an even greater impact on enterprise innovation than policy itself. We can understand that if policy uncertainty rises, enterprises’ investment environment will change greatly, which will affect R&D investment and technological innovation [18].

There are many kinds of technological innovation activities, and green technology innovation (GTI) is one of them. It has the characteristics of environmental protection and high efficiency, which can guarantee economic efficiency and give consideration to environmental protection at the same time [19]. Different from other technological innovation, it has double externalities, and that makes enterprises lack endogenous motivation, needing the support of external forces, such as government policies. Due to the dual externalities of green innovative technologies, the uncertainty of economic policies has a greater impact on their capital flows and R&D intentions, resulting in a relatively higher uncertainty of investment returns. Therefore, compared with other technologies, green innovative technologies are more vulnerable to policy uncertainty than other technologies.

However, few studies have given specific answers to what the interrelationship between GTI and EPU is, and whether it is promoting or inhibiting, but this question is necessary to study, because it can be used as a reference for the government to make relevant policies. Based on the EPU index of China from 2010 to 2018, as well as the GTI patent data and macroeconomic climate index of Chinese-listed companies, this paper conducts an in-depth study on the relationship between EPU and enterprises’ GTI.

2. Literature Review

Compared with general investment, research and development activities are more exploitative and exploratory. Among many investments, enterprise research and development belong to special investment. Bloom (2007) [20] pointed out that higher uncertainty will delay the investment of enterprises because they would wait for policy stability before investing. Similarly, economic policy uncertainty would delay investment by enterprises [21,22,23]. The results of Wang et al. (2017) [24] and Khan et al. (2020) [25] showed that economic policy uncertainty is not conducive to enterprise R&D investment, and this negative effect is more significant in enterprises with strong competitiveness, fewer government subsidies, and strong political connections. In fact, scholars generally believe that economic policy uncertainty has a negative impact on enterprises’ investment. However, Al-Thaqeb and Algharabali (2019) [26] believe that the investment impact of economic policy uncertainty on enterprises is difficult to predict, because it may depend on other factors or the market position of enterprises. It means that economic policy uncertainty may have a positive effect on R&D activities [27]. For instance, by taking A-share listed enterprises in Shanghai and Shenzhen Stock Exchange as samples from 2009 to 2017, Gu et al. (2020) [28] found that economic policy uncertainty would promote enterprises’ R&D investment. Jiang and Liu (2020) [29] used the data of 1163 A-share listed companies from 2008 to 2016, adopted linear Bayesian regression to study the influence of economic policy uncertainty on R&D expenditure, and got the same result as Gu et al. (2020) [28]. However, Chen et al. (2019) [30] believed that for enterprises with low systemic risk and large investment opportunities, there is an obvious nonlinear relationship between economic policy uncertainty and long-term investment.

How much R&D is invested is closely related to how much innovation is produced. Therefore, this paper also believes that the influence mechanism of EPU on enterprise investment will be transferred to the technological innovation of enterprises. Based on this conjecture, many scholars have investigated the impact of EPU on enterprise technological innovation. Gu et al. (2018) [31] found that EPU had a positive impact on patent-application volume of listed enterprises. In addition, the relationship between EPU and innovation behavior is affected by financial constraints, enterprise ownership, industry characteristics, and other factors. Xiong et al. (2021) [32] found that EPU is generally positively correlated with enterprise innovation. In addition, it has a stronger positive impact on state-owned enterprises, low-cashflow enterprises. Some studies also believe that the increase of policy uncertainty is not conducive to the innovation activities of enterprises [33,34,35], and He et al. (2020) [36] results shows that the relationship of policy uncertainty and innovation activities is not linear: when EPU is low, it has a significant positive impact; when EPU is high, it significantly inhibits enterprise innovation. Just like the relationship between R&D input and EPU, the relationship between EPU and enterprise innovation is still controversial.

With China’s economy stepping into a new stage, GTI as an important approach to solve environmental problems and economic development, more and more scholars have begun to study it. At present, the academic circle mainly studies GTI from the perspectives of system policy [22,37,38], market [39,40,41,42], and enterprise performance [43,44]. Among them, the policy of GTI has been widely discussed [45]. For example, Shen et al. (2020) [46] selected data of China’s 30 provinces from 1997 to 2014 and used multiple regression analysis to study the relationship between various environmental policy tools and regional green development. Results showed that different policy tools have different influences on innovation of green products, green process, and green technology.

Not only does policy have an impact on enterprises GTI but also the uncertainty of economic policy. Moreover, GTI is different from ordinary innovation activities. Due to its “double externality”, “high investment and low return”, and other characteristics, it may be more vulnerable to economic policy uncertainty. Li et al. (2021) [47] used the fixed-effect panel method and the provincial panel data of 30 administrative regions in China from 2000 to 2017, and found that economic policy uncertainty has a negative impact on green technology innovation. Xu and Yang (2021) [48] also explored the impact of economic policy uncertainty on green technology innovation by using data on economic conditions and urban conditions of 269 Chinese cities from 2005 to 2016. Different from Li et al. (2021) [47], Xu and Yang’s (2021) [48] results support the view that uncertainty can promote green technology innovation within a certain threshold range, while when uncertainty exceeds the inflection point, it has an inhibiting effect on green technology innovation. Those papers have different opinions about the relationship between EPU and GTI. For the question raised at the beginning, we still have no answer about the relationship of EPU to GTI. Moreover, the existing literature has studied the relationship between EPU and GTI from province or city perspectives, i.e., the macro view, but not the micro view.

Based on the above analysis, this study takes A-share listed enterprises as research objects from 2010 to 2018, based on China’s EPU index, and empirically tests the impact of EPU on GTI of enterprises. The contributions of this paper are as follows. (1) Existing literature mainly studies the relationship between EPU and enterprise technological innovation, but there is little research on the relationship between GTI and EPU from the perspective of microenterprises. Therefore, this study focuses on GTI and further explores the impact of policy uncertainty on GTI of enterprises. (2) Enterprises with different levels of pollution have different demands for GTI. Enterprises with higher pollution are more likely to be exposed to environmental pressure from the outside world, so they have a more urgent need for GTI. Therefore, this paper distinguishes the listed enterprises as high-polluting enterprises or non-high-polluting enterprises. Specifically, in combination with the unique characteristics of GTI, this paper also studies the heterogeneity impact of EPU on GTI for high-polluting enterprises and ordinary enterprises.

The following sections are arranged as follows. The third part is the theoretical analysis and research hypothesis, the fourth part is data and models, the fifth part is the empirical results, Section 6 is the robustness test, and the last section is the conclusion and policy implications.

3. Theoretical Analysis and Hypothesis

According to Knight (1921) [49], uncertainty is the only source of corporate profits. Assuming changes can be predicted, enterprises cannot make profits in the long run. Therefore, the existence of uncertainty encourages entrepreneurs to increase innovation input and improve the level of technological innovation. Although Bernanke (1983) [50] and Bloom (2009) [51] pointed out that, according to the real option theory, if the investment is irreversibility and investment-income information updates over time, the uncertainty would increase the value of waiting options, which would hinder the current investment of enterprises. However, according to Kulatilaka and Perotti (1998) [52], the initial investment under uncertainty is the acquisition of future growth options. When the strategic advantages are obvious, the increase in uncertainty will encourage the investment in growth options, especially in technological innovation, because it will enable the enterprise to obtain higher value of growth options. What is more, the most symbolic output of R&D is patents. By selling intellectual property rights, enterprises can partially recover the R&D investment ultimately reflected in patents, and partially offset the irreversibility of R&D investment. In this way, investing in green technology does not satisfy the premise of irreversibility mentioned by Bernanke (1983) [50]. Therefore, when the uncertainty of economic increases, enterprises tend to pay more attention to the potential benefits brought by the uncertainty and increase innovation investment [53,54]. For the GTI, it has a higher growth option value, and EPU may have a more obvious promoting effect on it. In recent years, people’s requirements for the environment are getting higher and higher, which also puts forward new requirements for the GTI. It is necessary to increase GTI and achieve high efficiency and low pollution production. Therefore, when EPU increases, enterprises will increase their R&D efforts for GTI.

According to the above material analysis, Hypothesis 1 is proposed:

Hypothesis 1.

EPU can promote the innovation of green technology.

Although Hypothesis 1 is proposed in the paper, it is rather general. As we all know, different industries and individual characteristics of enterprises are different, that is, the impact of EPU on GTI is different, which means there is a selection effect. State-owned enterprise and non-state-owned enterprise have different attributes. The former has more political resources and financial support, while the latter faces greater market competition. In a competitive environment [55,56], the impact on innovation is greater for enterprises that are financially constrained and those that rely on external financing [57]. Therefore, when faced with the uncertainty of economic policy, enterprises with two different attributes have different responses to GTI activities. Therefore, this study proposes Hypothesis 2:

Hypothesis 2.

The impact of EPU on green GTI activities of state-owned enterprises and non-state-owned enterprises is different, that is, EPU has a selection effect among enterprises with different ownership.

High-tech enterprises refer to those enterprises that, in the Key High-tech Fields supported by the State (this is a document issued by China in 2015, which identifies the high-tech sector—http://www.scio.gov.cn/32344/32345/32347/33665/xgzc33671/Document/1452811/1452811.htm, accessed on 13 September 2021), continuously carry out research and development and transformation of technological achievements, form core independent intellectual property rights, and carry out business activities on this basis. Compared with traditional enterprises, high-tech enterprises have the characteristics of high technical talent density, high growth and high profitability [41], and have stronger innovation strength. They tend to be more innovative in the face of uncertainty, considering the external investment risk and the dynamic continuity of innovation input, enterprises are more inclined to take innovation investment, so as to obtain first-mover advantage and competitive advantage [32]. Therefore, this paper speculates that EPU has a greater impact on GTI of high-tech enterprises. Hypothesis 3 is proposed:

Hypothesis 3.

The impact of EPU on the GTI level of enterprises in non-high-tech industries is less than that in high-tech industries.

High-polluting enterprises include steel, cement and other industries, generally traditional types of enterprises, with high emissions and high pollution characteristics [37]. Compared with non-high-polluting enterprises, high-pollution enterprises are environmentally sensitive enterprises, its environmental activities have received much attention [58]. In this case, high-polluting enterprises invest more in GTI. However, in the case of increasing EPU, compared with ordinary enterprises, high-polluting enterprises may have a more moderate response when they are stimulated by EPU. In the case of unstable macropolicies, traditional high-polluting enterprises may pay more attention to production efficiency and ignore GTI in order to maintain their operating efficiency. Hypothesis 4 is proposed:

Hypothesis 4.

The GTI level of high-polluting enterprises is less affected by EPU.

4. Data and Models

4.1. Sample Selection and Data Resource

This study selects China’s A-share listed enterprises from 2010 to 2018 as the research object. There are more than 4000 enterprises in total. Because of missing data, 910 listed enterprises were studied after excluding enterprises with missing data for the whole year: ST enterprises and listed enterprises without green patents. The green patent is from the patent system database of the State Intellectual Property Office of China, and the other data are from the CSMAR database.

4.2. Variables’ Description

4.2.1. Green Technology Innovation

A variety of indicators have been adapted to measure GTI, such as comprehensive indices, but it may lead to serious endogenous problems. If one component of the composite index is related to the error term, the parameter estimator may be biased and inconsistent [48]. Other studies use R&D investment to represent the innovation level, but it only represents the local government’s emphasis on innovation, not the innovation level. And due to data limitations, R&D spending cannot effectively measure GTI [59,60]. Given all this and reference previous study [10,43], we used green technology patent to measure GTI, which can directly reflect the GTI’s level, and it can be accurately measured.

First, we searched for patent information applied by A-share listed enterprises in China State Intellectual Property Office patent system database (http://pss-system.cnipa.gov.cn/sipopublicsearch/portal/uiIndex.shtml, accessed on 13 September 2021), then according to the list of green technologies issued by the World Intellectual Property Organization (WIPO) in 2010, combined with the international patent classification number and the found patents of listed enterprises to match, finally, the number of green patents of each listed enterprise is calculated every year. At the same time, this paper also distinguishes between green invention patents and green utility patents, taking them as the core index to measure the GTI of enterprises. Considering that application time of green patent is long and it has been put into use in the application process, which has an impact on the performance of enterprises. In addition, there is political intervention in whether green patents are authorized, which is unstable and uncertain, so the explained variable is the green patent applications rather than the green patent grants.

4.2.2. Economic Policy Uncertainty

Considering the accuracy and availability of the data, this study used the economic policy uncertainty index (EPU). There are two types of EPU index: one is constructed by Davis et al. (2019) [61] based on newspapers in mainland China: the Guangming Daily and Renmin Daily. Another one is developed by Baker et al. (2016) [4] based on the South China Morning Post. Those indices depict the EPU of China through textual and semantic analysis of the contents of newspapers. Even through there are differences between the two EPU indices, for example, Davis et al. (2019) [61] chose “economic”, “policy”, and “uncertainty” the three term sets, while Baker et al. (2016) [4] selected “China, Chinese”, “economic, economy”, and “uncertain, uncertainty” the three term sets, construction procedure is similar. First, they constructed the term sets which can be describe the economic policy uncertainty; then, according the term sets, filtered the papers including those terms at least once and count the paper monthly. In the last, standardize each newspaper’s monthly series of scaled frequency counts to have a unit standardization. In this paper, we collected the original data from the website http://www.policyuncertainty.com, accessed on 13 September 2021. Both types of raw data are monthly data, which we convert into annual data using average arithmetic. Davis constructed an EPU index used in fundamental regression and heterogeneity analysis regression, Baker constructed EPU index used in the robustness test.

4.2.3. Characteristic Variables

Considered the EPU has heterogeneity for different types enterprises, this study set three characteristic variables to illustrate this heterogeneity. The first characteristic is the enterprise attributes, denoted by $ownership$. The listed enterprises were divided into state-owned enterprises and non-state-owned enterprises according to the classification of ownership nature of listed enterprises in China Stock Market and Accounting Research (CSMAR) database. Second characteristic variable: high-tech enterprises ($hightech$). According to the qualification standards of listed enterprises in CSMAR database, the enterprises were classified into two types: high-tech enterprises and non-high-tech enterprises. The third characteristic is high-polluting enterprise ($highpollu$). The heavily polluting industries cover 16 types of industries, including thermal power, steel, cement, electrolytic aluminum and other industries (see Guidelines on Environmental Information Disclosure of Listed Enterprises published by the Ministry of Environmental Protection). In this paper, according to what industry the enterprise belongs to, to judge whether it is high-polluting enterprise. This information can be extracted from CSMAR database.

4.2.4. Control Variables

According to previous literature [6,18], we chose flowing variables to control in our model: profitability earning power ($ROA$); enterprise scale ($size$); asset-liability ratio ($lev$); cashflow ratio ($cashflow$); tangible assets ratio ($capital$); Tobin Q ($tobinq$); financing constraints ($SA$); enterprise age ($age$); and macroeconomic climate index ($macroecoindex$). The specific variable names and definitions are shown in

Table 1. Variable definition and description.

Category	Variable Name	Symbol	Index Calculation
Dependent variables	Green patents	grnpat	The number of green patents, add 1 and take the log
	Green invention patents	grninv	The number of green invention patents, add 1 and take the log
	Green utility model patents	grnuty	The number of green utility model patents, add 1 and take the log
Independent variable	Economic policy uncertain index	EPU	Monthly data is converted into annual data by arithmetic average method
Control variables	Return on Assets	ROA	Net profit/(Ending balance of total assets + Beginning balance of total assets)/2
	Enterprise scale	size	Total assets
	Asset-liability ratio	leva	Liabilities/Total Assets
	Cash flow ratio	cashflow	Net cash flow from operations/Total assets
	Tangible assets ratio	capital	Fixed assets/Total assets
	Tobin Q value	tobinq	Market value/Total assets
	Financing constraints	SA	(Cash holdings + Inventories + Accounts receivable accounts payable)/Total assets
	Enterprise age	age	The period from the establishment of the enterprise to 2018
	Macro-economic climate index	Macroecoindex	Macro-economic climate index
Characteristic variables	Enterprise attributes	owership	State-owned enterprise, Ownership = 1; if not, Ownership = 0
	High-tech enterprises	hightech	High-tech enterprises, hightech = 1; If not, hightech = 0.
	High-polluting enterprise	highpollu	High-polluting enterprise, highpollu = 1; if not, highpollu = 0

.

4.3. Model

Based on the above theories, the basic model is individual fixed model, as follows:

$$grnpa{t}_t={\beta }_0+{\beta }_{1}EP{U}_{t-1}+{\beta }_{i}contro{l}_{i,t-1}+{\epsilon }_{it}$$

(1)

In this model, $i$ denotes the ith listed enterprise, $t$ means the tth, $grnpat$ is the green patents, in order to smooth the data, this paper takes logarithm of $grnpat$. And because green patent applications number of some enterprises is 0 in some years, the logarithm cannot be directly taken. In reference to the practice of other scholars, we add 1 before taking the logarithm. Dependent variables include green patents, green invention patents, green utility model patents, they’ re denoted by $grnpat$, $grninv$, $grnuty$. Due to the delay effect of innovation output, research and development takes a certain time, so the patent is usually delayed for one or two years. Refer to Gu et al. (2018) [31], all explanatory variables and control variables (except age variable) were delayed by one period. Explanatory variable $EP{U}_{t-1}$ represents the EPU in the $t-1$ year. $contro{l}_{i,t-1}$ represents all the control variables, including profitability, enterprise size, asset-liability ratio, cash flow ratio, tangible asset ratio, Tobin Q value, financing constraints, enterprise age, macroeconomic climate index. ${\eta }_i$ represents individual fixed effects, ${\epsilon }_{it}$ represents residual term.

To verify Hypotheses 2 and 3, the following econometric models are set:

$$grnpa{t}_t={\beta }_0+{\beta }_{1}EP{U}_{t-1}+{\beta }_2{X}_{i,t}+{\beta }_{3}EP{U}_t\times X_{i,t}+{\beta }_{i}contro{l}_{\mathrm{i},t-1}+{\epsilon }_{\mathrm{it}}$$

(2)

In this model, adjustment variables ($X$) are added to this model, the interaction terms of the moderating variables ($X$) and economic policy uncertainty ($EPU$) are added. The moderating variables are enterprise nature (ownership) and high-tech enterprises ($hightech$) respectively.

If you test Hypothesis 2, $X_{i,t}$ = $ownershi{p}_{i,t}$, where $X_{i,t}$ represents the ownership attribute of enterprise i in the t year. If it is a state-owned enterprise, $Ownershi{p}_{i,t}=1$; If not, $owershi{p}_{i,t}=0$. Similarly, when using this model to test Hypothesis 3, $X_{i,t}$ = $hightec{h}_{i,t}$, where $X_{i,t}$ represents whether enterprise $i$ belongs to the high-tech industry. If it is high-tech enterprise, then $hightec{h}_{i,t}=1$; If not, then $hightec{h}_{i,t}=0$.

In view of the difference between GTI and general technology innovation, the reactions of high-polluting enterprises and ordinary enterprises in the face of EPU deserve further study. Therefore, this paper designs an econometric model different from the verification of Hypothesis 4, as follows:

$$\begin{array}{l}grnpa{t}_t={\beta }_0+{\beta }_{1}EP{U}_t+{\beta }_{2}EP{U}_t\times highpoll{u}_{i,t}\\ +{\beta }_{3}highpoll{u}_{i,t}+{\beta }_{i}contro{l}_{\mathrm{i},t-1}+{\epsilon }_{\mathrm{it}}\end{array}$$

(3)

If the enterprise belongs to the high-polluting industry, $highpoll{u}_{i,t}$ = 1; if it does not belong, $highpoll{u}_{i,t}$ = 0. Unlike model 2, $EPU$ is no one-stage lag processing. Considering that GTI of high-polluting enterprises is more sensitive to the EPU, this paper does not treat $EPU$ with a delay of one period. Figure 2 shows the research steps of this paper.

5. Analysis of Empirical Results

5.1. Basic Statistics

Descriptive statistics of major variables in this paper are shown in

Table 2. Descriptive statistics of major variables.

Variable	Observations	Average	Standard Deviation	Median	Minimum	Maximum
$grnpat$	8190	3.048	19.426	0.000	0.000	781.000
$grninv$	8190	1.312	13.892	0.000	0.000	664.000
$grnuty$	8190	1.735	8.172	0.000	0.000	283.000
$EPU$	8190	140.955	57.976	124.336	92.114	277.813
$ROA$	8190	0.051	0.263	0.043	−6.714	20.788
$size$	8190	10.920	108.079	0.331	0.000	2218.098
$lev$	8190	0.432	0.396	0.422	0.001	29.454
$cashflow$	8190	0.180	0.154	0.131	0.000	1.073
$capital$	8190	0.934	0.078	0.957	0.222	1.227
$tobinq$	8190	2.050	2.845	1.607	0.169	192.705
$SA$	8190	0.367	0.201	0.356	−1.073	2.477
$age$	8190	16.315	5.998	16	1	106
$macroecoindex$	8190	98.874	3.425	98.240	92.550	103.600
$ownership$	8190	0.403	0.491	0.000	0.000	1.000
$hightech$	8190	0.768	0.422	1.000	0.000	1.000
$highpollu$	8190	0.281	0.450	0.000	0.000	1.000

below. The total sample consists of 910 enterprises from 2010 to 2018, among which the annual average number of each enterprise’s green patent applications was 3.048, the annual average number of green invention patents was 1.312, and that of utility model patents was 1.735. Among the research objects, the number of state-owned enterprises and non-state-owned enterprises is relatively balanced; the number of high-tech enterprises is too large; the number of highly polluting enterprises is small. The number of highly polluting enterprises is small. EPU changed greatly, the minimum value was 92.114, the maximum value was 277.813, standard deviation reached 58.

5.2. Analysis of Regression Results

5.2.1. Analysis of Correlation

As shown in

Table 3. Correlation analysis results.

	$\mathit{E}\mathit{P}\mathit{U}$	$\mathit{R}\mathit{O}\mathit{A}$	$\mathit{S}\mathit{i}\mathit{z}\mathit{e}$	$\mathit{L}\mathit{e}\mathit{v}$	$\mathit{C}\mathit{a}\mathit{s}\mathit{h}\mathit{f}\mathit{l}\mathit{o}\mathit{w}$	$\mathit{C}\mathit{a}\mathit{p}\mathit{i}\mathit{t}\mathit{a}\mathit{l}$	$\mathit{T}\mathit{o}\mathit{b}\mathit{i}\mathit{n}\mathit{q}$	$\mathit{S}\mathit{A}$	$\mathit{M}\mathit{a}\mathit{c}\mathit{r}\mathit{o}\mathit{e}\mathit{c}\mathit{o}\mathit{i}\mathit{n}\mathit{d}\mathit{e}\mathit{x}$
$EPU$	1.0000
$ROA$	−0.0259	1.0000
$size$	0.0134	−0.0131	1.0000
$lev$	0.0250	−0.2841	0.1122	1.0000
$cashflow$	−0.1415	0.0959	−0.0761	−0.2936	1.0000
$capital$	−0.0996	0.0208	0.0718	0.0265	0.1593	1.0000
$tobinq$	−0.0862	0.2330	−0.0357	0.0935	0.0904	−0.0066	1.0000
$SA$	−0.1112	0.0742	−0.1462	−0.2842	0.6827	0.1977	0.0505	1.0000
$macroecoindex$	0.2877	0.0287	−0.0025	−0.0085	0.1368	0.0414	−0.0445	0.0849	1.0000

, the correlation index is far less than 0.5, indicating that there is no complete linear correlation. The largest correlation coefficient is EPU and macroeconomic index, but it is also less than 0.3, indicating that the next step of regression can be carried out.

5.2.2. Test Hypothesis 1

Table 4. The relationship between economic policy uncertainty and enterprises’ green patent application.

	$\mathit{G}\mathit{r}\mathit{n}\mathit{p}\mathit{a}\mathit{t}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{v}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{u}\mathit{t}\mathit{y}$
$EP{U}_{t-1}$	0.0038 *** (8.84)	0.0018 *** (5.98)	0.00300 *** (8.18)
$RO{A}_{t-1}$	0.0934 ** (2.25)	0.0425 (1.44)	0.0940 *** (2.63)
$siz{e}_{t-1}$	0.0003 *** (3.19)	0.0003 *** (4.68)	0.0001 (1.01)
$le{v}_{t-1}$	0.1347 *** (4.66)	0.1347 *** (3.06)	0.1202 *** (4.83)
$cashflo{w}_{t-1}$	0.0755 (0.79)	0.0919 (1.35)	0.0404 (0.49)
$capita{l}_{t-1}$	0.56533 *** (3.89)	0.1785 * (1.73)	0.6296 *** (5.04)
$tobin{q}_{t-1}$	−0.0177 *** (−4.73)	−0.0092 *** (−3.43)	−0.0156 *** (−4.83)
$S{A}_{t-1}$	−0.3182 *** (−4.30)	−0.1635 *** (−3.10)	−0.3334 *** (−5.23)
$Age$	−0.0080 *** (−3.94)	−0.0003 (−0.21)	−0.0097 *** (−5.56)
$macroecoinde{x}_{t-1}$	−0.030490 *** (−7.62)	−0.01848 *** (−6.48)	−0.0202 *** (−5.87)
F	19.07	12.64	19.57
R2	0.0256	0.0171	0.0262

Description: The t-statistic is in parentheses. *, ** and *** indicate significant at 10%, 5% and 1% (double-tailed) levels, respectively.

shows the impact of EPU on GTI of enterprises from 2010 to 2018. According to the regression result, no matter be explained variables are $grnpat$ (Green Technology Patents), or $grninv$ (Green Invention Patent), or $grnuty$ (Green New Patents), $EPU$ in 1% of the regression coefficient of level was significantly positive, uncertainty of economic policy to promote the Chinese-listed enterprise green patent application, which verified hypothesis 1. For listed enterprises, the uncertainty of economic policy is both an opportunity and a challenge. No matter how the policy changes, GTI is an unstoppable trend. Therefore, if EPU is improved, enterprises will strengthen the research and development of green patents, so as to improve their long-term competitiveness. However, it is important to note that this positive effect is relatively weak.

From the coefficient of control variables, we found that the stronger the profitability, the stronger the ability of green technology innovation at the 1% significance level. Once enterprise has more profit, and they will put more investment to innovation, and product more patents. For enterprise with larger scale, higher ratio of tangible assets and smaller financing constraints have stronger GTI ability. It is not hard to understand that if enterprises have more money, they will invest more money in green technology research and development, which will produce more green technology patents. He et al. (2020) [36] also believes that the innovation performance of larger enterprises with stronger profitability and cash ratio is higher than that of other listed enterprises. The older an enterprise is, the more mature it is. Even if the EPU is high, the enterprise with high maturity still has strong traditional products and other channels to maintain its development. Therefore, the investment in GTI is relatively weak. On the contrary, when faced with high EPU, young enterprises are more likely to take a leap of faith, and GTI is more motivated.

5.2.3. Test Hypothesis 2

To test whether enterprise nature has an impact on linear correlation between EPU and GTI of enterprises, the interaction term between the dummy variable of enterprise nature and EPU index was introduced. The regression results are shown in the

Table 5. The selection effect of economic policy uncertainty: enterprise ownership.

	$\mathit{G}\mathit{r}\mathit{n}\mathit{p}\mathit{a}\mathit{t}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{v}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{u}\mathit{t}\mathit{y}$
$EP{U}_{t-1}$	0.0034 *** (7.14)	0.0013 *** (3.73)	0.0030 *** (7.23)
$control variable$	Yes	Yes	Yes
$ownership$	0.1113 *** (4.54)	0.0956 *** (5.49)	0.0710 *** (3.36)
$ownership\ast EPU$	0.0010 *** (2.84)	0.0012 *** (5.04)	0.0003 (0.94)
F	19.23	17.05	17.60
R2	0.0308	0.0274	0.0283

Description: The t-statistic is in parentheses. *** indicate significant at 1% (double-tailed) levels.

. When the explained variables are respectively $grnpat$, $grninv$ and $grnuty$, the $ownership\ast EPU$ coefficient is all positive at the significance level of 1%, which fully proves hypothesis 2, that is, compared with private enterprises, state-owned enterprises produce more GTI in the face of EPU. This is consistent with the conclusion of Gu et al. (2018) [31]. State-owned enterprises have access to more bank loans with longer maturity, which can finance their own research and development activities [56], even when EPU is high, they are still able to carry out green technology research. However, due to financing constraints, political resources and other factors, non-state-owned enterprises are faced with greater market competition. In the face of economic policy uncertainties, they are more inclined than state-owned enterprises to maintain the status quo and not make risky investments, such as green technology research and development, in order to obtain short-term benefits and competitiveness, so as to maintain the survival of enterprises.

5.2.4. Test Hypothesis 3

From

Table 6. Choice effect of economic policy uncertainty: whether to belong to high-tech enterprises.

	$\mathit{G}\mathit{r}\mathit{n}\mathit{p}\mathit{a}\mathit{t}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{v}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{u}\mathit{t}\mathit{y}$
$EP{U}_{t-1}$	0.0018 *** (3.22)	0.0003 (0.83)	0.0019 *** (3.89)
$control variable$	Yes	Yes	Yes
$hightech$	0.0253 (0.95)	−0.0428 *** (−2.62)	0.0483 ** (2.11)
$hightech\ast EPU$	0.0018 *** (5.55)	0.0014 *** (5.92)	0.0010 *** (3.67)
F	19.08	13.55	18.35
R2	0.0306	0.0219	0.0294

Description: The t-statistic is in parentheses. ** and *** indicate significant at 5% and 1% (double-tailed) levels, respectively.

, we can know that compared with non-high-tech enterprises, the incentive effect of high-tech enterprises is more obvious when faced with the uncertainty of economic policy, that is, high-tech enterprises have more green technology patent applications, which verifies Hypothesis 3. The conclusions of this paper were consistent with the results of Gu et al. (2018) [31] and Xiong et al. (2021) [32]. It is generally believed that, in general, the product attributes of high-tech industry determine that enterprises must continue to maintain innovation vitality in order not to be eliminated, and the technology leader in the industry will get great returns. High- and new-technology enterprises are often enterprises with strong growth ability, strong innovation ability and high R&D investment. When economic policy uncertainty rises, high-tech enterprises will pay more attention to technological innovation to survive. Therefore, high-tech enterprises are more likely to be motivated by the uncertainty of economic policies, carry out GTI.

5.2.5. Test Hypothesis 4

The regression coefficient of the cross term ($highpollu\ast EPU$) between high-polluting enterprises and EPU is negative at the significance level of 5% (See

Table 7. Selection effect of economic policy uncertainty: whether to be a high polluter or not.

	$\mathit{G}\mathit{r}\mathit{n}\mathit{p}\mathit{a}\mathit{t}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{v}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{u}\mathit{t}\mathit{y}$
$EPU$	0.0031 *** (11.77)	0.0020 *** (10.57)	0.0020 *** (8.95)
$control variable$	Yes	Yes	Yes
$highpollu$	−0.1536 *** (−5.92)	−0.0294 (−1.6)	−0.1675 *** (−7.49)
$highpollu\ast EPU$	−0.0009 ** (−2.37)	−0.0006 ** (−2.27)	−0.0005 (−1.60)
F	26.80	18.27	24.76
R2	0.0424	0.0293	0.0393

Description: The t-statistic is in parentheses. ** and *** indicate significant at 5% and 1% (double-tailed) levels, respectively.

), that is, EPU has less promoting effect on GTI of enterprises in the sample of high-polluting enterprises. The regression results support Hypothesis 4. Cai et al. (2020) [37] believe that the uncertainty of economic policy has a restraining effect on environmental protection investment of high-polluting enterprises. The environmental protection investments of high-polluting enterprises were inhibited, and we further speculated that the number of green technology patent applications related to environmental protection was less than that of ordinary enterprises. It is not hard to see why EPU promotes GTI of enterprises on a whole, but the regression coefficient of $highpollu\ast EPU$ is negative. Under normal circumstances, enterprises with high pollution will pay more attention to GTI than ordinary enterprises due to greater external environmental pressure. When the external environment changes greatly, that is, the uncertainty of economic policy increases, the impact of GTI and R&D of high-polluting enterprises is relatively small. Therefore, the EPU increases, but the promotion effect on GTI is relatively moderate.

6. Robustness

6.1. Re-Selecting EPU

To eliminate the EPU index regression results of above is accidental, by referring to the methods of Zhong et al. (2020) [62], this paper adopted EPU index constructed by Baker et al. (2016) [4]. At the same time, monthly data is converted into annual data. The experiment above is repeated. This study found that to explain variables, regulating, the influence of the important control variables to be explained is consistent with the above conclusion, indicating that the research results of this paper are robust (See the

Table A1. Robustness test of baseline regression: the explained variable is replaced by Baker’s EPU.

$\mathit{Y}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{p}\mathit{a}\mathit{t}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{v}$	$\mathit{G}\mathit{r}\mathit{n}\mathit{u}\mathit{t}\mathit{y}$
${EPU}_{t-1}^{\prime }$	0.0013 *** (10.79)	0.0009 *** (10.06)	0.0009 *** (8.10)
$RO{A}_{t-1}$	0.1071 *** (2.58)	0.530 * (1.80)	0.1017 *** (2.84)
$siz{e}_{t-1}$	0.0004 *** (3.37)	0.0004 *** (5.02)	0.0001 (1.15)
$le{v}_{t-1}$	0.1444 *** (4.99)	0.0721 *** (3.51)	0.1242 *** (4.98)
$cashflo{w}_{t-1}$	0.1058 (0.79)	01289 * (1.89)	0.0463 (0.58)
$capita{l}_{t-1}$	0.6569 *** (4.51)	0.2632 ** (2.54)	0.6695 *** (5.33)
$tobin{q}_{t-1}$	−0.0192 *** (−5.14)	−0.0101 *** (−3.8082)	−0.0165 *** (−5.13)
$S{A}_{t-1}$	−0.3114 *** (−4.22)	−0.1591 *** (−3.03)	−0.3289 *** (−5.16)
$Age$	−0.0099 *** (−4.84)	−0.0024 * (−1.65)	−0.0103 *** (−5.81)
$macroecoinde{x}_{t-1}$	−0.0217 *** (−6.65)	−0.0170 *** (−7.30)	−0.0110 *** (−3.91)
F	22.95	19.25	19.44
R2	0.0306	0.0258	0.0261

Description: The t-statistic is in parentheses. *, ** and *** indicate significant at 10%, 5% and 1% (double-tailed) levels, respectively.

and

Table A2. Robustness test of heterogeneity: the explained variable is replaced by Baker’s EPU index.

$\mathit{Y}=\mathit{G}\mathit{r}\mathit{n}\mathit{p}\mathit{a}\mathit{t}$
	$\mathit{X}=\mathit{O}\mathit{w}\mathit{e}\mathit{r}\mathit{s}\mathit{h}\mathit{i}\mathit{p}$	$\mathit{X}=\mathit{H}\mathit{i}\mathit{g}\mathit{h}\mathit{t}\mathit{e}\mathit{c}\mathit{h}$	$\mathit{X}=\mathit{H}\mathit{i}\mathit{g}\mathit{h}\mathit{p}\mathit{o}\mathit{l}\mathit{l}\mathit{u}$
${EPU}^{\prime }$	0.0013 *** (9.34)	0.0019 *** (5.65)	0.0010 *** (9.87)
$control variable$	Yes	Yes	Yes
$X$	0.1246 *** (5.08)	0.0346 (1.3)	−0.1687 *** (−6.51)
$X\ast EPU$	0.0006 * (1.75)	0.0013 *** (3.86)	−0.0005 ** (−1.2)
F	22.30	20.90	23.27
R2	0.0355	0.0334	0.037

Description: The t-statistic is in parentheses. *, ** and *** indicate significant at 10%, 5% and 1% (double-tailed) levels, respectively.

.).

In addition, the PRS group developed an economic risk index, which was used to replace the explanatory variable EPU index for regression. The regression results are shown in Appendix C. The results show that, when the explanatory variable is the economic risk index, the impact of economic risk on green technology innovation of enterprises is promoting at the significance level of 1%, and the results of heterogeneity test are completely consistent with the above. This shows that our regression is robust.

6.2. Re-Selecting Green Patent Indicator

The explained variable for text is the number of green patents plus one and then take the logarithm. In order to avoid the distortion of data processing, the number of green patent applications is directly used as the explained variable to perform regression on Model 1. Regression results of explanatory variables remain significant and sign unchanged, indicating that EPU has a promoting effect on GTI research of enterprises (See the Table A2 in Appendix B).

6.3. A Discussion of Endogeneity

Firstly, there is almost no endogeneity problem between macro data and enterprise micro data. Therefore, there is no reverse causality between green patent innovation and EPU. Secondly, all explanatory variables were delayed by a phase (except EPU in Model 4), which further avoided the reverse causality problem. Lastly, this paper selects several control variables that may affect the GTI of enterprises, including profitability, debt paying ability, Operation ability and financing constraint, which effectively avoids the endogenous problem.

In conclusion, the regression results of this paper are robust.

7. Conclusions and Policy Implications

Based on the relevant financial indicators, green patent applications and economic policy uncertainty (EPU) index of A-share listed enterprises from 2010 to 2018, this research discusses the impact of EPU on green technology innovation (GTI) of enterprises. The results show that EPU has a slight positive effect on GTI of listed enterprises. EPU promotes the number of GTI patent applications by listed enterprises. In addition, this study also finds that EPU has a selection effect on GTI. Compared with private enterprises, GTI of state-owned enterprises is more promoted by EPU. Compared with non-high-tech enterprises, high-tech enterprises’ GTI is more motivated by EPU. Compared with non-high-polluting enterprises, high-polluting enterprises are less positively affected by EPU.

Different from the conclusions of most literature-EPU inhibits enterprises’ investment and innovation. Results show that EPU promotes enterprises’ GTI, even though not so much. And there is also a selection effect due to enterprises’ heterogeneity. Our results are consistent with the “promote hypothesis” and corroborate previous studies [31,63], and this paper provides evidence on the real impact of EPU shocks on the enterprise. At the same time, this study expands the research content of economic policy uncertainty. We care about the relationship between a special innovation behavior (GTI) and EPU, which is directly related to the balance between environmental protection and economic development. Such an expansion certainly makes sense. In addition, the conclusions of this paper have unique policy implications.

First, the government and relevant departments can appropriately adjust economic policies according to the local actual situation, which is of certain positive significance to the GTI of enterprises, without worrying about the negative impact of changing economic policies on the GTI of enterprises. Especially for environmental policies, such as environmental tax and emission trading permits, the government can make full use of these policy tools to promote GTI of enterprises, accelerate economic transformation and development, and realize the coordinated development between environmental protection and economic growth.

Second, the government can also make different policy adjustments according to different attributes of enterprises (state-owned or non-state-owned) and categories of enterprises (high-tech or non-high-tech, high-pollution or non-high-pollution). For state-owned enterprises, there can be stricter environmental requirements; For private enterprises, more attention should be paid to stabilizing the market, to fully mobilize the enthusiasm of all parties for GTI.

In addition, due to the difference in the promotion degree of GTI between high-tech and non-high-tech enterprises by EPU, the GTI of non-high-tech enterprises is less promoted by EPU. Therefore, when adjusting economic policies, the government needs to consider the adaptability of traditional enterprises, try to balance the enthusiasm of GTI of high-tech enterprises and non-high-tech enterprises, and introduce effective economic policies to maximize the promotion of GTI of different types of enterprises.

Last, the government may also consider the heterogeneity between high-polluting enterprises and non-high-polluting enterprises, and formulate policies according to the conditions of enterprises. Governments can use the selective effect of EPU. For example, when adjusting relevant policies, the government can optimize China’s industrial structure through appropriate economic policies and create an innovative atmosphere more suitable for the survival and development of low-pollution enterprises. Although there is no need to worry about the negative impact of the uncertainty on enterprises’ GTI, the government should try its best to provide clear policy guidance for high-polluting enterprises and promote their GTI.

This paper provides proof that economic policy uncertainty can promote enterprise innovation, and enriches this research boundary (green technology innovation). Admittedly, there are still some limitations in this paper. First, the measurement of green technology innovation of enterprises is limited to the output of green patents of enterprises, but it ignores other possible measurement methods, such as the application and efficiency of the innovation process. In other words, the impact of future economic policy uncertainty on innovation efficiency and application is the direction to be studied. Second, this paper neglects the impact of environmental regulation on the relationship between economic policy uncertainty and enterprise green innovation. For example, Cai et al. (2020) [37] believes that stricter environmental regulation will improve enterprise green technology innovation, but we do not consider the impact of policy regulation on green technology innovation. Although we use the economic climate index to reflect the impact of the environment on enterprise green technology, how to add environmental regulation into the moderating variable and further study the moderating effect of environmental regulation is the direction of our future efforts.