The Impact of Multi-Type Policies on New Energy Vehicle Innovation Output in China: A Central–Local Perspective

1. Introduction

Technological innovations based on green energy, environmental protection, and low carbon concepts are emerging. In consideration of environmental protection and related regulations, new energy vehicles (NEVs) have begun to attract the attention of the market. Moreover, multilevel governments use a series of policies or policy mixes to promote NEVs to reduce CO2 emissions in China [1]. Policy mix affects the speed and direction of social technology’s transition to sustainability [2]. A policy mix is the combination of different policy tools from various governance levels [3]. Moreover, interactions may occur both within multiple policy types at the same government level and across different government levels [1]. The existence of one policy tool may increase (weaken or not affect) the effectiveness of another tool. However, existing research on policy evaluation focuses primarily on national-level NEV policies and individual NEV policies [4]. In contrast, local government-level studies are relatively scarce [5], and even more so for research on multi-type policy mixes across multilevel governments. Moreover, research on the synergistic effects of multi-type policies across multilevel governments on clean energy technology innovation remains relatively scarce [6]. Therefore, this paper assesses the effectiveness of China’s multilevel government multi-type policies on NEV innovation output, and further explores the presence of synergistic effects among them.

The Chinese government, which exhibits a typical multilevel structure, places great importance on the sustainable development of NEVs and frequently employs policy mixes to promote the advancement of green technologies. In previous research, scholars mainly studied the impact of policies on innovation output from three aspects: the supply side (technology-driven), the demand side, and systematic policy tools. For example, Veugelers (2012) classified innovation policy into supply-side policy, demand-side policy, and environmental-side policy to explore clean innovation [7]. Zhao and Su proposed the existence of strategic policy [8].They believed that strategic policy plays a forward-looking and systematic guiding role in the long-term development of science and technology. Strategic policy is parallel with the research and development (R&D) policy on the supply-side and regulatory instruments on the environmental-side as a policy category [9]. Heidrich et al. also confirmed the existence of strategic policy, and the analyzed the effect of UK’s EV strategic policy on the development of urban EV infrastructure [10]. Therefore, building on these previous studies and the actual application of policy tools in China’s NEV industry, this paper divides policies into central policies and local policies according to different levels of government. It then divides policies into strategic policies, supply-side policies, demand-side policies, and environmental-side policies to study the impact of policy mixes on the innovation output of NEV. Additionally, we explore the linkage effect of central policies and local policies.

Our findings contribute to existing research as follows. First, from the perspective of policy function, this study introduces strategic policies into the policy mix and categorizes policies by type (strategic, supply-side, demand-side, and environment-side) and level (national and local), thereby expanding the theory of policy instrument classification. Second, employing a quantitative approach from the vertical government perspective, this study assess the heterogeneous effects of multi-level and multi-type policies on NEV innovation output, and further decompose the individual contributions of each policy category within the overarching policy mix. Third, this paper further demonstrates that both same-type and multi-type policy coordination between central and local governments exert positive effects on NEV innovation output, thereby providing empirical evidence for evaluating the effectiveness of vertical policy synergy.Last but not least, by providing empirical evidence from China, this study extends the application scope of policy mix research to developing countries.

The rest of this paper is organized as follows. Section 2 outlines the theoretical basis and proposes the research hypotheses. Section 3 introduces the empirical data and method. Section 4 presents the negative binomial regression results, the measurement of policy lag effects, and robustness test results. Section 5 discusses the synergy effects arising from multilevel government policies. Section 6 concludes the study and puts forward policy implications.

2. Theory Development of Hypotheses

A policy mix is defined as a multi-tiered innovation policy combination, comprising both a horizontal configuration of instruments within the same governmental level and a vertical configuration of instruments across different governmental levels [11]. Lanahan and Feldman (2015) studied policy effects of the federal government and the state governments in the United States. They found that the innovation activities of SMEs in the United States are not only influenced by the policies of the national federal government but are also influenced by the local political and economic factors of state governments, as well as by pressures from horizontal and peer governments [12]. Freitas (2020) categorised public financial support in France into three tiers：local government, central government, and the European Union，and found that, despite some overlaps in the innovation activities supported by these policies, the three types of fiscal support tend to favour innovation activities undertaken by ambidextrous and open firms [13].Edler et al. (2013) proposed that there are three effects of the interactions between two policy tools in a policy mix [14]. In the policy mix, one policy tool may increase or decrease the effectiveness of the other, or they may not affect each other. Del Río (2014) divided the interaction types of policy instruments into four types: strong conflict, weak conflict, complete complementation, and coordination [15]. Policymakers should construct a reasonable policy structure, which is reflected by policies being consistent and complementary at different levels of government, as well as having a successful relationship in the time dimension.R&D subsidies at the national and local government levels exhibit complementarity [16]. Only in this way can policy mixes achieve the effect desired by policymakers.

In this study, a policy mix influences innovation activities from two aspects: diversity of policy tools and multilevel policy tools. Specifically, the variety of policy tools is where policy tools can be divided into strategic, supply-side, demand-side, and environmental-side policies. These four types of policies have played leading, driving, pulling, and influencing roles. A multilevel policy is a policy that is divided into central and local government policy. The strategic policy at the central level is designed at the top-level to solve systemic problems in the industry. The supply-side, the demand-side, and the environmental-side policies are the concrete implementation policies. Under the guidance of the central government, local governments implement supporting local-level policy tools on the supply-side, demand-side, and environmental-side.

2.1. National-Level Strategic Policies

Strategic policy refers to the top-level design of the national government and its innovative development strategy, medium-term and long-term development plans, and strategic plans for the country, industry, and region [17]. Through strategic coordination, the national government comprehensively guides the formulation, implementation, and supervision of specific policies by its governments at all levels. The national government uses three specific policy tools: strategy, planning, and organization [9]. Strategic policy runs through the entire process of policy implementation. Strategic planning is one of the NEVs innovation policy tools [1]. Strategic policies play top-level design and strategic coordination roles. The national government supports and guides the technology, market, and environment of NEVs through strategic policies [10,18]. The implementation of an NEV strategy defines the strategic positioning, development goals, and development direction of NEVs at the national level. Therefore, a strategic policy at the national level promotes the innovation output of NEVs. Therefore, Hypothesis 1 is proposed:

Hypotheses 1.

Strategic policies of the national government play a positive role in the innovation output of NEVs.

2.2. Supply-Side Policies

2.2.1. National-Level Supply-Side Policies

Supply-side policies refers to government measures that aim to improve the supply of enterprise innovation elements by providing talent, capital, technology, information, and public services to ultimately promote innovative developments for the economy [19]. Innovation generates negative externalities, which dampen enterprises’ enthusiasm for engaging in innovation activities. The government typically uses measures such as R&D subsidies and public service platforms to reduce the negative externalities of NEVs innovation activities. The effect of R&D subsidy on innovation is the most direct and effective. It effectively reduces the cost of enterprise R&D activities and incentivizes firms to increase investments in R&D activities [20]. Moreover, even if R&D subsidies exhibit a crowding out effect, they still effectively improve the output of enterprise innovations [21].During the initial stage of NEV innovation, the government’s R&D subsidy policy has played a positive role in promoting the technical innovation and technological progress of NEVs [22]. Supply-side policies at the national level have effectively reduced the R&D costs of NEVs through subsidies, tax incentives, and other measures. Moreover, they have improved the enthusiasm of market players to innovate, encouraging the investment of more R&D capitals. Therefore, a supply-side policy at the national level promotes the innovation output of NEVs. Accordingly, Hypothesis 2 presented.

Hypotheses 2.

Supply-side policies of the national government play a positive role in the innovation output of NEVs.

2.2.2. Local Government-Level Supply-Side Policies

Local governments also support the innovative development of NEVs in their region from the supply side. Local governments set up special funds, build engineering laboratories, and perform other functions to help the development of NEVs and support independent innovation. Pilot cities for NEVs issue supply-side policies to promote NEVs. For example, the Shenzhen municipal government disseminated a notice on the issuance of policies regarding the revitalization and development of Shenzhen’s NEV industry. According to this notice, from 2009 onward, a special fund dedicated to the development of new energy industries was to be established for seven consecutive years. The fund was intended to support independent innovation and R&D in NEVs, as well as the construction of R&D centers, engineering laboratories, and key laboratories. Local governments set up supply-side policies to encourage the development of local NEVs. When local and central government supply-side policies are aligned, their positive impact on innovation output persists [1]. Through the preceding analysis, Hypothesis 3 is proposed.

Hypotheses 3.

The supply-side policies of local governments have a positive effect on the innovation output of NEVs. Under national-level policy implementation condition, supply-side policies at the local government level still play a positive role in improving the innovation output of NEVs.

2.3. Demand-Side Policies

2.3.1. National-Level Demand-Side Policies

Demand-side policies refers to government policies that act on producer and consumer demands. On the one hand, the government guides enterprises in developing innovative products or services that are needed by the market. On the other hand, the government guides and stimulates consumers to purchase innovative products or services to reduce the uncertainty of supply and demand in the market for innovative products, ultimately promoting the diffusion of innovative products by influencing market demand [23]. Demand-side policies, such as consumer subsidies, positively influence the commercialization of clean technology patents by new energy vehicle enterprises and accelerate the diffusion of innovative products. However, China’s consumer subsidy policy for NEVs follows a phased reduction mechanism, under which the standard subsidy amount per purchase declines year by year. Furthermore, the policy target has gradually shifted from providing universal subsidies across all vehicle models to prioritizing technologically advanced models [24]. Although the subsidy phase-out policies acts as a guiding mechanism that raises the threshold for technological innovation and thereby promotes innovation in the new energy vehicle industry [25], it simultaneously exerts a negative impact on the sales volume of NEVs [26]. The subsidy phase-out policies does not severely hinder technological progress; rather, it can be well coordinated with or serve as a substitute for non-subsidy demand-side policies, such as the dual-credit policy, thereby guiding technological innovation in the NEV industry. These arguments generate the following hypothesis:

Hypotheses 4.

The demand-side policy of consumer subsidy phase-down at the national level exhibits an inverted U-shaped impact on the innovation output of NEVs.

2.3.2. Local Government-Level Demand-Side Policies

Local governments in pilot cities of China support the application and development of NEVs in their regions through demand-side policies and encourage public institutions and private individuals to buy and use NEVs. All these measures have created a market demand for NEVs and improved the enthusiasm for R&D in the NEV industry. Particularly, when subsidy phase-out policies are aligned between local and central governments, local demand-side policies exhibit a similar inverted U-shaped effect on innovation output.Therefore, Hypothesis 5 proposed:

Hypotheses 5.

The demand-side policies of local governments exert a U-shaped effect on the innovation output of NEVs. Under the implementation condition of national-level policies, the demand-side policy of consumer subsidy phase-down at the local government level still exhibits an inverted U-shaped impact on the innovation output of NEVs.

2.4. Environmental-Side Policies

2.4.1. National-Level Environmental-Side Policies

Environmental-side policies are government measures designed to create a supportive environment for innovative product cultivation and marketization. These measures include infrastructure construction and the establishment of technical standards, implemented through administrative norms and public services [17].These policies have heightened market entities’ focus on R&D for charging piles and related infrastructure, leading to an increase in patent output [22].The national government contributes to the development of NEV charging infrastructure technology innovation activities by using policy tools such as infrastructure construction incentives and electricity prices subsidies.The construction of infrastructure will create a good environment for the use of NEVs. Infrastructure construction for supporting NEVs is one of the critical links in their commercialization and industrialization [27]. Robust industrial supporting facilities and expansive market prospects incentivize market entities to increase R&D investment in NEV products. Overall, environmental-side policies promote innovation in the NEV industry leading to the following hypothesis 6:

Hypotheses 6.

Environmental-side policies of the national government play a positive role in the innovation output of NEVs.

2.4.2. Local-Level Environmental-Side Policies

The number of regional infrastructures, such as charging piles, is an essential factor that affects the sales of NEVs [28]. In China’s NEV pilot cities, local governments have begun constructing charging piles extremely early. Local governments have also introduced environmental-side policies such as infrastructure subsidies, operations management, and land concessions to support the innovative development of NEVs [29]. The implementation of these policies influences the decision-making of market entities regarding R&D in charging technologies and product development for NEVs. Consequently, environmental-side policies provided by multilevel governments positively affect R&D activities and technological innovation in charging infrastructure, ultimately contributing to the improvement of NEV innovation output.Therefore, Hypothesis 7 is presented.

Hypotheses 7.

The environmental-side policies of local governments exert a positive effect on the innovation output of NEVs. Under implementation condition of national-level policies, environmental-side policies at the local government level still play a positive role in improving the innovation output of NEVs.

3. Data and Methods

3.1. Sample Selection and Data Collection

This study tests the hypotheses using panel data (1993–2022) from 20 pilot cities in China, all of which were among the earliest pilot cities launched in 2009 and 2010. Policy data came from the Pkulaw database, the NEV statistical yearbook, and China’s NEV industry development report. Other data were taken the statistical database of China’s economic and social development, the city statistical yearbook of China, and the city statistical bulletin of China.

3.2. Dependent Variable

Patents are important indicators of innovation output .Therefore, we use the number of patent applications of NEVs as the dependent variable. Patent data comes from the website of China National Intellectual Property Administration (http://pss-system.cnipa.gov.cn). The patent keyword search method is used to search for NEV patents Keywords [1] include electric vehicles, hybrid vehicles, fuel cell vehicles, new energy vehicles, lithium-ion batteries, power battery systems, drive motor systems, and vehicle control. The number of applications patents is a counting variable. The patent variable is denoted by the symbol “totalpatent.”

3.3. Independent Variables

This study sets policy variables with dummy variables except for demand-side policies [30]. Strategic, supply-side and environmental-side policies adopt dummy variables (variable = “1” when the policy is implemented and “0” under other conditions). Demand-side policies refer to the setting of the demand subsidy policy variables, with the upper limit of consumption subsidy amount for pure passenger EVs representing the demand-side policies [30]. Given the time lag in policy implementation effects, policy variables are appropriately lagged. The lag periods follow the policy lag effect estimation results in Section 4.3. All demand-side policies, such as car purchase subsidies, occur in the current period, and thus, demand-side policy variables do not lag. The national strategic, supply-side, demand-side, and environmental-side policy variables are denoted by the symbols “strategy, supply_c, demand_c, and env_inf_c,” respectively. Similarly, local supply-side, demand-side, and environmental-side policy variable are denoted by “supply_d, demand_d, and env_inf_d,” respectively.

3.4. Control Variables

In this study, per capita gross domestic product (GDP), per capita income, and the patent ownership of 10,000 individuals in pilot cities are used as control variables. The three control variables are denoted by the symbols “per_GDP, per_income, and per_citypatent,” respectively. Per capita GDP represents the effect of pilot cities’ economic development on the innovation of NEVs [1,31]. Per capita income reflects the consumption of NEVs in pilot cities [1,32]. The patent ownership of 10,000 individuals is used to describe the innovation capacity of pilot cities [33].

3.5. Model

Based on the foregoing analysis, models are developed to estimate the effects of multi-type policies on NEV innovation output at the central government level, the local government level, and the multilevel government level. These models are presented in Equations (1), (2), and (3), respectively.

$${totalpatent}_{it}={{\alpha }_0+\alpha }_1{strategy}_{it-2}+{\alpha }_{2}suppl{y\_c}_{it-2}+{\alpha }_{3}deman{d\_c}_{it}+{\alpha }_4{demand\_c}_{it}^2+{\alpha }_5{env\_\mathrm{i}\mathrm{n}\mathrm{f}\_c}_{it-3}+{\alpha }_{6}lnper\_{gdp}_{it}+{\alpha }_{7}ln{per\_income}_{it}+{\alpha }_8{lnper\_citypatent}_{it}+{\epsilon }_{it}$$

(1)

$$total{patent}_{it}={{{\beta }_0+\beta }_{1}supply\_d}_{it}+{\beta }_2{demand\_d}_{it}+{\beta }_3{demand\_d}_{it}^2+{{\beta }_{4}env\_inf\_d}_{it-1}+{\beta }_{5}lnper\_{gdp}_{it}+{\beta }_{6}ln{per\_income}_{it}+{\beta }_7{lnper\_citypatent}_{it}+{\epsilon }_{it}$$

(2)

$$total{patent}_{it}={\gamma }_0+{{\gamma }_{1}suppl{y}_d}_{it}+{\gamma }_2{deman{d}_d}_{it}+{\gamma }_3{demand\_d}_{it}^2+{{\gamma }_{4}env\_inf\_d}_{it-1}+{\gamma }_5{strategy}_{it-2}+{\gamma }_{6}suppl{y\_c}_{it-2}+{\gamma }_{7}deman{d\_c}_{it}+{\gamma }_8{demand\_c}_{it}^2+{\gamma }_9{env\_inf\_c}_{it-3}+{\gamma }_{10}lnper\_{gdp}_{it}+{\gamma }_{11}ln{per\_income}_{it}+{\gamma }_{12}{lnper\_citypatent}_{it}+{\epsilon }_{it}$$

(3)

Equation (1) presents the effects of four types of policy tools (strategic, supply-side, demand-side, and environmental-side) at the central government level on the innovation output of new energy vehicles (NEVs). Equation (2) presents the effects of three types of policy tools (supply-side, demand-side, and environmental-side) at the local government level on NEV innovation output. Equation (3) presents the effects of three types of local government policy tools (supply-side, demand-side, and environmental-side) on NEV innovation output, under the influence of multi-type policies at the central government level.${\alpha }_i,{\beta }_i$ and ${\gamma }_i$are policy effect parameters to be estimated, ${\alpha }_0{,\beta }_0$ and ${\gamma }_0$ are constants, and ${\epsilon }_{it}$ is the random error term.

4. Results

4.1. Descriptive Statistical Analysis

Table 1 presents the descriptive statistics of all variables. Descriptive statistics are performed for the 11 variables selected for the model. The correlations among independent variables are minimal, and no multicollinearity is detected. The average patents of 20 cities are 189.98, and the standard deviation is 327.44. The variance-to-mean ratio is 564. The negative binomial regression model is suitable for the excessive dispersion of such variables [34].

4.2. Analysis of Multilevel and Multi-Type Policy Effects

According to the Hausman test results, this study employs a random-effects negative binomial panel regression to investigate the respective impacts of multi-type policies implemented by central governments, local governments, and multi-level governments on the innovation output of NEVs.

4.2.1. Policy Effects at the Central Government Level

Table 2 provides the regression estimate results of the negative binomial regression model that predicts the relationship between innovation output and national policies.Model (1) is the basic model with only control variables. All the control variables are significant at the 1% significance level. Per capita GDP, per capita income, and per capita patent ownership are positively correlated with the innovation output of NEVs. Models (2)-(8) show the effect of national-level policies on innovation output. Models (2), (3), and (6) indicate that strategic, supply-side, and environment-side policy instruments at the central government level, when implemented individually, have a positive and significant effect on NEV innovation output. Models (4) and (5) indicate that the demand-side policy of consumer subsidy phase-down at the central government level exerts an inverted U-shaped effect on the enhancement of innovation output in the NEV industry. In Model (8) , the four policy instruments are simultaneously employed by the central government, their effects on NEV innovation output rank in descending order as follows: supply-side (1.059*), environment-side (0.548*), strategic-side (0.373*), and demand-side (0.355*** and -0.0431*). Therefore, hypotheses 1, 2, 4, and 6 are supported.Moreover, relative to the single-variable models (Models 2, 3, and 6), the inclusion of the policy mix in the full model (Model 8) reveals that the effects of supply-side and environment-side policy instruments on NEV innovation output are enhanced to varying degrees, while the effects of demand-side and strategic-side policies are slightly weakened.

4.2.2. Policy Effect at the Local Government Level

Table 3 indicates the effect of local policies on the innovation output of NEVs. In Table 3, Models (9) and (12) show that the supply-side and environment-side policy instruments at the local government level, when implemented individually, each have a significant positive impact on the innovation output of NEVs. Models (10) and (11) reveal an inverted U-shaped effect of the local demand-side consumer subsidy phase-down policy on NEV innovation output.Model (14) is the full model. When all three policy instruments are simultaneously implemented at the local government level, their effects on the innovation output of NEVs, ranked in descending order of coefficient magnitude, are as follows: supply-side （0.421***）, environment-side （0.303***）, and demand-side (0.245*** and -0.0382***).Furthermore, compared with the single-variable models (Models 9, 10, 11, and 12), the effects of the three policy instruments at the local government level on NEV innovation output are each weakened to varying degrees. However, when employed as a policy mix, the overall policy effect of the three instruments remains greater than that of any single policy.

4.2.3. Policy Effects at the Multilevel Government Level

Table 4 indicates the effect of multilevel policies on the innovation output of NEVs. In Table 4, Model (8) serves as the baseline model. Models (15) and (18) examine the effects of the corresponding supply-side and environment-side policy instruments at the local government level, showing that both have a positive and significant impact on the enhancement of NEV innovation output under the influence of central government policy instruments. Models (16) and (17) indicate that the local demand-side policy of consumer subsidy phase-down still exhibits an inverted U-shaped effect. Model (20) is the full model. Compared with the baseline model (Model 8), when multi-type policies of both the central and local governments are simultaneously implemented, the effects of supply-side and demand-side policy instruments at the local government level on NEV innovation output, though slightly weakened, remain unchanged in direction. However, the positive effect of the local environment-side policy becomes insignificant. Therefore, hypotheses 3 and 5 are validated, while hypothesis 7 is not supported. Compared with the baseline model (8) and the single-variable models, the coefficients of the variables in the full model show little change, indicating that the multilevel policy effects are relatively stable.

4.3. Analysis of Policy Lag Effect Measurement

To further clarify the selection and effects of policy lags for the core explanatory variables, this study calculates the specific lag periods of multi-type policy lag effects in multilevel governments.

4.3.1. Lag Period Estimation of Multi-Type Policies at the Central Government Level

Table 5 presents the estimated lag periods of multi-type policies at the central government level. As shown in Table 5, the policy effects of strategic-side and supply-side policies at the central government level are maximized at a lag of two periods, after which they decline as the lag length increases. Although demand-side policy instruments at the central government level exhibit a certain lagged effect, their contemporaneous effect is the largest. The primary reason is that policy subsidies are directly deducted from the vehicle price at the time of purchase. Environment-side policy instruments at the central government level achieve their maximum effect at a lag of three periods.

4.3.2. Lag Period Estimation of Multi-Type Policies at the Local Government Level

Table 6 presents the estimated lag periods of multi-type policies at the local government level. Table 6 shows that local policies generally have shorter lag periods than central policies. At the local government level, supply-side policy instruments achieve their greatest effect in the current period. Although demand-side policies have longer lag periods, their implementation mechanism is consistent with that of central-level demand-side policies. Environment-side policies at the local level reach their maximum effect at lags of one and two periods.

4.4. Robustness Test

This study uses Akaike information criterion and Bayesian information criterion to test the suitability of the models. The values of Models (2)-(14) are all smaller than that of Model (1), indicating that the model is improved. Models (15)-(20) also exhibit improvements compared with previous models.Model (20) is the best-fitting model, as it effectively captures the relationship between the policy mix variables across different government tiers and the innovation output of new energy vehicles.

Moreover, two additional studies were conducted to ensure the robustness of results. The first study considered license plates. We control the complimentary license plates policy in pilot cities (Beijing, Shanghai, Guangzhou, Tianjin, Hangzhou, and Shenzhen), respectively. The license plate variable is a dummy variable (license = ‘1’ when the license policy is implemented and license = ‘0’ in other conditions). The regression results are shown in Table 7. The study found that the results are consistent with the positive relationship between policy and innovation.

The second study is to reduce the variance of patent outputs of pilot cities. We eliminated the pilot cities in the bottom 30% of the patent output ranking and retained 14 cities. The results in Table 8 also verify our hypotheses.

5.Synergy Effects of Multilevel Government Policies

This study further addresses the synergy effects of multilevel government policies. first, what are the effects of synergy or non-synergy between same-type policies at the central government level and the local government level on the innovation output of NEVs? Second, what are the synergy effects of multi-type policies in multilevel governments? Specifically, do the policy mixes consisting of supply-side, demand-side, and environment-side policy instruments at both the central and local levels generate a synergistic enhancing effect or a non-synergistic weakening effect on NEV innovation output?

Among the variables, supply_cd_Y, demand_cd_Y, and env_inf_cd_Y denote same-type policy synergy within multilevel governments, namely supply-side policy synergy, demand-side policy synergy, and environment-side policy synergy, respectively. The variables supply_cd_N, demand_cd_N, and env_inf_cd_N denote supply-side policy non-synergy, demand-side policy non-synergy, and environment-side policy non-synergy, respectively. SD_cd_Y and SD_cd_N denote the synergy and non-synergy, respectively, of supply-side and demand-side policies within multilevel governments. SDE_cd_Y and SDE_cd_N denote the synergy and non-synergy, respectively, of supply-side, demand-side, and environment-side policies within multilevel governments.

The study adopts a dummy variable approach to construct the policy synergy variables. Specifically, supply_cd_Y_it is coded as 1 if both the central government and local governments implemented NEV R&D support policies in the same year, indicating multilevel government supply-side policy synergy; otherwise, it is coded as 0. Other variables are defined analogously. The negative binomial regression results are presented in Table 9 and Table 10.

5.1. Same-Type Policy Synergy Effects in Multilevel Governments

Table 9 presents the negative binomial regression results for same-type policy synergy in multilevel governments. In Table 9, Model (21) is the baseline model. Models (22), (24), and (26) indicate that the synergy of same-type policies across multilevel governments—namely supply-side, demand-side, and environment-side policy synergy, respectively—has a significant positive effect on the innovation output of NEVs. Models (23) and (24) show that same-type policy non-synergy across multilevel governments impedes NEV innovation output, while Model (25) indicates that the non-synergy effect of environment-side policies is insignificant. Overall, Table 9 demonstrates that same-type policies across multilevel governments in China exhibit synergy effects.

5.2. Multi-Type Policy Synergy Effects in Multilevel Governments

Table 10 presents the negative binomial regression results for multi-type policy synergy in multilevel governments. In Table 10, Model (21) is the baseline model. Model (28) shows that the synergy of supply-side and demand-side policies in multilevel governments has a positive effect on NEV innovation output, whereas Model (29) reveals that the non-synergy of these policies exerts a negative impact. Model (30) shows that the synergy of supply-side, demand-side, and environment-side policies in multilevel governments has a significant positive effect on the innovation output of NEVs. In contrast, Model (31) reveals that the non-synergy of these policies exerts a negative impact on NEV innovation output. Overall, Table 10 demonstrates that multi-type policies across multilevel governments have positive synergy effects in China.

The study also conducts a robustness test of the policy synergy effects. By replacing the dependent variable with the number of invention patent applications, this paper obtains the regression results presented in Table A2. By adding the license plate policy as an additional control variable, this paper obtains the regression results shown in Table A3. The results of both tests are consistent with the regression results reported in Section 5. Therefore, the findings of this study exhibit good robustness.

6. Conclusion and Discussion

6.1. Main Conclusions

This work studies the effect of multilevel policy mixes on the innovation output of NEVs. Policies are divided into strategic, supply-side, demand-side, and environmental-side policies. In addition, policies are divided into national and local levels. Several results are noteworthy.

First, when considered individually, both national and local multi-type policies significantly enhance the innovation output of NEVs. Specifically, the demand-side policy of consumer subsidy phase-down exhibits an inverted U-shaped impact on NEV innovation output, whereas supply-side and environment-side policies demonstrate a more pronounced positive effect.

Second, when all policies are implemented simultaneously, the inclusion of supply-side and demand-side policies at the local government level generates a policy superposition effect, whereas the effect of local environment-side policy is crowded out.

Third, in China, multilevel governments exhibit policy synergy both within and across policy types. Just as same-type (supply-side, demand-side, environment-side) and multi-type synergies positively drive NEV innovation, policy non-synergy actively hinders it.

6.2. Discussion and Recommendations

In accordance with the study, policymakers should pay attention to the collocation of different types and levels of policies. First, although the effect of a national strategic policy on the innovation output of NEVs is insignificant when all policies are simultaneously in effect, the role of strategic policies in leading the development direction of NEVs can not be disregarded. Second, the national government should focus on supply-side policies, while local governments should focus on environmental-side policies. Supply-side policies, such as R&D subsidies, are the most direct and effective policies for promoting innovation output. Local governments are the primary implementer of environmental-side policies such as the construction of charging piles. Third, the NEV purchase subsidy policy will inevitably be withdrawn in the long run. The demand-side subsidy policy should be inclined toward high-tech innovative vehicle products, rather than subsidies for all NEV products. And, regulatory policies such as the dual credit policy are a good alternative and supplement to demand-side policies [4,35].Finally, governments should accordingly strive to enhance the synergistic effects among policies across different levels and of various types, aiming to amplify their combined efficacy while mitigating the adverse consequences arising from policy misalignment.

The study also exhibits some limitations. First, research on the policy implementation process uses virtual variables. We only study the effect of policies on innovation output promotion or inhibition and cannot quantify the intensity of the effect of each policy. Quantitative research on policy effectiveness is a crucial research direction in the future. The text analysis method based on machine learning is also a good method for quantifying policy intensity. Finally, we only empirically study China’s NEV policy, and thus, whether the research conclusions are applicable to other countries and industries must still be investigated.

Appendix A.1

Table A1. Robustness results of central and local multi-type policy effects.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
strategy0	0.646***	0.413***	0.613***	0.457***	0.690***	0.633***	0.626***	-0.101	0.567***	0.329***	0.627***
	(0.108)	(0.119)	(0.110)	(0.117)	(0.110)	(0.114)	(0.110)	(0.191)	(0.119)	(0.124)	(0.111)
lnper_GDP	0.613***	0.567***	0.572***	0.496***	0.579***	0.475***	0.613***	0.536***	0.576***	0.453***	0.493***
	(0.0958)	(0.0970)	(0.0947)	(0.0990)	(0.0970)	(0.101)	(0.0956)	(0.0964)	(0.0965)	(0.0984)	(0.0942)
lnper_income	0.857***	0.810***	0.816***	0.810***	0.842***	0.717***	0.863***	0.817***	0.833***	0.738***	0.759***
	(0.0551)	(0.0580)	(0.0564)	(0.0592)	(0.0565)	(0.0673)	(0.0552)	(0.0575)	(0.0564)	(0.0629)	(0.0585)
lnper_citypatent	0.422***	0.401***	0.428***	0.467***	0.435***	0.404***	0.427***	0.445***	0.441***	0.430***	0.452***
	(0.0488)	(0.0487)	(0.0484)	(0.0495)	(0.0491)	(0.0493)	(0.0491)	(0.0485)	(0.0498)	(0.0483)	(0.0481)
supply_cd_Y		0.471***
		(0.0839)
supply_cd_N			-0.214***
			(0.0704)
demand_cd_Y				0.338***
				(0.0630)
demand_cd_N					-0.168**
					(0.0698)
env_inf_cd_Y						0.454***
						(0.0680)
env_inf_cd_N							0.0682
							(0.0779)
SD_cd_Y								0.872***
								(0.170)
SD_cd_N									-0.160*
									(0.0950)
SDE_cd_Y										0.721***
										(0.0901)
SDE_cd_N											-0.383***
											(0.0719)
_cons	-16.16***	-15.17***	-15.19***	-14.48***	-15.67***	-13.25***	-16.23***	-14.93***	-15.45***	-13.21***	-13.68***
	(1.021)	(1.035)	(1.048)	(1.062)	(1.039)	(1.116)	(1.023)	(1.028)	(1.085)	(1.062)	(1.065)
ln_r	1.219***	1.294***	1.260***	1.359***	1.283***	1.261***	1.214***	1.249***	1.228***	1.265***	1.251***
	(0.313)	(0.314)	(0.314)	(0.317)	(0.316)	(0.313)	(0.313)	(0.313)	(0.313)	(0.312)	(0.313)
ln_s	4.048***	4.049***	4.059***	4.155***	4.118***	3.994***	4.037***	3.984***	4.042***	3.902***	3.977***
	(0.348)	(0.347)	(0.347)	(0.349)	(0.350)	(0.346)	(0.348)	(0.346)	(0.348)	(0.345)	(0.346)
N	580	580	580	580	580	580	580	580	580	580	580

* p < 0.1, ** p < 0.05, *** p < 0.01, Standard errors in parentheses.

Table A1. Robustness results of central and local multi-type policy effects.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
strategy0	0.646***	0.413***	0.613***	0.457***	0.690***	0.633***	0.626***	-0.101	0.567***	0.329***	0.627***
	(0.108)	(0.119)	(0.110)	(0.117)	(0.110)	(0.114)	(0.110)	(0.191)	(0.119)	(0.124)	(0.111)
lnper_GDP	0.613***	0.567***	0.572***	0.496***	0.579***	0.475***	0.613***	0.536***	0.576***	0.453***	0.493***
	(0.0958)	(0.0970)	(0.0947)	(0.0990)	(0.0970)	(0.101)	(0.0956)	(0.0964)	(0.0965)	(0.0984)	(0.0942)
lnper_income	0.857***	0.810***	0.816***	0.810***	0.842***	0.717***	0.863***	0.817***	0.833***	0.738***	0.759***
	(0.0551)	(0.0580)	(0.0564)	(0.0592)	(0.0565)	(0.0673)	(0.0552)	(0.0575)	(0.0564)	(0.0629)	(0.0585)
lnper_citypatent	0.422***	0.401***	0.428***	0.467***	0.435***	0.404***	0.427***	0.445***	0.441***	0.430***	0.452***
	(0.0488)	(0.0487)	(0.0484)	(0.0495)	(0.0491)	(0.0493)	(0.0491)	(0.0485)	(0.0498)	(0.0483)	(0.0481)
supply_cd_Y		0.471***
		(0.0839)
supply_cd_N			-0.214***
			(0.0704)
demand_cd_Y				0.338***
				(0.0630)
demand_cd_N					-0.168**
					(0.0698)
env_inf_cd_Y						0.454***
						(0.0680)
env_inf_cd_N							0.0682
							(0.0779)
SD_cd_Y								0.872***
								(0.170)
SD_cd_N									-0.160*
									(0.0950)
SDE_cd_Y										0.721***
										(0.0901)
SDE_cd_N											-0.383***
											(0.0719)
_cons	-16.16***	-15.17***	-15.19***	-14.48***	-15.67***	-13.25***	-16.23***	-14.93***	-15.45***	-13.21***	-13.68***
	(1.021)	(1.035)	(1.048)	(1.062)	(1.039)	(1.116)	(1.023)	(1.028)	(1.085)	(1.062)	(1.065)
ln_r	1.219***	1.294***	1.260***	1.359***	1.283***	1.261***	1.214***	1.249***	1.228***	1.265***	1.251***
	(0.313)	(0.314)	(0.314)	(0.317)	(0.316)	(0.313)	(0.313)	(0.313)	(0.313)	(0.312)	(0.313)
ln_s	4.048***	4.049***	4.059***	4.155***	4.118***	3.994***	4.037***	3.984***	4.042***	3.902***	3.977***
	(0.348)	(0.347)	(0.347)	(0.349)	(0.350)	(0.346)	(0.348)	(0.346)	(0.348)	(0.345)	(0.346)
N	580	580	580	580	580	580	580	580	580	580	580

* p < 0.1, ** p < 0.05, *** p < 0.01, Standard errors in parentheses.

Table A2. Robustness results for policy synergy effects.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
strategy0	0.645***	0.359***	0.596***	0.480***	0.680***	0.622***	0.641***	-0.103	0.549***	0.337***	0.613***
	(0.104)	(0.113)	(0.105)	(0.112)	(0.105)	(0.109)	(0.107)	(0.183)	(0.114)	(0.120)	(0.107)
lnper_GDP	0.504***	0.455***	0.458***	0.408***	0.478***	0.377***	0.504***	0.426***	0.465***	0.351***	0.400***
	(0.0914)	(0.0912)	(0.0882)	(0.0941)	(0.0925)	(0.0955)	(0.0913)	(0.0916)	(0.0910)	(0.0941)	(0.0899)
lnper_income	0.806***	0.773***	0.764***	0.776***	0.798***	0.704***	0.807***	0.777***	0.782***	0.720***	0.733***
	(0.0549)	(0.0559)	(0.0542)	(0.0576)	(0.0557)	(0.0629)	(0.0551)	(0.0563)	(0.0552)	(0.0602)	(0.0565)
lnper_citypatent	0.437***	0.429***	0.457***	0.482***	0.449***	0.444***	0.438***	0.469***	0.462***	0.473***	0.481***
	(0.0482)	(0.0468)	(0.0470)	(0.0491)	(0.0486)	(0.0487)	(0.0486)	(0.0477)	(0.0491)	(0.0478)	(0.0480)
license	0.194***	0.103	0.132*	0.138*	0.177**	0.0202	0.194***	0.152**	0.179**	0.0712	0.114
	(0.0750)	(0.0687)	(0.0717)	(0.0728)	(0.0749)	(0.0763)	(0.0749)	(0.0708)	(0.0739)	(0.0693)	(0.0722)
supply_cd_Y		0.578***
		(0.0810)
supply_cd_N			-0.310***
			(0.0676)
demand_cd_Y				0.294***
				(0.0613)
demand_cd_N					-0.135**
					(0.0683)
env_inf_cd_Y						0.446***
						(0.0671)
env_inf_cd_N							0.0143
							(0.0764)
SD_cd_Y								0.879***
								(0.163)
SD_cd_N									-0.193**
									(0.0900)
SDE_cd_Y										0.664***
										(0.0865)
SDE_cd_N											-0.349***
											(0.0691)
_cons	-14.43***	-13.55***	-13.34***	-13.16***	-14.09***	-12.07***	-14.44***	-13.31***	-13.67***	-11.95***	-12.42***
	(0.983)	(0.975)	(0.970)	(1.015)	(0.998)	(1.050)	(0.984)	(0.981)	(1.017)	(1.017)	(1.009)
ln_r	1.204***	1.265***	1.246***	1.324***	1.258***	1.197***	1.203***	1.226***	1.214***	1.223***	1.220***
	(0.314)	(0.313)	(0.314)	(0.318)	(0.317)	(0.313)	(0.314)	(0.313)	(0.314)	(0.312)	(0.313)
ln_s	4.469***	4.387***	4.436***	4.566***	4.530***	4.358***	4.468***	4.383***	4.457***	4.314***	4.392***
	(0.351)	(0.348)	(0.349)	(0.353)	(0.353)	(0.349)	(0.351)	(0.349)	(0.351)	(0.348)	(0.349)
N	580	580	580	580	580	580	580	580	580	580	580

* p < 0.1, ** p < 0.05, *** p < 0.01, Standard errors in parentheses.

Table A2. Robustness results for policy synergy effects.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
strategy0	0.645***	0.359***	0.596***	0.480***	0.680***	0.622***	0.641***	-0.103	0.549***	0.337***	0.613***
	(0.104)	(0.113)	(0.105)	(0.112)	(0.105)	(0.109)	(0.107)	(0.183)	(0.114)	(0.120)	(0.107)
lnper_GDP	0.504***	0.455***	0.458***	0.408***	0.478***	0.377***	0.504***	0.426***	0.465***	0.351***	0.400***
	(0.0914)	(0.0912)	(0.0882)	(0.0941)	(0.0925)	(0.0955)	(0.0913)	(0.0916)	(0.0910)	(0.0941)	(0.0899)
lnper_income	0.806***	0.773***	0.764***	0.776***	0.798***	0.704***	0.807***	0.777***	0.782***	0.720***	0.733***
	(0.0549)	(0.0559)	(0.0542)	(0.0576)	(0.0557)	(0.0629)	(0.0551)	(0.0563)	(0.0552)	(0.0602)	(0.0565)
lnper_citypatent	0.437***	0.429***	0.457***	0.482***	0.449***	0.444***	0.438***	0.469***	0.462***	0.473***	0.481***
	(0.0482)	(0.0468)	(0.0470)	(0.0491)	(0.0486)	(0.0487)	(0.0486)	(0.0477)	(0.0491)	(0.0478)	(0.0480)
license	0.194***	0.103	0.132*	0.138*	0.177**	0.0202	0.194***	0.152**	0.179**	0.0712	0.114
	(0.0750)	(0.0687)	(0.0717)	(0.0728)	(0.0749)	(0.0763)	(0.0749)	(0.0708)	(0.0739)	(0.0693)	(0.0722)
supply_cd_Y		0.578***
		(0.0810)
supply_cd_N			-0.310***
			(0.0676)
demand_cd_Y				0.294***
				(0.0613)
demand_cd_N					-0.135**
					(0.0683)
env_inf_cd_Y						0.446***
						(0.0671)
env_inf_cd_N							0.0143
							(0.0764)
SD_cd_Y								0.879***
								(0.163)
SD_cd_N									-0.193**
									(0.0900)
SDE_cd_Y										0.664***
										(0.0865)
SDE_cd_N											-0.349***
											(0.0691)
_cons	-14.43***	-13.55***	-13.34***	-13.16***	-14.09***	-12.07***	-14.44***	-13.31***	-13.67***	-11.95***	-12.42***
	(0.983)	(0.975)	(0.970)	(1.015)	(0.998)	(1.050)	(0.984)	(0.981)	(1.017)	(1.017)	(1.009)
ln_r	1.204***	1.265***	1.246***	1.324***	1.258***	1.197***	1.203***	1.226***	1.214***	1.223***	1.220***
	(0.314)	(0.313)	(0.314)	(0.318)	(0.317)	(0.313)	(0.314)	(0.313)	(0.314)	(0.312)	(0.313)
ln_s	4.469***	4.387***	4.436***	4.566***	4.530***	4.358***	4.468***	4.383***	4.457***	4.314***	4.392***
	(0.351)	(0.348)	(0.349)	(0.353)	(0.353)	(0.349)	(0.351)	(0.349)	(0.351)	(0.348)	(0.349)
N	580	580	580	580	580	580	580	580	580	580	580

* p < 0.1, ** p < 0.05, *** p < 0.01, Standard errors in parentheses.