Does regional competition hinder renewable energy innovation? evidence from Chinese cities

1 Introduction

In recent decades, China has sustained rapid economic growth, creating a miracle of leapfrog development. Yet, much like the trajectories of many developed countries, China’s cities concentrated on accelerating growth and expanding economic scale during the early stages of development, especially before 2019. As regional competition has intensified, the drawbacks of this crude growth model have become increasingly apparent. The high-input, high-energy-consumption approach often inflicts severe environmental damage, undermining a region’s green and sustainable development. As one of the fastest-growing economies of recent years, China simultaneously confronts serious environmental pollution, and despite a succession of policies and institutional arrangements, ecological governance has yet to deliver immediate and substantial results [1]. According to the “2018 China Ecological and Environmental Status Report,” the share of Chinese cities experiencing acid rain reached 37.6%, a year-on-year increase of 1.5%, and among the 818 nationally designated ecological functional areas, 11.4% of county-level cities recorded a decline in ecological environmental quality.

As a crucial force in environmental governance and green development, renewable energy technology innovation (RETI) acts as a catalyst for achieving the energy transition [2]. Despite notable progress in recent years, a certain gap remains between China and the developed world. In the first half of 2016, China’s average wind-power curtailment rate reached a historical high, signalling an unfavorable trend in curtailment problems. Against the backdrop of a widening global imbalance between energy supply and demand, issues of energy security, climate change, and sustainable development have drawn growing attention at home and abroad. The limitations of China’s RETI not only hinder the green development of regional economies but also weaken the capacity to respond effectively to sudden energy-security events.

Research suggests that one source of inefficiency in environmental governance is excessive regional competition. Within the Chinese governmental system, decentralized governance is characterized by “fiscal power moving up and administrative power moving down” between the central and local governments [3]. Because the work objectives of different government levels are broadly similar, competition arises among them. In this setting, officials’ promotions are largely determined by observable economic-performance indicators, producing a situation that resembles an “official promotion competition” [4]. Driven by promotion incentives, governments tend to exhibit a spending preference, prioritizing fiscal revenue for initiatives that yield rapid results over short cycles. What matters for our purposes is not that competition exists but that it is conducted through instruments with eligibility rules. Preferential access to land, fiscal support, and credit is allocated on criteria written in terms of indicators observable within an official’s assessment horizon, and those criteria are not neutral across sectors. Renewable energy technology development, which is research-intensive, intangible in its capital, and slow to register on measurable output, satisfies none of them. Under performance-assessment mechanisms centred on economic growth, whenever growth conflicts with environmental governance, governments often resort to various means to secure growth, and this form of regional competition ultimately depresses environmental-governance efficiency and dampens innovative applications [4].

From this perspective, both regional competition and RETI exert tangible influence in the fields of environmental governance and green development. Regional competition has been shown to lower both environmental quality efficiency and green total factor productivity, whereas RETI has significantly improved regional environmental quality [5,6]. Existing research, however, is largely confined to the application domains of RETI, with limited understanding of how the macroeconomic environment, and especially the characteristics of the governmental system, shapes RETI. Understanding the impact of regional competition on RETI is therefore crucial for addressing pressing environmental problems and is a prerequisite for achieving high-quality economic development [7].

This empirical study examines how regional competition influences RETI, addressing three specific questions. First, does regional competition have a discernible impact on RETI? Second, if such an impact exists, through which mechanisms is it transmitted? Third, does this competitive effect display notable heterogeneity? By answering these questions, the paper seeks to provide a theoretical foundation and policy recommendations for regional environmental-governance practice, ultimately accelerating effective and sustainable urban development.

Two challenges complicate the analysis. The first is endogeneity: the degree of regional competition is not exogenous and may stand in a reciprocal relationship with the level of RETI, so ordinary least squares (OLS) estimation may yield biased results [8]. To address this, the Bartik instrumental-variable method is employed for robustness testing. The second challenge concerns the sample window, which we end in 2018 for reasons of measurement rather than analytical convenience. Our RETI measure is constructed from renewable-energy patents dated by year of application, since the application date marks the timing of the underlying innovative activity. Because the interval between application and grant in our data has a median of approximately 2 years, counts for application years after 2018 were mechanically incomplete at the time of data collection, and their inclusion would generate a spurious downward trend that our estimates would absorb [9]. Extending the panel further would also strain the shift-share instrument, whose validity rests on base-period industrial shares remaining a credible predictor of exposure, and over a window approaching 15 years both relevance and the exclusion restriction weaken. Finally, the restrictions on factor mobility that followed 2019 altered the very margin through which our mechanism operates, namely, competition for mobile capital, so a coefficient estimated on that period would not be a different value of the same parameter but a different parameter.

We note that our window should not be read as a period of institutional stability. It spans the incorporation of ecological civilization into the national policy framework in 2012, the revised Environmental Protection Law effective in 2015, the nationwide rollout of the Central Environmental Inspection from 2016, and the Environmental Protection Tax Law in 2018. Regulatory stringency tightened substantially within our sample, and our estimates are therefore not identified off a single homogeneous regulatory regime [10].

To investigate these concerns, the study evaluates the effect of regional competition on RETI using data on Chinese prefecture-level cities from 2011 to 2018. The results indicate that regional competition inhibits urban RETI, and an examination of the underlying mechanisms reveals that this inhibition operates mainly through regional green economic efficiency, the entry of low-tech manufacturing enterprises, and resource misallocation. The inhibiting effect is further found to be more pronounced in cities with higher levels of digital development, heavier environmental pollution, and weaker environmental regulation.

The primary contributions of this paper lie in three areas. In terms of research content, it explores the effect of regional competition on RETI, filling gaps in both the understanding of regional-competition effects and the determinants of RETI. In terms of influencing mechanisms, it specifies how competition sorts economic activity into and out of preferential treatment. Rather than describing competition as a uniform relaxation of environmental stringency, we identify the eligibility rules governing land, fiscal, and credit allocation as the operative instrument, and show that RETI falls outside the resulting perimeter not because it is judged undesirable but because it is illegible to those rules. The three mechanism tests then trace a single sequence of composition, allocation and efficiency rather than three parallel channels. Third, by drawing the heterogeneity results together, the paper shows that the cities in which competition most strongly suppresses RETI are systematically those with the weakest capacity to resist it, and draws out the implication for institutional design: uniform benchmarks applied to heterogeneous capacity would compound the distortions we document rather than correct them.

The remainder of the paper is organized as follows. Section 2 presents the literature review and theoretical framework; Section 3 details the model setup and data; Section 4 reports the empirical analysis; and Section 5 concludes.

2 Literature review and theoretical framework

2.1 Regional competition

One strand of relevant literature concerns regional competition, with studies analysing its causes and consequences and proposing improvements. Against the backdrop of official promotion competition, the elevation of officials depends heavily on observable economic indicators, shaping a pattern of “competing for economic growth” that strongly influences government fiscal revenues and expenditures as well as regional industrial and development structures. On one hand, intensified regional competition raises administrative barriers among governments, aggravates market segmentation, triggers redundant construction, and diminishes the efficiency of resource allocation and fiscal-fund utilization across regions [11]. On the other hand, regional competition may itself become excessive, and the inefficiency it generates becomes an impediment to improving regional economic efficiency [12].

Moreover, the interaction between regional competition and industrial clustering may produce overcrowding and concentrated-emission effects, magnifying negative externalities and ultimately constraining the improvement of regional green total factor productivity [13]. Given the array of problems that regional competition brings, it has been argued that the central government should foster a scientific development outlook among local governments so as to curb malicious inter-regional competition [3].

2.2 Renewable energy technology innovation

Another relevant strand concerns RETI, which occupies a central position in the third global energy transition [14]. Amid this transition, scholars have shown broad interest in how to raise the level of RETI, and although no unified conclusion has been reached, the literature on its determinants spans four dimensions: policy, energy, environment, and economy. Policy factors are crucial determinants of RETI [15]. With the timely introduction of the “dual-carbon goals,” China’s sustainable-development philosophy has gained international recognition, creating a favourable environment for RETI. Energy factors, such as the scarcity of fossil fuels and frequent fluctuations in energy prices, act as catalysts that accelerate the advance of RETI [16]. Because environmental problems arising from greenhouse-gas emissions have attracted widespread public concern, countries increasingly prioritize green development, making the promotion of RETI a prevailing trend and a popular demand. From an economic standpoint, economic expansion has been identified as the primary driving force of RETI [17]. Financial development has also been shown to exert a positive impact on it [18].

Existing research has focused mainly on how regional competition affects business inflows and regional ecosystems, while studies on its relationship with RETI tend to concentrate on energy- and economy-related factors. Literature exploring the link between policy frameworks and RETI remains scarce. This paper therefore examines the impact of regional competition on RETI, with a view to offering empirical references and theoretical recommendations from the perspective of regional competition.

2.3 Hypotheses

Before setting out the hypotheses, it is worth stating concretely how competition sorts activity, since this determines which sectors it favours and which it passes over. Competition for mobile capital is conducted by allocating discretionary local resources, and allocation requires eligibility rules written in terms of indicators observable within an official’s assessment horizon. Three instruments carry most of the weight. Industrial land is discounted conditional on investment intensity and projected tax yield per unit area, criteria that research-intensive activity holding intangible capital cannot satisfy. Fiscal incentives are tied to near-term turnover, whereas research expenditure reduces current taxable profit before producing any measurable output. Credit intermediated by local financing platforms is screened on collateral, which again favours tangible fixed assets. Together these rules define a perimeter of qualifying activity, and renewable energy technology falls outside it not because it is judged undesirable but because it is illegible to the instruments through which competition is conducted. The hypotheses developed below, and the mechanism tests that examine them, follow from this account.

RETI exhibits significant externalities and contributes positively to environmental protection and sustainable economic development [19]. In this context, the private sector frequently undersupplies RETI [11]. RETI must therefore be led and promoted by governments. As China’s awareness of renewable energy has risen, local-government assessment criteria have gradually shifted from a single-minded pursuit of growth toward more diversified standards that place greater weight on environmental factors [20]. Nevertheless, RETI is a capital- and technology-intensive activity characterized by long investment cycles, unclear short-term returns, and substantial technological risk [21]. Because regional competition is still motivated by economic catch-up and oriented toward short-term gains, factors conducive to long-term benefits, such as RETI, are excluded from the core of performance assessment [22]. Governments imitate and compete with one another while focusing on short-term development, which in turn leads to neglect of RETI [20]. Accordingly, the following hypothesis is proposed.

Hypothesis 1 Regional competition stifles urban RETI. Beyond this direct effect, several latent mechanisms may operate between regional competition and urban RETI, and the full set of hypothesized paths is illustrated in Figure 1. First, in regional competition a favourable business environment may induce a “profit-seeking” inflow of resources, yet where local governments lack competitive advantages, they tend to attract productive projects or secondary industries that are fast-acting, highly polluting, and energy-intensive [23]. Such firms typically lack research-and-development capacity, and although they may generate agglomeration effects, they can lower the level of RETI or even burden the high-quality development of the regional economy [24]. In addition, when competition intensifies, governments often relax environmental regulation in order to compete for productive factors, especially liquid capital, triggering a large influx of low-end manufacturing that suppresses regional RETI [6]. Accordingly, the following hypothesis is proposed.

Hypothesis 2 The stifling effect of regional competition on RETI operates through an increase in low-tech manufacturing enterprises. Second, to strengthen the inflow of external resources and the “siphoning effect” on neighbouring regions, governments commonly relax environmental regulation as they compete [25]. When governments regard themselves as “market players,” they frequently intervene in regional economic development [26,27]. Because most RETI belongs to the environmental-protection sector and such expenditures contribute little to traditional economic sectors, local outlays are inevitably tilted toward education, science and technology, and culture [28]. When fiscal autonomy is insufficient, spending on RETI becomes very small, which not only lowers regional green productivity but also aggravates the imbalance of resource allocation across cities, thereby hindering the improvement of RETI. Accordingly, the following hypothesis is proposed.

Hypothesis 3 The stifling effect of regional competition is transmitted through an exacerbation of resource misallocation. Finally, the further imbalance of fiscal revenues and expenditures under regional competition leads to insufficient investment in public goods such as infrastructure within a jurisdiction [29]. This weakens the competitiveness of product markets and deepens market segmentation, thereby reducing regional green economic efficiency. Moreover, while competing for economic resources, governments often sacrifice resource and environmental protection as well as residents’ welfare in pursuit of short-term growth [25]. In such circumstances, the resulting neglect of green economic efficiency hinders the improvement of RETI. Accordingly, the following hypothesis is proposed.

Hypothesis 4 The stifling effect of regional competition is transmitted through a reduction in urban green economy efficiency.

3 Methods

3.1 Empirical model specification

This section examines the effect of regional competition on local RETI. Consistent with prior evidence that the levels of regional competition and RETI differ across cities, the following the following Equation 1 is used to solve this issue is used to solve this issue [30]:where RETI i , t denotes city i ‘s renewable energy technology innovation in year t, computed from Equation 2, and RC i , t captures the intensity of regional competition (RC), measured by the per capita amount of actually utilized foreign direct investment (FDI, in log). The term CONTROLS i , t represents the control variables described in Section 3.2.4, while γ i , μ t, and ε i , t denote the city fixed effect (CFE), the year fixed effect (YFE), and the random error, respectively.

R E T I i , t = α 0 + α 1 R C i , t + α 2 C O N T R O L S i , t + γ i + μ t + ε i , t (1)

3.2 Variables

3.2.1 Dependent variable

The dependent variable is the stock of urban renewable energy technological innovation, denoted RETI i , t, constructed by the perpetual-inventory knowledge-stock method [31]. Let PAT i , k denote the number of renewable-energy invention patents filed by applicants located in city i in year k and subsequently granted, retrieved from the INCOPAT database. Patents are dated by application year rather than grant year, since the application date marks the timing of the underlying innovative activity. The knowledge stock is as follows:where k 0 = 2009 is the first application year for which patent counts are collected, and ω 1 and ω 2 are the depreciation and diffusion rates, set to 0.36 and 0.3, respectively [7]. Two features of Equation 2 warrant comment. First, the weight applied to a vintage-k patent in Equation 3,is zero at t = k, rises to a maximum at t − k = ln ⁡ ω 1 + ω 2 / ω 1 / ω 2 ≈ 2.7 years, and declines thereafter. Contemporaneous filings therefore contribute nothing to the current stock, which is the intended property of the specification, since newly filed knowledge is assumed not to be immediately usable. Second, because ω 1 > 0, vintages older than roughly a decade receive weights below one percent, so the stock is insensitive to the choice of k 0 except in the years immediately following initialization. This is why patent counts are collected from 2009 while the estimation sample begins in 2011: the 2-year accumulation interval allows the weight function to reach the neighbourhood of its maximum before the first estimation year, so that no city enters the panel with a mechanically null or near-null stock.

R E T I i , t = ∑ k = k 0 t P A T i , k · ⁡ exp ⁡ − ω 1 t − k · 1 − exp ⁡ − ω 2 t − k (2)

w t − k = exp − ω 1 t − k 1 − exp − ω 2 t − k (3)

3.2.2 Independent variable

The independent variable is regional competition. During the phase of expansive development from 2011 to 2018, competition among regions for foreign capital emerged as a pivotal strategy for local governments [6]. Such competition aims to secure a broad array of developmental resources, with a particular emphasis on capital acquisition, and FDI is widely adopted to gauge the intensity of intergovernmental rivalry. Following this perspective, the per capita amount of actually utilized FDI (in log) is used as an indicator of regional competition [5].

3.2.3 Mechanism variables

To improve the accuracy of the mechanism analysis, three sets of mechanism variables are employed [32]. The first set captures manufacturing enterprises at different technical levels, for which the total number of urban manufacturing firms is counted separately for high-tech (HIGH), medium-high-tech (MEDHIGH), medium-low-tech (MEDLOW), and low-tech (LOW) enterprises, each serving as a substitution variable.

The second set measures resource misallocation. The underlying idea is that factors are efficiently allocated when their marginal revenue products are equalised, so that any departure from this condition can be read off the gap between the share of a factor a unit actually holds and the share its output contribution would warrant. The original formulation is set at the establishment level, whereas our unit of observation is the city, so the index is constructed from city aggregates as follows [33]:where K i and L i are the capital and labour employed in city i , K and L are the corresponding national totals, and s i = Y i / Y is the city’s share of national output. The output elasticities of capital and labour are obtained from a Cobb-Douglas production function Y i = A i K i β K L i β L estimated in logarithms with city and year fixed effects, the city effects absorbing time-invariant differences in total factor productivity; because the estimated elasticities are common across cities, they cancel from the ratio of the factor share to the output share, which is why they do not appear in Equation 4. A value of γ above unity indicates that the city absorbs more of the factor than its output contribution warrants and a value below unity the reverse, so what the index records is the absolute deviation from efficient allocation, and larger values of TKI and TLI correspond to greater distortion in either direction. The capital stock is constructed by the perpetual-inventory method at a depreciation rate of 9.6 percent, output and investment are deflated to constant 2011 prices, and labour is measured by year-end employment.

T K I i = 1 γ K i − 1 , γ K i = K i / K s i T L I i = 1 γ L i − 1 , γ L i = L i / L s i (4)

The third set is green economy efficiency (GEE) calculated in Equation 5. Because competition is expected to alter both the composition of production and the environmental burden it generates, the measure must admit undesirable outputs rather than treating pollution as a free by-product, and it must discriminate among cities that all lie on the frontier. A super-efficiency slacks-based measure with undesirable outputs satisfies both requirements and is estimated under variable returns to scale [34]. For the evaluated city k, with input vector x k, desirable output y k g and undesirable output y k b, the programme iswhere m is the number of inputs indexed by p , s 1 and s 2 are the numbers of desirable and undesirable outputs indexed by q and l, the summations run over all cities j other than the evaluated city k, and λ j are the intensity weights. Undesirable outputs enter the programme in the same direction as inputs, so that a city is credited for producing a given real output with less pollution rather than with more. Because the evaluated city is excluded from its own reference set, scores may exceed unity, which allows cities on the frontier to be ranked rather than tied at a common ceiling. The objective is a ratio of linear forms and is converted into an equivalent linear programme before solution. Inputs are the capital stock, year-end employment, and total energy consumption; the desirable output is real GDP; and the undesirable outputs are industrial wastewater discharge, industrial sulphur dioxide emissions, and industrial smoke and dust emissions.

θ * = min 1 m ∑ p = 1 m x ¯ p x p k 1 s 1 + s 2 ∑ q = 1 s 1 y ¯ q g y q k g + ∑ l = 1 s 2 y ¯ l b y l k b s . t . x ¯ ≥ ∑ j ≠ k λ j x j , y ¯ g ≤ ∑ j ≠ k λ j y j g , y ¯ b ≥ ∑ j ≠ k λ j y j b x ¯ ≥ x k , y ¯ g ≤ y k g , y ¯ b ≥ y k b ∑ j ≠ k λ j = 1 , λ j ≥ 0 (5)

3.2.4 Control variables

Following the related literature, a set of city-level control variables is included to account for potential confounders [35,36]. Per capita GDP (PGDP) is gross domestic product divided by the urban population. Fiscal expenditure (EXPEND) is the ratio of public-finance disbursement to GDP. Industrial structure (INDUS) is proxied by the share of secondary-industry value added in GDP. Educational resource input (EDUEXP) is captured by the proportion of local financial expenditure devoted to education. Urban financial level (FINANCE) is the ratio of the year-end deposit balance of financial institutions to GDP. Finally, urban innovational input (INNOPUT) is measured by the share of scientific expenditure in the general budget of local governments, expressed in ten thousand yuan.

3.3 Data sources and descriptive statistics

This study systematically explores how regional competition affects urban RETI, drawing on data from INCOPAT and the China City Statistical Yearbook. Renewable-energy patent applications are retrieved from 2009 onward, which provides the 2-year accumulation interval required by Equation 2, while the city-level dependent and control variables drawn from the China City Statistical Yearbook are available on a consistent basis only from 2011. The estimation sample therefore covers Chinese prefecture-level cities over 2011–2018. The closing year also reflects the policy environment. Following a national directive on environmental governance issued by the State Council in mid-2018, a series of environmental policies, including low-carbon-city and emissions-trading initiatives, was subsequently rolled out, and the Key Points of National Air Pollution Prevention and Control in 2019 further acted as a leading deterrent to unsustainable regional development [37,38]. Table 1 reports the descriptive statistics for all variables entering Model (1).

TABLE 1. Descriptive statistics of variables.

RETI, denotes renewable energy technology innovation, measured as the knowledge stock from Equation 2; RC, denotes regional competition, measured as per capita actually utilized foreign direct investment in log; PGDP, denotes per capita GDP; EXPEND, denotes fiscal expenditure as a share of GDP; INDUS, denotes industrial structure, the share of secondary-industry value added in GDP; EDUEXP, denotes educational resource input, the share of local financial expenditure devoted to education; FINANCE, denotes urban financial level, the year-end deposit balance of financial institutions over GDP; INNOPUT, denotes urban innovational input, the share of scientific expenditure in the general local budget.

4 Results

4.1 Benchmark test

Table 2 reports the regression results as control variables are added step by step. Column (1) of Table 2 shows that regional competition significantly reduces RETI, and after the successive inclusion of additional controls the coefficient on RC retains the same sign and significance throughout. This pattern reflects the spillover effects of regional competition, such as excessive economic expansion, market fragmentation, and the underutilization of resources, which together neglect environmental costs and the importance of sustainable development.

TABLE 2. Benchmark test.

RETI, denotes renewable energy technology innovation; RC, denotes regional competition; PGDP, denotes per capita GDP; EXPEND, denotes fiscal expenditure; INDUS, denotes industrial structure; EDUEXP, denotes educational resource input; FINANCE, denotes urban financial level; INNOPUT, denotes urban innovational input; CFE, denotes city fixed effects; YFE, denotes year fixed effects.

p < 0.1.

p < 0.05.

p < 0.01; standard errors clustered at the city level are reported in parentheses. This note applies to the following tables.

4.2 Robustness test

4.2.1 Substituting the dependent variable

To probe robustness, per capita RETI patent applications (PRETI) and total RETI patent applications (TRETI) are used in turn to replace RETI in Model (1) [39]. The estimated coefficients remain significant at the 1% level and continue to indicate an adverse effect, confirming that the baseline estimates are robust, as shown in columns (1) and (2) of Table 3.

TABLE 3. Robustness test results.

PRETI, denotes per capita RETI, patent applications; TRETI, denotes total RETI, patent applications; HIGHTECH, denotes high-technology industry agglomeration measured by the location-entropy method; CITYCLUSTER, denotes a dummy equal to one for cities in one of the top four city clusters; Bartik-RC, denotes the Bartik-simulated regional competition instrument, reported in scientific notation because of its scale. Column (5) reports the first stage of the two-stage least-squares estimation.

4.2.2 Other alternative factors

Other potential drivers of RETI may also be at work, so the analysis examines several confounders by incorporating high-technology industry agglomeration and city clusters, both recognized as engines of cleaner and more sustainable economic activity that may affect urban RETI. High-technology agglomeration (HIGHTECH) is gauged by the location-entropy method, and city cluster (CITYCLUSTER) is a dummy variable equal to one for cities located in one of the top four city clusters and zero otherwise. The estimates, reported in columns (3) and (4) of Table 3, remain statistically significant and consistent with the baseline. Reverse causality or omitted variables may nonetheless affect Model (1), so the Bartik instrumental-variable method is used to simulate the level of regional competition, yielding an estimate that is highly correlated with the actual value but uncorrelated with other residuals [40]. This satisfies the conditions of a valid instrument and mitigates endogeneity within a two-stage least-squares framework. The simulated instrument (Bartik-RC) exerts a direct positive effect on regional competition, and the Cragg-Donald Wald F statistic exceeds 10, confirming the robustness of the findings, as reported in columns (5) and (6) of Table 3.

4.3 Mechanism test

Building on the analysis above, the mechanisms involving the number of low-tech manufacturing enterprises, green economy efficiency, and resource misallocation are tested using the Equation 6 [41]:where MED i , t denotes city i ‘s mechanism variable, namely, the number of manufacturing enterprises at different technical levels, capital misallocation, labour misallocation, and green economy efficiency. All results are reported in Table 4.

M E D i , t = ρ 0 + ρ 1 R C i , t + ρ 2 C O N T R O L S i , t + γ i + μ t + ε i , t (6)

TABLE 4. Mechanism test.

HIGH, MEDHIGH, MEDLOW, and LOW, denote the number of high-tech, medium-high-tech, medium-low-tech and low-tech manufacturing enterprises respectively; TKI, denotes capital misallocation; TLI, denotes labour misallocation; GEE, denotes green economy efficiency.

The three tests trace a single sequence rather than three separate stories: competition alters what is produced (Section 4.3.1), which alters how factors are allocated (Section 4.3.2), which is summarized in the efficiency with which the local economy operates (Section 4.3.3).

4.3.1 Manufacturing enterprises

If competition operates through the eligibility rules described in Section 2.3, its first observable consequence should be compositional, in that activity should shift toward sectors satisfying the criteria for preferential land and fiscal treatment. The number of urban manufacturing enterprises at different technical levels is first used as the mechanism variable. As reported in columns (1) to (4) of Table 4, regional competition significantly increases the number of low-tech manufacturing enterprises, while its effect on high-tech and medium-high-tech enterprises is insignificant, indicating that competition depresses urban RETI mainly through the marginal expansion of low-tech manufacturing. An increasing number of low-tech enterprises is itself detrimental to sustainable development and cannot promote RETI, given their high energy consumption, carbon-dioxide emissions, and pollutant discharge [11].

The pattern of significance is itself informative. Competition raises entry only at the low-technology margin, leaving high- and medium-high-technology entry unaffected, which is what the perimeter argument predicts. Preferential land pricing conditioned on investment and tax yield per unit area, and fiscal incentives tied to near-term turnover, are instruments to which low-technology manufacturing responds strongly and technology-intensive activity responds weakly, since the latter’s returns fall outside the horizon over which the commitments are assessed. Competition is not indiscriminately pro-entry; it is selectively pro-entry for activities the instruments can reach.

4.3.2 Resource misallocation

Preferential allocation directs land, credit, and labour toward qualifying activities irrespective of their relative productivity. The second implication is therefore a widening dispersion in marginal revenue products, that is, misallocation in the sense of Section 3.2.3. Resource misallocation can hinder the innovation process by reducing technical and production efficiency, and may therefore exert a negative effect on RETI [32]. To test Hypothesis 3, capital and labour misallocation are examined, and columns (5) and (6) of Table 4 show that regional competition plays a meaningful role with respect to labour, such as employment and job opportunities. Urban RETI has been shown to bear a positive relationship with human capital, and in particular with labour structure [42]. These results therefore further confirm the adverse role of regional competition during this early period.

The asymmetry between the two estimates deserves comment rather than omission. Competition significantly worsens labour misallocation, while its effect on capital misallocation is not statistically distinguishable from zero. This is consistent with the mechanism rather than in tension with it. Capital in Chinese cities during our sample period was mobile across jurisdictions and responsive to the incentives cities offered, so the marginal return to capital was partially equalised by the competitive process itself. Labour was not, since household registration, housing costs, and the local provision of public services constrain mobility, so the composition of local employment adjusts to the composition of local production rather than to the returns available elsewhere. When competition tilts the industrial structure toward low-technology manufacturing, it is therefore the allocation of labour that bears the adjustment. The consequence for RETI is direct, since clean-technology development depends on the accumulation of specialised human capital, and a labour force locked into low-technology employment does not accumulate it.

4.3.3 Green economy efficiency

The preceding channels describe a reallocation of local activity toward the interior of the policy perimeter. Green economic efficiency provides a summary measure of what that reallocation costs. Because GEE is computed as a super-efficiency score in which capital, labour, and energy are inputs, pollutants are undesirable outputs, and real output is the desirable output, it registers jointly the composition of production and the environmental burden per unit of output. A decline in GEE is therefore the aggregate expression of the two mechanisms already documented, not an independent third channel.

The estimates in column (7) of Table 4 show that regional competition directly lowers urban green economy efficiency. Evaluated at the sample dispersion of the regressor reported in Table 1, a one-standard-deviation increase in competitive intensity is associated with a decline of approximately 0.010 units in GEE. The interpretation follows from Section 2.3. Where preferential access to land, fiscal support, and credit is conditioned on short-horizon indicators, the qualifying activities are those converting factors into measurable output quickly and at high environmental cost, while non-qualifying activities, including renewable energy technology development, face the ordinary cost of factors without offsetting support. The production structure that emerges obtains each unit of output with more energy and more pollution than the efficient frontier would require [43,44].

This matters for RETI through two further links rather than by definition. First, a production structure weighted toward low-technology, pollution-intensive activity generates weak derived demand for clean technology, since the local market for renewable energy solutions is thin precisely where the industrial base is least equipped to absorb them. Second, low green efficiency indicates that the environmental cost of conventional production is not priced locally, and where that cost is not borne by the firms generating it, the private return to substituting clean for conventional technology falls, and with it the incentive to develop such technology. Competition does not suppress RETI by prohibition. It suppresses RETI by sustaining a local economic structure in which the returns to clean-technology innovation are low and the returns to remaining inside the perimeter are high.

4.4 Heterogeneity test

4.4.1 Digital development

In digitally developed cities, business competition is intense and firms are frequently pressured to pursue short-term financial gains [30]. This subsection therefore asks whether the inhibiting effect of regional competition on urban RETI varies with the level of digital development, using the median level of urban digital finance as a proxy and splitting the sample at that median into a high group and a low group. As shown in columns (1) and (2) of Table 5, the inhibiting effect is more pronounced where digital development is higher.

TABLE 5. Heterogeneity test.

subsamples are formed by splitting the full sample at the median of urban digital finance, environmental pollution intensity, and the number of environmental laws and regulations respectively.

This result makes sense, since digitally advanced cities are not otherwise disadvantaged. The reading consistent with Section 2.3 is one of competing claims on a common resource rather than of weak capacity. Digital infrastructure is itself capital-intensive, tangible, and rapidly installed. It registers on precisely the indicators that govern preferential access to land, fiscal support, and credit, and it therefore sits inside the perimeter. Where digital expansion is proceeding quickly, it absorbs the discretionary resources a city has available, and it does so in competition with clean-technology development, which sits outside. The result is that intensified competition bites harder on RETI where an alternative qualifying use of local support is most readily available. Digital development also raises local energy demand, which strengthens the near-term case for reliable conventional supply over investment in renewable alternatives whose returns accrue later [35].

4.4.2 Environmental pollution

Urban environmental pollution can arise from a long-term dependence on traditional industries and a thin technological and knowledge base for renewable-energy innovation, which may leave competing cities at a relative disadvantage in RETI. Environmental pollution intensity is therefore computed to split the sample into a high group, with intensity above the median, and a low group below it [45]. Columns (3) and (4) of Table 5 indicate an analogous inhibiting effect in both groups, but the coefficient for the high group is larger.

The interpretation follows from the perimeter argument rather than from regulatory pressure. Where the existing industrial base is pollution-intensive, the activities qualifying for preferential treatment are already concentrated in those sectors, so intensified competition operates on a margin where the qualifying set is large and the non-qualifying set, including RETI, is correspondingly marginal. Two features reinforce this. The local fiscal base in such cities is disproportionately dependent on pollution-intensive output, which raises the political cost of any policy disadvantaging it and makes relaxation the path of least resistance. And the local technological and human-capital base in clean technology is thin, so the absorptive capacity required to convert any given level of support into realized innovation is low.

We note that this phenomenon is the opposite of what a regulatory-pressure account would predict. If heavily polluted cities faced binding and more urgent pollution-control obligations, those obligations should attenuate rather than amplify the suppressive effect of competition. The estimates indicate the reverse, which suggests that in these cities such obligations are not the operative constraint, a reading consistent with the regulation results reported next.

4.4.3 Environmental regulation

The stringency of environmental regulation differs under intensified competition, and weak regulation may fail to generate positive environmental policies and incentives capable of effectively guiding RETI. The number of laws and regulations related to environmental governance is used as a measure of urban environmental regulation, and the full sample is split at the median into a high group and a low group [46]. As columns (5) and (6) of Table 5 show, regional competition more prominently depresses RETI in cities with weaker regulation.

This is the cross-sectional counterpart of the mechanism developed in Section 2.3. Environmental regulation determines whether the environmental cost of conventional production is borne by the firms generating it. Where stringency is high, that cost is at least partially internalized, which raises the relative private return to clean technology and narrows the gap between activities inside and outside the perimeter, so that competition then has less room to widen it. Where stringency is low, the perimeter operates unconstrained: relaxation remains an available instrument of investment attraction, conventional production retains its cost advantage, and derived demand for renewable energy technology stays weak. Regulation, on this reading, does not act on RETI directly so much as it bounds the extent to which competition can be conducted on the environmental margin [47].

4.4.4 Heterogeneity discussion

The heterogeneity results reported in Table 5 are not independent. The cities in which competition most strongly suppresses RETI are disproportionately the same cities, namely, those with a pollution-intensive industrial base and weak regulatory capacity. This coincidence has an implication that the coefficients alone do not convey, and which we regard as the most consequential reading of our results.

Consider the position of such a city under competitive pressure. Its fiscal base rests on pollution-intensive output, which constrains its ability to tighten environmental standards without immediate revenue loss. Its low regulatory stringency is itself an instrument of investment attraction, so relaxing it further is the cheapest available response to competition from neighboring jurisdictions. The capital attracted on those terms deepens the dependence of its fiscal base on pollution-intensive activity, tightening the original constraint. Meanwhile its clean-technology base receives no accretion, since renewable energy technology remains outside the perimeter, local absorptive capacity does not develop, and the returns to future investment in it fall further relative to the alternative. Each round of competition makes the next round’s choice more constrained in the same direction.

This is the mechanism of directed technical change under path dependence: because innovation builds on the accumulated stock of knowledge in each direction, an economy specialised in dirty technology faces increasing costs of redirection over time, and delayed intervention is strictly more expensive than early intervention [48]. Applied across jurisdictions rather than across an economy, it implies that the cross-city dispersion in clean-technology capability we observe is unlikely to be self-correcting. Cities positioned as low-regulation destinations for pollution-intensive capital are not simply at a temporary disadvantage in RETI; they are on a trajectory along which that disadvantage compounds, while cities entering the period with a cleaner base and stronger regulatory capacity face progressively lower costs of clean-technology development.

Two consequences follow. The first is the sub-national analogue of the pollution haven: a stable spatial division of labour in which certain jurisdictions specialize in the pollution-intensive segment of national production, not as a transitional stage but as an equilibrium sustained by the competitive incentives themselves. The second is distributional. Because environmental burden is borne locally while the output it generates is traded nationally, this division of labour concentrates environmental damage on residents of the cities least equipped to abate it, and does so cumulatively.

We are explicit about the limits of what our design can say here. Our specification estimates an average effect and a set of subsample contrasts. It does not identify a threshold, and we make no claim that any critical level of competitive intensity exists beyond which divergence becomes irreversible. What our estimates establish is that the effect of competition is systematically stronger where initial conditions are worse, that is, the sign of the interaction between competitive pressure and initial disadvantage rather than its functional form. Whether that interaction generates a genuine absorbing state requires dynamic methods we do not use. The policy argument below does not depend on the stronger claim, since an effect that merely compounds, without any threshold, is sufficient to make the timing of intervention consequential.

5 Conclusion and implication

5.1 Conclusion

Using data on Chinese prefecture-level cities from 2011 to 2018, this study empirically assesses how regional competition influences RETI. The baseline regressions reveal a notably adverse effect of regional competition on RETI. The mechanism tests show that this effect operates as a sequence: competition expands entry at the low-technology margin of manufacturing while leaving technology-intensive entry unaffected, this compositional shift is absorbed principally by the allocation of labour, whose mobility is constrained, and the combined result is registered in a decline in green economic efficiency. The heterogeneity tests further demonstrate that the inhibiting effect is stronger in cities with more advanced digital development, heavier pollution, and weaker environmental regulation. Because the latter two disadvantages are mutually reinforcing, the dispersion in clean-technology capability across cities is unlikely to correct itself.

5.2 Policy implication

Our results locate the problem not in competition as such but in the criteria by which competition is conducted. Local governments respond rationally to an incentive structure rewarding outcomes observable within a short horizon and appropriable within a single jurisdiction, and renewable energy technology satisfies neither condition. Policy addressing the symptom by exhorting cities to invest more in clean technology, without altering those criteria, will be absorbed by the same logic that produced the problem. We therefore group our recommendations by the constraint each is intended to relax.

Redefining the perimeter. The eligibility rules governing preferential access to land, fiscal support, and credit are the operative instrument through which competition sorts sectors, and they are directly adjustable. Land-use assessment conditioned on investment and tax yield per unit area could admit research intensity and clean-technology output as qualifying criteria, so that innovation-intensive activity ceases to be illegible to the instrument. The local share of renewable energy support could be raised, so that a city promoting clean technology retains part of the fiscal return rather than bearing the local cost of a centrally captured benefit. And guided funds could be given explicit mandates for intangible-asset-intensive projects, which collateral-based screening otherwise excludes.

Removing the competitive margin. Because environmental relaxation is effective as an instrument of investment attraction only when neighboring jurisdictions do not match it, the incentive to relax weakens if minimum standards are set and enforced above the jurisdictional level. Coordinated enforcement floors within economic regions, and verification of environmental performance by bodies independent of the assessed government, would remove the margin on which the mechanism we document operates. This is a more direct remedy than encouraging cities to compete differently.

Reasonable accommodation for cities with disadvantageous initial conditions. The preceding recommendations raise a design problem our heterogeneity results make unavoidable. The cities in which competition most strongly suppresses RETI are those with the weakest fiscal base, the most pollution-dependent industrial structure, and the thinnest clean-technology capability. Imposing uniform and stringent environmental and innovation benchmarks on all cities would bind hardest precisely where the capacity to comply is lowest.

The predictable result is not compliance but its simulation. When a target is added to a tournament in which agents are evaluated comparatively, and when some agents cannot meet it through real performance, effort shifts toward the dimension cheapest to manipulate, namely, the measurement itself. This follows from the structure of the incentive rather than from any failing of the officials subject to it. Where several tasks are rewarded through a single observable index, effort moves toward whichever task the index registers at least cost, and adjusting what is reported is registered at far lower cost than abating what is emitted. Uniform stringency applied to heterogeneous capacity would therefore intensify rather than correct the administrative distortions we document, adding falsification and short-horizon compliance to the existing bias toward locally appropriable indicators while leaving the underlying constraint untouched.

Reasonable accommodation is accordingly not a concession to weak performers but a condition for the system to function as intended. Differentiated benchmarks would assess cities on improvement relative to their own baseline rather than on absolute attainment, so that the target is achievable through real effort in every jurisdiction and manipulation ceases to be the lower-cost route to compliance. Transitional grace periods, granted against a binding and published convergence trajectory rather than open-ended, would separate the timing of compliance from the timing of capacity acquisition. Capacity transfers would address the binding constraint directly, through fiscal transfers offsetting the revenue cost of retiring pollution-intensive capacity and, of greater long-run consequence, through technical transfers in the form of shared testing infrastructure, personnel exchange, and structured access to clean-technology licensing, since the deficit in these cities is one of absorptive capacity and not only of funds. Independent verification is the precondition for all three, because differentiated treatment allocated on the basis of self-reported performance would itself become an object of manipulation.

The case for acting early rests on the compounding documented in Section 4.4.4. Because the cost of redirecting technological development rises with the accumulated stock of knowledge in the incumbent direction, accommodation extended while cross-city dispersion remains moderate is substantially less expensive than the same accommodation extended once specialization has consolidated. This is an argument about the timing of intervention, and it does not require the existence of any threshold to hold.

5.3 Limitation

This study has three boundaries, which we state plainly. First, its design speaks to the average effect of competitive pressure and not to the structure of interaction that produces it. Our estimates come from a two-way fixed-effects specification identified by a shift-share instrument, with inference clustered at the city level. We do not estimate reaction functions between jurisdictions, and we do not test whether the effect of competition propagates spatially beyond the competing city. The theoretical account in Section 2.3 implies that such propagation is plausible, since environmental relaxation is effective only relative to neighboring jurisdictions and clean technology diffuses across administrative boundaries, but plausibility is not evidence, and we make no claim on the point. We note also that spatial dependence confined to the disturbances would leave our point estimates consistent while affecting inference, whereas bias in the coefficient of interest requires an omitted endogenous spatial lag of the outcome. The direction of any such bias is not signed a priori. If competitive pressure in neighbouring cities also depresses local RETI and competition is positively autocorrelated across space, our estimate absorbs part of the neighbourhood effect and is too large in absolute value; if competition is instead zero-sum in mobile factors, the omitted spillover carries the opposite sign and our estimate is conservative. We do not have the design to adjudicate between these two cases, and we draw no conclusion from either.

Second, our specification is a local linear approximation. The heterogeneity results establish that the marginal effect of competition varies systematically with city characteristics, and Section 4.4.4 sets out why this variation is likely to compound over time. Neither result identifies a threshold, and we do not claim that any critical level of competitive intensity exists beyond which cross-city divergence becomes self-sustaining. Testing that proposition would require dynamic methods capable of detecting multiple equilibria or absorbing states.

Third, our estimates pertain to 2011–2018. This window is determined by the availability of city-level covariates at the start of the panel, by the truncation of application-year patent counts at the time of data collection, and by the base-period requirements of the shift-share instrument, as set out in Section 1 and Section 3.3, rather than by analytical convenience. Whether the relationship between interjurisdictional competition and clean-technology innovation was altered by the regulatory tightening and the disruptions to factor mobility that followed is an important question we do not answer here.

These boundaries converge on a single extension. A spatial dynamic panel estimated over a longer window, in which cross-city interaction is modelled rather than assumed away, in which the possibility of a structural break is tested rather than imposed, and in which nonlinearity in competitive intensity is permitted rather than excluded by functional form, would address all three, and would allow the entrenchment mechanism described in Section 4.4.4 to be tested directly. We regard this as the natural next step and offer the present results as the reduced-form evidence that motivates it. In addition, given data constraints, subsequent studies could extend the analysis to a broader set of developing economies, and to comparisons with developed nations, in order to trace how the impact of regional competition on RETI evolves.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Peerapong Uthansakul, Suranaree University of Technology, Thailand

Reviewed by: Yasuko Kawahata, Rikkyo University, Japan

Yanting Ji, Zhejiang University of Technology, China