Can innovative city pilot policy reduce income inequality?

Li, Xiang; Shen, Guihua; Zhang, Xiuwu Article Can innovative city pilot policy reduce income inequality? Journal of Innovation & Knowledge (JIK) Provided in Cooperation with: Elsevier Suggested Citation: Li, Xiang; Shen, Guihua; Zhang, Xiuwu (2025) : Can innovative city pilot policy reduce income inequality?, Journal of Innovation & Knowledge (JIK), ISSN 2444-569X, Elsevier, Amsterdam, Vol. 10, Iss. 4, pp. 1-14, https://doi.org/10.1016/j.jik.2025.100749 This Version is available at: https://hdl.handle.net/10419/327644 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Can innovative city pilot policy reduce income inequality? Xiang Li a,* , Guihua Shen b , Xiuwu Zhang c a School of Economics and Finance, Huaqiao University, Quanzhou 362021, China b Jiangxi Vocational Education and Industrial Research Institute, Jiangxi Science & Technology Normal University, Nanchang 330000, China c Research Center for Quantitative Economics, Huaqiao University, Xiamen 361021, China ARTICLE INFO JEL classification: C33 D31 Keywords: Innovative city pilot policy Income inequality Multi-period difference-in-differences model ABSTRACT Ensuring fairness in income distribution is a fundamental requirement in achieving common prosperity. This study employed panel data sourced from 276 cities throughout China, covering the time frame from 2003 to 2022. It sets up a multi-stage difference-in-differences (DID) model to explore the impacts of the innovative city pilot policy on income inequality. Results indicate that this policy significantly reduces income inequality, a conclusion that remains robust across various tests. Meanwhile, the effects are more pronounced in central and western regions, non-border cities, Yangtze River Economic Belt cities, non-old industrial bases, ethnic-minority areas, low administrative cities, low initial income inequality and non-resource-based cities. The mechanism analysis indicates that these policies mitigate income inequality largely by fostering labour resource agglomeration, structure optimisation and innovative vitality. Through an analysis of the impacts of innovative city pilot initiatives, this study enriches our comprehension and provides significant perspectives for promoting income equality in the new epoch. In addition, it provides strategic guidance for expanding and scaling these pilot policies to broader contexts. Introduction The rapid progress of global economic integration and technological advancements has made income inequality a major challenge. It poses notable risks to social cohesion and sustainable economic development. In China, the income gap has been further exacerbated by the urban- –rural dual structure, constraining resource efficiency and undermining social equity. The growing disparity intensifies social tensions and threatens stability and societal well-being. Therefore, developing effective policy measures to address income inequality has become a top priority for policymakers and researchers globally. The pilot programme for innovative cities is a pivotal strategy to promote the development of an innovation-oriented nation. It plays a key role in guiding the economic transition of China from a phase of factor-driven expansion to one characterised by innovation-led development. This policy aimed to strengthen the capacity of the cities for independent innovation, optimise industrial structures and cultivate an innovation-conducive environment. Its goal was to drive sustainable, high-quality growth. Since the pilot work of building innovative cities was launched in 2010, some cities have been included in the scope of the pilot programme. In the development of innovative cities, local governments have taken a leading role. They actively harness the decisive function of market mechanisms in allocating innovation resources. This has resulted in a distinctive model defined by the synergy between government and market forces. There is a synergy between government leadership and market dynamics. This synergy has played a vital role in reducing the risks linked to corporate research and development. For instance, the policy has implemented various measures, such as attracting high-calibre talent, increasing financial investments, strengthening intellectual property protections and building innovation infrastructure. These measures have effectively improved the innovation ecosystem. These efforts have significantly enhanced cities’ innovation capacity and attracted clusters of high-end industries (Berrone et al., 2013). However, whether these benefits are distributed equitably among all social groups remains unclear, particularly among rural and low-income populations. Innovation activities can boost the demand for highly skilled labour and increase their income levels. However, rural and low-income groups may be left behind if innovation remains concentrated in urban centres. Such exclusion risks further exacerbate the income gap. Consequently, how innovative city pilot policies affect income inequality and their effectiveness remain vital concerns that merit more extensive research. * Corresponding author. E-mail address: [email protected] (X. Li). Contents lists available at ScienceDirect Journal of Innovation & Knowledge journal homepage: www.elsevier.com/locate/jik https://doi.org/10.1016/j.jik.2025.100749 Received 24 February 2025; Accepted 31 May 2025 Journal of Innovation & Knowledge 10 (2025) 100749 Available online 18 June 2025 2444-569X/© 2025 The Authors. Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ). This study employed panel data sourced from 276 cities throughout China from 2003 to 2022. This framework establishes a multi-stage difference-in-differences (DID) framework. The aim of this study was to explore the impacts of the innovative urban pilot policy on the income inequality situation. The outcomes indicate that the enforcement of the pilot policy significantly reduces income inequality. This conclusion holds true after multiple robustness checks. Notably, the effects of this policy manifest more prominently in specific geographical and administrative contexts. These include the central-western regions, non-border cities, cities positioned within the ambit of the Yangtze River Economic Belt, cities with lower administrative hierarchies, locales that initially display a lower level of income disparity and non-resource-dependent cities. Mechanism analysis indicates that the policy promotes fair income distribution primarily through labour resource agglomeration, structure optimisation and innovative vitality. The innovations of this study are as follows. First, this study analysed the impact of innovation policies on income distribution. It fills the gap in social effect analysis from the perspective of innovative development. In addition, it found that innovative urban development helps narrow income gaps and significantly reduces income inequality. Moreover, this study investigated the dynamic evolution of such influence, considering policy marginal effects and external shocks. It provides decision-making guidance for expanding the scope of innovative urban pilot programmes and also offers a new perspective on achieving common prosperity. Second, regarding research content, this study dynamically demonstrates the long-term impact of innovation policies on income inequality. In this process, theory and empirical evidence are effectively integrated to conduct a detailed demonstration. By elaborating on the institutional background, this study becomes more specific and reliable. In addition, this study explores multiple influence paths in depth, highlighting three key pathways: labour resource agglomeration, structure optimisation and enhanced innovative vitality. Moreover, this study addresses the limitations of extant research on income distribution. In particular, previous studies often struggle to isolate income inequality from broader concepts such as common prosperity and tend to offer only shallow analyses of the underlying mechanisms. This study clarifies how the policies mitigate income inequality and provides theoretical support for designing further measures to reduce income gaps. Third, in terms of research methods, this study employed a multi-stage DID model, which is a rigorous and sophisticated econometric approach. Robustness was verified using propensity score matching DID (PSM-DID) and instrumental variable methods, effectively minimising estimation bias. By analysing heterogeneity, this study explored the complicated implications of innovation policies on income inequality. This enhances the understanding of the social effects of innovative urban development. Literature review Innovation and income inequality Academia has extensively debated the impact of innovation on income gaps, yielding two contrasting perspectives. Some studies suggest that technological innovation tends to widen income disparities. In particular, income growth among the wealthiest groups is more pronounced (Aghion et al., 2019). This effect is primarily attributed to the unequal distribution of innovation returns, increasing labour income inequality (Permana et al., 2018). In particular, skill-oriented technological advancements have spurred a substantial increase in the demand for highly skilled labour. Consequently, this has caused a significant elevation in skill-related premiums and income inequality (Acemoglu & Restrepo, 2018, 2019). In addition, technological progress in China has exhibited a notable skill bias. This skill bias is the primary cause of widening regional wage disparities (Wang et al., 2022). Moreover, technological innovation shifts skill distribution in the labour market through a ‘screening effect’. This effect benefits high-skilled workers; however, it may sideline low-skilled workers (Lee & Pose, 2013; Michaels et al., 2014). This, in turn, intensifies income disparities. Variances in substitution elasticity among technological advancements and diverse tasks hold significant sway over income inequality. High-skilled labour is typically complementary to technology, whereas low-skilled labour is vulnerable to displacement (Yu et al., 2021). Some researchers contend that technological innovation can reduce income disparities via mechanisms such as ‘knowledge spillovers’ and ‘capital conservation’. The ‘learning-by-doing’ phenomenon related to technological innovation allows low-skilled workers to upgrade their skills by learning, which fosters knowledge spillover. These knowledge spillovers subsequently result in a narrowing of wage differences. Technological innovation, characterised by capital conservation, may alleviate income inequality indirectly by reducing rental expenditures (Antonelli & Gehringer, 2017). Neutral technological advancements can promote growth in the availability of a skilled workforce. Therefore, the wage gap separating skilled from unskilled labourers can be narrowed (Dong et al., 2014; Liu & Zhang 2017). Moreover, the combined influence of urbanisation and technological innovation can mitigate income inequality among residents (Zhao et al., 2018). Generally, workers with high human capital are directly involved in innovation and obtain high returns through research or technological complementarity (Aghion et al., 2019). By contrast, those with low human capital indirectly participate in innovation. If workers with low human capital successfully acquire new technologies, their income may rise. However, if they fail to do so, they may face a risk of decreased income or even marginalisation. Consequently, as technology diffusion and skill upgrading proceed, the adverse influence of innovation on income inequality may progressively weaken (Yan et al., 2023). Economic effect of innovation policies Cities play a crucial function as the main drivers of economic growth in China, particularly in terms of promoting innovation (Davis & Dingel, 2019). Major urban agglomerations in China host ~90% of innovation activities within ~20% of its land area (Zhou et al., 2021). Currently, the majority of research efforts have been mainly focused on economic impacts. These impacts are engendered by executing the innovation pilot city policy and are one sided. For instance, research has highlighted the influence of fostering the synergy between pollution abatement and carbon reduction in urban settings. Innovation-driven policies can promote the coordinated improvement of pollution reduction and carbon reduction in cities (Yang & Xue, 2024). In addition, these policies can enhance urban green ecological efficiency. Meanwhile, in entrepreneurial vitality studies, innovative city pilot policies strongly promote urban green entrepreneurship (Yang & Liu, 2024). The impact of pilot policies on urban innovation follows an asymmetric inverted-V trend, rising then falling (Li & Yang, 2019) and enhancing industrial efficiency and structure via factor and technology agglomeration (Hu et al., 2020). However, the construction of innovation-oriented cities causes changes in economic effects. In addition, it triggers alterations in social effects, such as the pattern of income distribution. However, extant research on this issue is comparatively limited. Most studies on the impact of innovation in China on income inequality have indirectly examined it through certain angles such as technological change. They overlook the social impacts of the pilot policy, particularly income inequality caused by changes in the innovation policy environment. Social effects of innovation policies Literature on the relationship between innovation policies and income distribution is scarce. For instance, Yang and Li (2023) examined the impact of innovation policies on common prosperity. Their study constructed a comprehensive index system for this purpose. However, the findings heavily depend on index measurements without a unified standard. These results merely reflect the overall macro-situation and do not thoroughly explore income inequality. Common prosperity, in X. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 2 essence, is a comprehensive development objective, embodying ‘shared development’ and ‘equitable distribution of outcomes’. The data and methodology issues arise because research using ‘common prosperity’ indicators may not fully capture income inequality. This characteristic makes it impossible to separate income distributions from the research framework. Therefore, it precludes in-depth research on income inequality. Although there is a heterogeneity analysis, an analysis of internal differences among different regions and city types is not detailed enough. To address these research shortcomings, this study focused on income inequality. It reduces biases using accurate indicator data, applying instrumental variables and conducting multiple robustness tests. In addition, it strengthens heterogeneity analysis to enhance the generalisability of the results. Meanwhile, Xu and Zeng (2024) studied the impact of innovative city pilot policies on the income gap. However, by focusing on only one mechanism, their analysis is too superficial, yielding infeasible countermeasures and suggestions. This study confines itself to the short-term outcomes of policies, neglecting the long-term impacts of policy implementation from a dynamic perspective. This study explored the impact of innovation policies on income inequality. Moreover, a more comprehensive analysis was conducted in three dimensions: labour agglomeration, structural optimisation and innovation vitality. Institutional background and theoretical analysis Institutional background (1) The Innovative City Pilot Policy The innovative urban pilot policy is a gradual reform under the innovation-driven strategy in China. It extends innovation efforts from individual actors to the city level and integrates innovation activities into urban governance. Cities are the key spatial platform for implementing this strategy. The pilot policy is a unique Chinese policy tool. Its diffusion principle lies in the central government granting local governments institutional space for ‘early experimentation’. This approach encourages local policy innovation and selects successful practices for wider adoption. In 2005, the State Council released a National Mediumand Long-term Science and Technology Development Plan (2006–2020). This plan sets the strategic goal of building an innovationdriven nation. In 2008, Shenzhen was selected as the first innovative city pilot, marking the official launch of an innovative urban development initiative. In 2010, the National Development and Reform Commission (NDRC) and the Ministry of Science and Technology (MOST) jointly approved 44 cities and districts as innovative city pilots. This approval further advanced the strategy of building an innovation-driven nation. By 2016, the NDRC and MOST consolidated earlier pilot programmes, establishing 61 innovative city pilots. This number grew to 78 by 2018, encompassing national-level innovative city pilot programmes in cities and districts. By 2022, MOST supported an additional 25 cities, including Baoding, to implement innovative urban development. The total number of innovative city pilots approved by the two ministries was 103. Among them, there were 97 prefecture-level cities, 4 districts in municipalities directly under the central government and 2 countylevel cities. (2) Household Registration System and Income Gap During the economic transformation in China, institutional breakthroughs and policy innovation are crucial for realising common prosperity. As the cornerstone of the urban–rural dual system, the household registration system has long restricted labour mobility. The traditional household registration system creates multiple barriers to urban employment, social security and public services for rural residents. Consequently, this situation reduces the spatial allocation efficiency of the labour force. Despite reform-relaxed household registration rules, slow rural labour urbanisation hinders urban–rural income gap reduction. For instance, despite the influx of migrant workers into cities, household registration restricts their equal access to urban education and healthcare. This hinders family-level migration, undermining longterm labour supply stability and impeding the natural narrowing of the urban–rural income gap through urbanisation. Meanwhile, the Chinese government-led innovation policies have deeply intervened in the market mechanism through institutional design and resource allocation. Therefore, these policies reshape the regional economic development landscape. The government relies on policy tools such as innovative city pilots and development zone construction to break administrative barriers and guide factors in gathering in key areas. Through these efforts, a ‘policy-driven’ development model is formed (Hu et al., 2020). This top–down policy intervention not only addresses market failures but also releases reform dividends through institutional innovation. This approach provides new paths to narrowing income gaps. Theoretical analysis of marginal effects Promoting income distribution through technological innovation is an integral part of achieving common prosperity. The core of income equity is ensuring that development benefits are fairly and reasonably shared across all social strata. Moreover, it focuses on fostering economic growth and innovation efficiency. Institutional innovation and policy optimisation enhance the inclusiveness and sharing of socioeconomic development. Therefore, they narrow income gaps and promote common prosperity. The hypothesis is that the economic workforce is divided into highly skilled (H) and low-skilled (L) workers. Innovation policies that promote technological progress (A) improve the productivity of low-skilled workers and thus decrease the income gap. The production function can be calculated as follows: Y=A(H α L1− α ),0< α <1 (1) where A is the growth in total factor productivity generated by innovative policies and α is the output elasticity of highly skilled labour. The incomes of the high-skilled and the low-skilled labour force can be calculated as follows: WH= α ⋅Y H,WL= (1− α )⋅Y L(2) The indicator of income inequality is as follows: G=WH WL = α 1− α ⋅L H(3) Assume that innovation policy input I decreases G through an increase in A. For instance, policies provide subsidies for low-skilled workers’ skills training to improve productivity. With an increase in I, the growth of A declines, meaning that ∂ A ∂ I>0, ∂ 2A ∂ I2<0 (4) This results in a reduced pace of narrowing the income gap G, i.e. ∂ G ∂ I<0, ∂ 2G ∂ I2>0 (5) Given the above analysis, we can infer that the impact of innovation policies on narrowing income inequality declines as the input of innovation policies increases. Theoretical mechanism analysis The mechanism through which innovation-oriented policies affect income inequality is primarily manifested in three respects: labour resource agglomeration, structure optimisation and stimulated X. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 3 innovative vitality. (1) The Aggregation Mechanism of Labour Resources The pilot policy for innovative cities promotes income equalisation through the agglomeration of labour resources. This process is a crucial channel in reducing income disparities. The Chinese household registration system causes the ‘semi-urbanisation’ of the labour force. Migrant workers cannot fully enjoy urban resident benefits, affecting family migration and labour force stability. From the perspective of the spatial agglomeration of the population, urbanisation optimises the urban–rural spatial layout, breaks barriers, promotes resource flow and deepens integration (Portnov & Schwartz, 2009). This series of effects of urbanisation is fundamental to narrowing the income disparity. The pilot policy for innovative cities has been broken down into two aspects. First, innovative policies should upgrade infrastructure and promote public service equalisation (Zhao et al., 2023). This includes building affordable housing and improving education policies for migrant workers’ children. These actions reduce the cost of rural labour migration and weaken the constraints of the household registration system. Second, ‘talent policies’ serve as a breakthrough. Housing subsidies, entrepreneurship support and other measures are implemented to attract highly skilled talents. These initiatives foster an agglomeration effect in which talents attract more talents. This generates economies of scale and knowledge spillover. Therefore, more jobs are created, and workers earn more, thereby improving income equality (Dougal et al., 2015; Shen et al., 2019). For instance, through talent policies, innovative pilot cities such as Shenzhen and Hangzhou have attracted high-end elements, substantially reducing the income gap with traditional industrial cities (Sun et al., 2022). (2) Structural Optimisation Mechanism The institutional environment plays a shaping role in regional development and industrial upgrading strategies in China. Innovative pilot cities can leverage technological breakthroughs to promote balanced industrial growth, drive equipment upgrades and boost efficiency. In addition, these cities can use such breakthroughs to foster innovative business models (Bartelsman et al., 2013; Uzunidis, 2016). China is at a critical juncture in its shift towards high-quality economic development. During this period, industrial upgrading not only enhances resource allocation but also significantly improves the fairness of income distribution (Wu et al., 2018). In this progression, labour increasingly migrates to high-value-added sectors, creating wide-ranging employment opportunities and enhancing household earnings (Deng & He, 2018). Therefore, this process reduces income disparity. For instance, ‘Made in China 2025’ drives the intelligent transformation of the manufacturing industry. This initiative generates numerous high-skilled employment opportunities and stimulates income growth. Industrial structure transformation is increasingly linked to efficiency gains, equitable distribution and the interplay between production and distribution (Guo & Luo, 2021). Regarding the employment structure, the government has sponsored vocational training and service improvements. These efforts have increased the alignment between workers’ human capital and job opportunities (Zhou & Chen, 2021). In particular, the ‘Vocational Skills Enhancement Initiative’ offers tailored training to migrant workers and other groups, equipping them with the necessary capabilities. This combination of policy intervention and market mechanisms fully leverages the guiding role of the government. Moreover, it stimulates the resource-allocation efficiency in the market, thereby achieving a balance between ‘efficiency and fairness’ in income distribution. (3) Innovative Vitality Stimulation Mechanism The stimulation of entrepreneurial vitality is another crucial aspect through which innovation-driven policies exert a positive impact on income inequality. The convergence of venture capital and the expansion of financing channels have effectively alleviated the financial pressures faced by start-ups (Stiglitz, 2015; Mulier & Samarin, 2021). The agglomeration of innovative talent accelerates the flow of knowledge and technological innovation. The continuous augmentation of human resource endowment holds the key to fostering innovation within high-tech firms (Huang et al., 2023). In addition, policies aimed to refine the business environment and enhance government service efficiency. They provide entrepreneurs with a more robust support system and strengthen their ability to withstand market uncertainties (Ding et al., 2021; Juan et al., 2024). The flourishing of entrepreneurial activities is a powerful catalyst for the rise of novel industries and innovative business paradigms. Meanwhile, increased market competition and efficient resource integration concurrently optimise income distribution structures (Zhao et al., 2020). At the urban scale, entrepreneurial activities disrupt market disequilibria and foster a substantial number of job opportunities. Conversely, at the county level, they play a pivotal role in augmenting farmers’ incomes and reducing the urban- –rural income gap (Ye et al., 2022). The concentration of talent accelerates knowledge flow and technological innovation. Moreover, it reduces knowledge exchange costs, promoting the rapid diffusion and application of new technologies and ideas. Therefore, innovation policies have lowered entrepreneurial barriers and enhanced support and financing (Bai et al., 2022). These policies have ignited societal enthusiasm for innovation and entrepreneurship. This promotes the growth of micro and small enterprises, creates economic growth points and jobs and increases income opportunities. Consequently, it helps alleviate income inequality. Considering this, the subsequent hypotheses are proposed in this study: H1: The enactment of the innovative city pilot policy has a positive impact on alleviating income inequality. However, with an increased input of innovation policies, their impact on narrowing income inequality declines. H2a: Income inequality is reduced through an innovative city pilot policy via labour resource agglomeration. H2b: Income inequality is reduced by the innovative city pilot policy through the optimisation of industrial and employment structures. H2c: Income inequality is reduced through an innovative city pilot policy that enhances urban innovative vitality. Model, variables and data Model The DID technique is a frequently used econometric tool to evaluate the influence of policy enactment. The proposed model allows the analysis and quantification of policy impacts while minimising interference from other factors. The fundamental concept of DID is to regard policy implementation and institutional changes as exogenous factors. In particular, these factors are considered ‘quasi-experiments’ or ‘natural experiments’ within an economic system. This approach assumes that policy implementation follows a mechanism similar to random assignment. This mechanism guarantees that the characteristics and tendencies of the treatment and control groups are comparable. This methodology examines the changes in outcomes within the experimental and control groups before and after the policy is enacted. Through this examination, the differences that can be attributed to the policy can be identified. This approach mitigates endogeneity issues arising from external factors, allowing the estimation of the net effect of the policy. The pilot project for innovative cities exhibits the traits typical of a ‘quasi-natural experiment’. Drawing upon the features of the DID model, this approach offers two key advantages. First, it harnesses the timeX. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 4 series data across multiple periods. This allows the tracking of dynamic shifts in income inequality at various intervals following policy rollout. Income inequality is not instantaneously affected by the innovative city pilot policy; rather, it unfolds gradually over time. Second, the multiperiod DID method can effectively distinguish the treatment group from the control group in the preand post-policy implementation periods. Therefore, it allows for a precise estimation of the causal connection between the innovative city pilot policy and income inequality. This makes it suitable for evaluation via the DID method. However, as the policy was rolled out in several stages, a multi-period DID methodology was adopted in this research to formulate the model (Beck & Levkov, 2010; Wang et al., 2023). To assess whether the policy effectively reduces income inequality, the following equation was established: Giniit =β0+β1DIDit +β2∑Controlit +ui+λt+ ε it (6) where Gini it is the degree of income inequality in city i in year t and DID it depicts whether city i is designated as an innovative trial city in year t. If city i belongs to the innovative trial cities, its value is set to 1. Otherwise, it is 0. u i and λ t are the city-fixed and year-fixed effects, respectively. ε it is the random disturbance term, and Control it is the set of control variables. Variables (1) Dependent Variable In the empirical analysis, the Gini coefficient is employed as a proxy to assess income inequality. According to Fang and Meng (2024), the Gini coefficients for each city were calculated. The detailed calculation of the Gini coefficient for city i in year t is presented as follows: Giniit =∑ ni k=1∑ ni r=1 |Likt −Lirt| 2n2 iLit (7) where ni is the total number of urban and township units in the i prefecture level or above the city. Lit is the average nighttime light intensity of the i prefecture level or above the city in year t. Likt is the nighttime light intensity of urban or township unit within city i during the year t. Lirt is the nighttime light intensity of the r urban or township unit within city i during the year t. (2) Core Independent Variable A binary variable is created to mirror the pilot policy of innovative cities, considering the time and range of its implementation. This variable is denoted as DID and is derived by multiplying the variable of time with treat. When a city is selected as an innovative trial city, it is assigned to the treatment group in which the value of treat is set to 1. In cases where it is not, treat is set to 0. In case a city is identified as innovative in a certain year, the value of time is set to 1 from that year onwards. For all years before the designation and for cities not identified as innovative, the value is set at 0. The treatment and control groups comprise 97 and 179 innovative pilot and non-pilot cities, respectively. (3) Control Variables Referencing the existing body of literature, the subsequent control variables were chosen. The economic development level (avgdp) is measured by the inflation-adjusted per capita real gross domestic product (GDP) of the city. This GDP was converted to constant 2003 prices. Subsequently, the resulting value of per capita real GDP is logtransformed. Government intervention (gov) is calculated as the ratio of the budgetary spending of the local government in relation to the regional GDP. Average wage level (income) is the log-transformed average wage of urban employees. Fixed asset investment (asset) is the log-transformed total fixed asset investment. Degree of trade openness (open) is calculated as the percentage of the combined value of imports and exports representative of GDP. Population density (popm) is measured as the population per square kilometre. Data sources and descriptive statistics Cities with substantial data gaps were excluded to maintain data completeness. The dataset covers 276 Chinese urban areas from 2003 to 2022. Among them, 97 were identified as innovation pilot cities, and the remaining 179 cities acted as non-pilot counterparts. The data are primarily from various editions of the China Urban Statistical Yearbook and statistical yearbooks for provinces, cities and counties. Data regarding the innovative trial cities were sourced from the document titled ‘Guidelines for Establishing Innovative Cities’. Table 1 presents a synopsis of the descriptive statistical figures for the key variables. Empirical analysis Regression results Table 2 presents the results of the impacts of the innovative city trial initiative on income disparity. Column 1 presents the outcomes without considering other factors into account. The regression analysis indicates a statistically significant negative coefficient for the policy intervention dummy (−0.0145), indicating that the innovative city pilot initiative effectively mitigates income inequality. Column 2 presents control variables but only accounts for city-fixed effects. The policy-related coefficient is −0.0261, though subject to slight modifications, and consistently retains its significantly negative value across different model specifications. These findings indicate that policy interventions can reduce economic disparity and promote shared prosperity. This effectiveness remains even when unobserved heterogeneity at the city level is considered through fixed-effects estimation. To mitigate potential confounding biases, the specification presented in column 3 incorporates time and city-fixed effects to assess the causal relationship between the policy intervention and income distribution outcomes. The empirical analysis reveals a policy coefficient estimate of −0.0101 for the intervention indicator variable. While there is some fluctuation compared with the results without the control variables, the negative effect, statistically significant at the 1% significance level, remains evident. Assuming other factors are held constant, the policy causes a reduction of ~1.01% in the average Gini coefficient of pilot cities in comparison with the coefficient of non-pilot cities. The results indicate that the pilot initiative for innovative cities, serving as a Table 1 Definitions of variables. Variable Symbols Obs Mean Std. dev. Min Max Gini coefficient Gini 5520 0.7481 0.1982 0.0388 0.9952 Policy dummy variable DID 5520 0.1486 0.3557 0 1 Economic development avgdp 5520 6.7109 1.0606 3.2122 10.2859 Government intervention gov 5520 0.1717 0.0945 0.0313 1.4852 Average wage level income 5520 10.5261 0.6947 2.2834 12.6780 Fixed asset investments asset 5520 15.7184 1.3228 10.2518 19.0834 Trade openness open 5520 0.2127 0.4026 0.0004 7.6201 Population density popm 5520 5.8047 0.9593 1.5476 9.2350 X. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 5 cornerstone of the innovation-led growth strategy in China, significantly enhances income distribution equality. Thus, Hypothesis 1 is confirmed. Parallel trend test Satisfying the assumption regarding parallel trends is a requirement for the multi-period DID. If this condition is violated, the estimated coefficients cannot accurately reflect the policy effect. Owing to the phased rollout of the intervention, the composition of city groups changes in each phase. To overcome this empirical issue, this research employed the event study methodology, following the precedent set by Jacobson et al. (1993). This methodological framework simultaneously validates the parallel trend hypothesis and assesses the temporal evolution of policy impacts. Considering the post-policy sample size, the time variable ranges from −6 (6 years pre-policy) to 4 (4 years post-policy). As shown in Fig. 1, in the period before the innovative city programme took effect, the intervention and control groups had no significant trend differences. This evidence indicates the satisfaction of the parallel trend assumption. The post-implementation period reveals statistically significant negative coefficients for income inequality. This finding reveals a significant difference between cities participating in the pilot scheme and those not participating. This difference verifies that the policy measures are effective in reducing income inequality. In conclusion, the observed income inequality reduction is not attributed to pre-policy trends. Dynamic effect analysis As shown in Fig. 1, the dynamic effects of policy influence are discernible. Notably, after the third period, there is a minor decrease in the positive influence that innovative city buildings have on income inequality. In essence, the efficacy of innovation policies in reducing income inequality first rises and then falls. This phenomenon can be comprehensively interpreted from two key perspectives: the theory of diminishing marginal effects and external shocks. In line with the law of diminishing marginal effects, at the onset of innovation policy implementation, two aspects contribute to the initial impact of the policy. First, R&D subsidies and patent incentives fuel high-skilled industry growth, drive economic expansion and generate numerous high-paying jobs. In addition, these policies upgrade traditional industries, boosting low-skilled workers’ incomes and narrowing the income gap. Second, the intensifying agglomeration of innovation resources draws talents, funds and technologies to cities. This generates entrepreneurial and investment opportunities, diversifies income sources for various groups and particularly elevates the earnings of active innovators. However, with the continuous injection of policy resources, saturation of key innovative elements (such as talent and capital) occurs. Concurrently, policy implementation costs are rising, and the ability to attract talent is weakening. These two factors combine to undermine the efficacy of a policy in narrowing income gaps, eventually leading to diminished policy outcomes. This study takes action to verify the compliance of the marginal effects of innovation policies with the law of diminishing returns. The model uses a dynamic panel model and a generalised method of moments estimation to analyse the impact of innovation policy intensity on income inequality. Herein, the intensity of innovation policies is indicated by the ratio of R&D expenditure relative to GDP (RD_ratio). Column 1 in Table 3 presents that the coefficient for the R&D expenditure proportion is significantly positive (−0.601), whereas the coefficient for its squared term is significantly negative (0.7069). This indicates a Ushaped relationship between innovation policy intensity and income inequality. The correctness of the theoretical hypothesis 1 was verified. In addition, a dummy variable DT is constructed for the three-period Table 2 Impact of the innovative city pilot policy on income inequality. Variables (1) (2) (3) DID −0.0145** −0.0261*** −0.0101*** (0.0056) (0.0050) (0.0030) avgdp −0.0709*** −0.0125*   (0.0081) (0.0068) open 0.0073 0.0295***   (0.0055) (0.0034) gov 0.2658*** 0.0199   (0.0264) (0.0165) income −0.0240*** 0.0053   (0.0058) (0.0040) asset 0.0197*** −0.0005   (0.0028) (0.0018) popm −0.0100*** 0.0087***   (0.0030) (0.0018) Time FE Yes No Yes City FE Yes Yes Yes R 2 0.704 0.149 0.710 N5520 5520 5520 Note: *, ** and *** represent statistical significance at the 10%, 5% and 1% levels, respectively, with t values based on city-level clustering provided in parentheses. The same applies to the following tables. Fig 1. Parallel trend test. X. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 6 time frame around policy implementation. A value of 1 is assigned after the third-period policy implementation, and 0 is assigned before it. The implementation strength of talent policies is measured by extracting talent-related terms from government work reports. These reports cover policy evaluations, effectiveness and future-oriented intensity. Column 2 in Table 3 presents the impact of the innovation policy dummy variable DID and its interaction term with DT on talent-introduction intensity. The analysis indicates that local policies have enhanced the intensity of talent introduction since the innovation policy took effect. However, this intensity wanes after the third implementation phase, further diminishing the positive influence of the talent agglomeration effect on income inequality. The stable implementation of innovation policies can be disrupted by external events, causing variations in their influence on income inequality. Among the various external shock factors, the home-return entrepreneurship policy is a quintessential example. Piloted in certain regions in 2016 and 2017, this policy may disrupt the labour force aggregation pattern following the implementation of urban innovation policies. A dummy variable FX is constructed by this paper for the policy of returning to hometown for entrepreneurship. That is to say, when a county/district in a city enforces the policy, it gets a value of 1; otherwise, it gets 0. As shown in column 3, the coefficient of DID#FX is significantly positive with the inclusion of the effects of the homereturning entrepreneurship policy. This indicates that the positive impact of the innovative city pilot policy on income inequality has declined. The decline is particularly notable after the shock of the homereturning entrepreneurship policy. As stated otherwise, the policy of returning to one’s hometown to start a business offsets some of the effectiveness of the innovation policy. Consequently, it has increased the difficulty of narrowing the income gap. Robustness tests Results from the baseline regression verify that a city’s inclusion in the innovative pilot programme notably cuts down income inequality. To ensure that the conclusions are not influenced by confounding factors, a series of robustness tests were conducted. These tests address various dimensions, such as sample selection, exclusion of other policy interferences, PSM-DID analysis, non-random sample selection and instrumental variable regression. (1) Sample Data Filtering To address the impact of extreme values, numerical variables were winsorised at the first and fifth percentiles, and the model was reevaluated. In addition, certain special years in the sample may have impacted the accuracy of the results, prompting their exclusion from the analysis. For instance, the 2008 worldwide financial turmoil caused a marked decrease in the import and export activities within China. Despite the introduction of economic stimulus policies globally, the crisis resulted in certain challenges such as financing difficulties, rising costs, increased unemployment and reduced wages. Similarly, the COVID-19 outbreak in 2020 caused widespread city and business shutdowns, further disrupting economic activity. To eliminate the influence of these special years, data from 2008 and 2020 were excluded, and the model was re-estimated using the remaining sample. As shown in columns 1, 2 and 3 in Table 4, the coefficients of the DID policy are notably negative at the 10% significance level (−0.0078, −0.0073 and −0.0060, respectively). The consistency of these results with previous research bolsters the soundness of the findings. (2) Excluding the Influence of Other Policies To precisely evaluate the influence of the innovative city pilot programme on income inequality, excluding the impacts of other policy measures is crucial. Through a review of relevant literature and policy documents, the study identified smart city policy as a potential confounding factor during the sample period. The policy has facilitated the advancement of intelligent technologies, which are used to reinforce urban infrastructure and promote economic development. In addition, the policy has had a beneficial impact on urban innovation. This may, in turn, influence income inequality. The smart city policy was implemented in three phases starting in 2010, and it coincided temporally with the innovative city pilot policy. To test for robustness, an indicative binary variable, SMA, for the smart city initiative was included. In particular, the SMA variable is coded as 1 for cities participating in the smart city initiative and as 0 for those that do not. Column 4 in Table 4 presents the relevant results. The results indicate that when the SMA variable is included, the sign of the coefficient related to the DID policy variable (−0.0077) does not change. In addition, the DID policy variable remained statistically significant. This underscores that the results are independent of the smart city policy or other potential confounding factors. Moreover, the coefficient associated with the innovative city policy remained stable, further strengthening the reliability of the study’s findings. (3) PSM-DID Analysis Owing to the relatively large size of China, cities significantly vary in terms of economic development and policy enforcement. The treatment and control groups may exhibit distinct characteristics. Moreover, biases could arise from sample selection, reverse causality or other sources of endogeneity. To address these challenges, this study used the PSM-DID method for validation. Table 5 presents the results of the process employing radius matching, kernel matching and nearest-neighbour matching techniques. The obtained coefficients were significant under all three methods. The results (−0.0108, −0.0080 and −0.0078) affirm that the pilot policy for innovative cities effectively mitigates income inequality. This confirmation strengthens the robustness of the conclusions. Table 3 The intensity of talent introduction and home-returning innovation policy. Variables (1) Gini (2) The intensity of talent introduction (3) Gini RD_ratio −0.6010**   (0.2625)   RD_ratio2 0.7069**   (0.3418)   DID 0.0048** −0.0171***   (0.0022) (0.0038) DID#DT −0.0071***    (0.0023)  DID#FX   0.0152***    (0.0053) Controls Yes Yes Yes N5520 5520 5520 R 2 0.277 0.710 Wald test 1543.06   Table 4 Robustness tests. Variable (1) 1% Winsorisation (2) 5% Winsorisation (3) Excluding special years (4) Excluding other policies DID −0.0078** −0.0073** −0.0060* −0.0077** (0.0031) (0.0031) (0.0034) (0.0031) SMA    0.0025     (0.0020) Controls Yes Yes Yes Yes Time FE Yes Yes Yes Yes City FE Yes Yes Yes Yes R 2 0.707 0.704 0.719 0.711 N5520 5520 4968 5520 X. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 7 (4) Non-random Sample Selection When selecting the list of pilot cities for innovation, governments often consider specific attributes such as geographic location and economic development level. Over time, these attributes may have differential impacts on income inequality. Thus, it is crucial to consider and control these factors to maintain the robustness of the results. The DID method adopted in this study assumes a quasi-natural experiment in which the treatment and control groups are ideally randomly selected. The actual selection of innovative pilot cities, however, is influenced by various factors such as economic development, geographic location and social conditions, which are not entirely random. This study aimed to address potential biases from ‘non-random’ selection and reduce their impact. To achieve this, it incorporates interaction terms between baseline characteristics and linear time trends into the baseline regression model (1). The updated model is expressed as follows: Giniit =γ0+γ1DIDit +γ2∑Controlit +γ3∑Dumc×trendt+ui+λt + ε it (8) In this model, Dum c represents a set of dummy variables that capture specific city characteristics. These entail aspects such as whether the city belongs to the Yangtze River Economic Belt (Dum 1 ), whether it holds the status of a municipality directly under the central government (Dum 2 ) and whether it is designated as a special economic zone (Dum 3 ). trend t denotes the time trend term. Other variable definitions are consistent with those in previous sections. Table 6 presents the results. Columns 1–3 individually incorporate each interaction term, and column 4 simultaneously includes all three interaction terms. The DID coefficient consistently and significantly remains negative across all specifications, with values of −0.0066, −0.0066, −0.0080 and −0.0054. This confirms that the innovative city pilot policy significantly influences the reduction of income inequality. Moreover, the results indicate that, while certain city-specific factors were considered during the selection of pilot cities, the process retains some degree of randomness. (5) Instrumental Variable Regression Determining innovative pilot cities is not based on the principle of randomness. Instead, it takes into comprehensive consideration various factors such as the regional positioning, innovation capabilities and economic development levels of the cities. Owing to the non-random selection, the treatment group is highly likely to be interfered with by policy endogeneity. There may also be potential endogeneity issues with the policy variable, ultimately leading to deviations in the research results. This study employed the instrumental variable method for estimation to address the interference of endogeneity issues in research results. It designated National Historical and Cultural Cities as the instrumental variable for policy. This choice is well-founded. On the one hand, there is a similarity in economic status between innovative cities and National Historical and Cultural Cities because they are economic centres. Innovative cities aim to create innovative centres with strong radiating and driving effects and are key forces in promoting the economic development of modern society. National Historical and Cultural Cities were mostly important economic and political areas in history and served as the economic core regions in ancient society. Therefore, there is a strong correlation between them. On the other hand, National Historical and Cultural Cities cannot directly affect the income inequality of current cities. These can only exert their influence through the establishment of innovative cities. In this way, they meet the ‘exclusion restriction’ and conform to the requirement of exogeneity. As shown in Table 7, the instrumental and policy variables are significantly and positively correlated. In addition, the relevant test results demonstrate that the instrumental variable meets the weak identification requirement. The estimated coefficient of DID remains notably negative. This implies that even potential endogeneity issues are further considered. It can still be concluded that this innovative city pilot policy will reduce income inequality. Heterogeneity analysis Heterogeneity by geographical location Municipalities located in various geographical regions exhibit substantial disparities in terms of economic development levels and approaches. These differences might have an impact on the execution and outcomes of the innovative city pilot initiative. Therefore, it is crucial to analyse whether the impact of a policy on income inequality varies by geographical location. According to the classification in the ‘China National Economic and Social Development Statistics Bulletin’, cities are categorised into two regions: eastern and central western. This study further classifies cities based on their geographical locations, using the ‘Hu Huanyong Line’ and ‘the Yangtze River Economic Belt’ as boundaries. Moreover, this study examined the differences between border and non-border cities, as well as between cities along the Yangtze River and those in other regions. The findings of this study presented in Table 8 indicate that the innovative city pilot policy exerts a negative yet statistically insignificant influence in the eastern region. Conversely, in the central and western regions, its impact is remarkably negative, with a coefficient of −0.0081. Such a difference might be attributed to the distinct levels of economic development present in these different regions. By contrast to Table 5 PSM-DID test. Variable (1) (2) (3) Radius Kernel Nearest-neighbour DID −0.0108** −0.0080*** −0.0078** (−2.5656) (−2.6213) (−2.5398) Controls Yes Yes Yes Time FE Yes Yes Yes City FE Yes Yes Yes N 3463 5474 5520 R 2 0.7391 0.7126 0.7105 Table 6 Analysis of non-random sample selection. Variable (1) (2) (3) (4) DID −0.0066** −0.0066** −0.0080*** −0.0054* (0.0030) (0.0031) (0.0030) (0.0030) Controls Yes Yes Yes Yes Time FE Yes Yes Yes Yes City FE Yes Yes Yes Yes R 2 0.717 0.715 0.719 0.731 N5520 5520 5520 5520 Table 7 Instrumental variable regression. Variable (1) The first stage (2) The second stage DID −0.0113**   (0.0043) IV 0.8638**  (0.0119)  Controls Yes Yes Time FE Yes Yes City FE Yes Yes R 2 0.810 0.943 N5520 5520 F-value in the first stage 10782.27  Kleibergen-Paap LM statistic 726.06 (P value =0.0000) Kleibergen-Paap Wald statistic 10782.27 (P value =0.0000) X. Li et al. Journal of Innovation & Knowledge 10 (2025) 100749 8