How Does R&D Investment Persistence Boost SRUN Firms’ Growth Quality? A Mediation Analysis

Abstract

Specialized, Refined, Unique and Novel (SRUN) listed firms are pivotal to the high-quality development of China’s real economy, and their growth quality underpins the security of industrial and supply chains. This study empirically examines the relationship between R&D investment persistence and growth quality of Chinese A-share SRUN listed firms from 2006 to 2024, with technology conversion efficiency as the mediating variable. R&D investment persistence is measured from the dual dimensions of investment intensity and stability, and firm growth quality is a comprehensive indicator constructed via principal component analysis (PCA) from revenue growth, profitability and risk resilience. Panel data regression models, combined with mechanism, endogeneity, robustness and heterogeneity tests, are adopted for empirical analysis. The results show a significantly positive correlation between R&D investment persistence and SRUN firms’ growth quality, with the regression coefficient of R&D investment persistence on growth quality reaching 0.189 (p < 0.01); both investment intensity and stability exert significant positive effects on all dimensions of growth quality, with their regression coefficients on growth quality being 0.156 and 0.132 (both p < 0.01) respectively. Technology conversion efficiency plays a partial mediating role in this relationship, with the mediating effect ratio of R&D investment persistence on growth quality through technology conversion efficiency at 34.2%, as R&D investment persistence indirectly improves growth quality by enhancing patent output and new product conversion efficiency. Heterogeneity analysis indicates that this positive correlation is more pronounced in high-tech industries, small and medium-sized enterprises (SMEs) and eastern China-based firms, driven by differences in industrial R&D dependence, resource endowments and financing frictions. Though endogeneity is mitigated by instrumental variables, propensity score matching (PSM) and difference-in-differences (DID), strict causal identification is constrained by data availability. This study enriches the theories of R&D investment and firm growth, and provides empirical insights for SRUN firms to optimize their R&D strategies and for the government to formulate targeted support policies, so as to promote the high-quality development of SRUN firms and the transformation of China’s manufacturing industry.

1. Introduction

Amid the intensifying global industrial competition and under the national strategy of pursuing self-reliance and self-improvement in science and technology, SRUN (Specialized, Refined, Unique and Novel) enterprises have emerged as the core driving force for the transformation and upgrading of China’s manufacturing sector. Bolstered by their strengths in specialization, sophistication, differentiation and innovation, they have become pivotal players in breaking through core technology bottlenecks and advancing the independence and controllability of industrial and supply chains [1]. By the end of 2024, China had nurtured more than 1500 listed SRUN enterprises, according to the official statistics from the Small and Medium Enterprises Bureau of the Ministry of Industry and Information Technology and the Wind Database, spanning strategic emerging industries such as high-end manufacturing, new materials and biomedicine. The growth quality of these enterprises is directly tied to the overall high-quality development of the manufacturing industry [2]. Notably, the distribution mechanism of public innovation support exerts a profound impact on the R&D investment behavior of SRUN enterprises: the current clustered allocation of R&D subsidies in some regions may lead enterprises to adopt short-term, utilitarian R&D practices, while the lack of long-term, stable policy support—compounded by financing frictions plaguing small and medium-sized SRUN enterprises [3,4]—hinders the formation of sustained R&D investment. This mismatch between policy support patterns and the long-term cumulative nature of R&D activities not only undermines the technological iteration pace of SRUN enterprises but also restricts the translation of R&D input into growth quality, making it crucial to explore how R&D investment persistence (rather than one-off intensity) shapes their high-quality development.

R&D investment is the core driver for SRUN enterprises to sustain technological advantages and achieve sustainable growth [5]. Given the long-term and cumulative nature of R&D activities, sustained investment is far more conducive to forging irreplicable core competitiveness than one-off investment. Unlike conventional enterprises, SRUN enterprises focus on niche markets and rely more heavily on core technologies; as such, the stability and intensity of their R&D investment directly determine their pace of technological iteration, product competitiveness and risk resilience. However, many SRUN enterprises are currently grappling with prominent challenges, including severe volatility in R&D investment, the prevalence of short-term utilitarian investment behaviors and low efficiency in technology commercialization. These issues impede the effective translation of R&D investment into growth momentum, thus hindering the high-quality development of SRUN enterprises [6].

Existing research has largely centered on the impact of R&D investment intensity on firm performance [7]. In recent years, relevant literature has expanded its scope to explore R&D investment persistence [8,9], yet most such studies focus on traditional enterprises in developed countries, with insufficient attention paid to SRUN enterprises in emerging markets. Moreover, the internal mechanism through which R&D investment persistence influences the growth quality of SRUN enterprises remains unclear—particularly the mediating role of technology conversion efficiency in linking R&D input and growth outcomes—and its heterogeneous characteristics across industry, firm size and regional contexts call for further in-depth exploration [10]. Against this backdrop, this study takes Chinese A-share listed SRUN enterprises from 2006 to 2024 as research samples to systematically examine the relationship between R&D investment persistence and firm growth quality. It introduces technology conversion efficiency as a mediating variable to probe the transmission channel between the two and analyzes the heterogeneous effects across dimensions, including industry type, firm size, regional development level, technological capability and the direction of R&D shocks. Drawing on the synergistic insights of Schumpeter’s Innovation Theory [11], the Resource-Based View [12] and the Resource Dependence Theory [13], this study constructs a theoretical framework of “R&D investment persistence → technology conversion efficiency → SRUN enterprise growth quality”, which lays a solid theoretical foundation for the research hypotheses.

This study makes three key marginal contributions: First, it deconstructs R&D investment persistence into dual dimensions of intensity and stability, moving beyond the single-dimensional focus on R&D intensity or time-series continuity in existing research, and empirically verifies its comprehensive impact on the multi-dimensional growth quality (revenue growth, profitability, risk resilience) of SRUN enterprises. Second, it clarifies the partial mediating role of technology conversion efficiency (integrating patent output and new product conversion) and quantifies its mediating effect ratio (34.2% for overall R&D persistence), filling the gap in the internal transmission mechanism between R&D persistence and firm growth quality. Third, it systematically identifies heterogeneous effects across high-tech vs. non-high-tech industries, SMEs vs. large enterprises, and eastern vs. non-eastern regions, and further verifies the financing friction channel underlying SME heterogeneity, providing targeted empirical evidence for differentiated policy design. Ultimately, this study seeks to provide theoretical support and practical references for SRUN enterprises to optimize their R&D strategies and for the government to formulate targeted support policies, thereby advancing the high-quality development of SRUN enterprises and the transformation of China’s manufacturing industry.

2. Literature Review and Research Hypotheses

2.1. Literature Review

The relationship between R&D investment and firm growth has long been a focal research topic in academia. Grounded in Schumpeter’s Innovation Theory, early studies have identified technological innovation as the core driver of enterprises’ sustainable growth, with R&D investment serving as the material underpinning for such innovation [14,15]. In particular, sustained and stable R&D investment can accelerate technological iteration and help enterprises build core competitive advantages [16]. As research in this field has advanced, subsequent scholars have extended the research scope and found that the relationship between R&D investment and firm growth is not simply linear [17]; instead, factors including R&D investment intensity, persistence and structure all modulate the degree to which R&D investment drives firm growth [7]. Existing studies even present conflicting conclusions: some argue that R&D investment persistence can enhance innovation performance through knowledge accumulation and technological accumulation, thereby boosting firm scale expansion and profitability improvement; others, however, contend that excessively high R&D investment persistence may lead to resource misallocation and elevated operational risks, thus exerting a negative impact on firm growth quality [18,19].

With the in-depth exploration of the R&D investment-firm growth nexus, the measurement methods and connotations of R&D investment persistence have been continuously enriched in existing research [20,21,22]. Brown & Petersen (2011) defined R&D smoothing as enterprises’ behavior of maintaining stable R&D investment through internal and external financing under financing constraints, with a specific focus on stabilizing R&D investment in the face of external shocks [20]. Czarnitzki & Hottenrott (2011) measured R&D persistence from a time-series continuity perspective, taking the probability of enterprises continuing R&D investment in subsequent periods as the core indicator [21]. Building on this, Kang et al. (2017) further expanded the connotation of R&D persistence by integrating R&D investment intensity and technological capability, noting that R&D persistence is reflected not only in the stability of investment scale but also in the matching degree between investment and an enterprise’s technological capability [23]. Nevertheless, a notable research gap remains in existing studies: most focus on R&D smoothing under financing constraints or the time-series continuity of R&D investment, while few decompose R&D persistence into the dual dimensions of investment intensity and stability to explore its impact on firm growth quality.

To address this research gap, this study defines R&D investment persistence as the long-term adequacy and time-series stability of R&D investment, and measures it from the dual dimensions of investment intensity and investment stability. This measurement approach differs from the “R&D smoothing” proposed by Brown & Petersen (2011)—which centers on stabilizing R&D investment under financing constraints—and the “time-series continuity” proposed by Czarnitzki & Hottenrott (2011) [20,21]. It takes into account both the “scale dimension” of R&D investment (i.e., intensity) and the “temporal stability” of R&D investment (i.e., stability), which is more aligned with the characteristics of SRUN enterprises—enterprises that focus on segmented markets and rely on the iteration of core technologies. Furthermore, drawing on the research of Ipinnaiye et al. (2025), this study further decomposes R&D persistence and combines it with the growth characteristics of SRUN enterprises, thereby enriching the measurement system of R&D investment persistence [22].

In empirical research specific to SRUN enterprises, Ge et al. (2025) found that R&D investment intensity is significantly and positively correlated with the growth quality of SRUN enterprises, and this correlation is more pronounced in high-tech industries [24]. In addition, Liu et al. (2015) pointed out that R&D investment persistence is more effective than one-off R&D investment in enhancing firm innovation performance, a conclusion that is particularly applicable to SRUN enterprises focused on niche markets [25]. Despite these valuable findings, existing studies still exhibit four distinct limitations: first, most focus on R&D investment intensity, with research on R&D persistence being scattered, and its heterogeneous impacts on the various dimensions of growth quality remaining unclear; second, the internal transmission mechanism through which R&D investment persistence affects firm growth quality has not been thoroughly explored, and the mediating role of technology conversion efficiency has not been fully verified; third, the analysis of heterogeneous characteristics is inadequate, with no in-depth research conducted by integrating the industrial, scale and regional differences in SRUN enterprises; fourth, the concept of R&D persistence has not been benchmarked against classic existing studies, and the heterogeneous impacts of technological capability and positive/negative R&D shocks on the relationship between R&D investment persistence and firm growth quality have not been considered.

Drawing on relevant theories and existing literature, this study constructs an integrated analytical framework based on Schumpeter’s Innovation Theory, the Resource-Based View (RBV) and the Resource Dependence Theory (RDT), with each theory specifically supporting the dual dimensions of R&D investment persistence and forming a synergistic mechanism aligned with SRUN enterprises’ growth characteristics:

Resource-Based View (RBV) [12]: This theory directly underpins the investment intensity dimension of R&D persistence. RBV emphasizes that heterogeneous and inimitable resources are the source of enterprises’ sustainable competitive advantages [12]. For SRUN enterprises focusing on segmented markets, high-intensity R&D investment ensures sufficient input of human, material and financial resources, which are transformed into unique technological knowledge and core technical capabilities. These heterogeneous resources enable enterprises to develop differentiated products, enhance market pricing power and improve profitability, forming a solid resource foundation for the improvement of growth quality. In other words, the adequacy of R&D investment (intensity) is the key to building heterogeneous technical resources, which is exactly the core logic of RBV in supporting the relationship between R&D persistence and firm growth quality.

Schumpeter’s Innovation Theory [11]: This theory provides core support for the investment stability dimension of R&D persistence. Schumpeter’s Innovation Theory holds that sustained technological innovation is the fundamental driving force for breaking market equilibrium and achieving excess profits [11]. R&D activities are inherently long-term, cumulative and path-dependent. Stable R&D investment ensures the continuity of technological research, avoids the interruption of knowledge accumulation and project development, and enables enterprises to continuously promote product iteration and process innovation. For SRUN enterprises relying on core technology iteration, stable investment maintains the continuity of technological innovation, helps them seize market opportunities in niche segments, and realizes sustained revenue growth, which is consistent with the core implication of Schumpeter’s Innovation Theory.

Resource Dependence Theory (RDT) [13]: This theory also supplements the investment stability dimension by focusing on enterprises’ ability to reduce external dependence through resource accumulation [13]. Stable R&D investment enables SRUN enterprises to continuously accumulate internal technical resources and build independent core technology systems, thereby reducing their reliance on external technologies, core components and supply chains. This reduction in external dependence enhances enterprises’ ability to withstand industrial fluctuations and market risks, improves financial stability and risk resilience, and ensures the stability of growth quality—this exactly responds to the core logic of RDT that enterprises can enhance their autonomy and anti-risk capabilities by controlling key resources.

The synergistic effect of the three theories forms a complete theoretical support system for the research framework: RBV guarantees the “resource foundation” through the intensity dimension, while Schumpeter’s Innovation Theory and RDT ensure the “continuous driving force” and “risk resilience” through the stability dimension. Together, they provide a solid theoretical basis for the positive correlation between R&D investment persistence (intensity + stability) and the growth quality of SRUN enterprises. As a key link connecting R&D investment and firm growth, technology conversion efficiency is hypothesized to play a mediating role in this relationship—converting the resource advantages formed by R&D intensity and the continuous innovation capacity formed by R&D stability into tangible market performance (revenue growth, profitability, risk resilience).

2.2. Hypotheses on the Positive Correlation Between R&D Investment Persistence and the Growth Quality of Chinese A-Share SRUN Listed Companies

R&D investment persistence consists of two core dimensions: investment intensity and investment stability, which jointly drive the improvement of firm growth quality. From the perspective of investment intensity, SRUN enterprises focus on niche markets, and breakthroughs in core technologies depend on long-term and sufficient R&D investment. A relatively high level of R&D investment intensity enables enterprises to attract high-caliber R&D talent, procure advanced R&D equipment, and conduct in-depth research on core technologies, thereby boosting product upgrading and process optimization, and elevating product market share and profitability [26]. Unlike conventional enterprises, the R&D investment intensity of SRUN enterprises directly shapes their technological leadership; sustaining high investment intensity prevents technological obsolescence amid industrial evolution and consolidates their market position in niche segments, thus enhancing overall growth quality.

From the perspective of investment stability, R&D activities are inherently long-term, cumulative and high-risk in nature. Volatile short-term R&D investment disrupts the continuity of technological research, leading to fractured knowledge accumulation, abandoned R&D projects and heightened risks of R&D failure. In contrast, stable R&D investment ensures the uninterrupted progress of R&D projects, fosters the gradual accumulation of technological knowledge and improves R&D efficiency. It also sends a clear market signal of an enterprise’s commitment to innovation, boosting the confidence of investors and partners and helping enterprises secure more resource support, which in turn drives improvements in revenue growth, profitability and risk resilience [27].

Firm growth quality is a comprehensive concept encompassing three core dimensions: revenue growth rate, profitability and risk resilience [28]. R&D investment persistence is positively correlated with all three dimensions. In terms of revenue growth, sustained R&D investment spurs the development of new products and the upgrading of existing ones, meets the diversified and high-quality demands of the market, expands market share and thus fuels steady revenue growth. In terms of profitability, technological advantages forged through sustained R&D cut production costs, raise product added value and create differentiated competitive edges, enabling enterprises to capture excess profits and lift profitability indicators such as return on net assets. In terms of risk resilience, sustained R&D drives technological diversification for enterprises, reduces over-reliance on a single product or technology, enhances the ability to respond to market changes and industrial competition, and bolsters financial stability by lowering the asset-liability ratio. Based on the above analysis, the following research hypotheses are proposed:

H1: 

R&D investment persistence has a significantly positive correlation with the growth quality of Chinese A-share SRUN listed companies.

Drawing on the synergistic insights of Schumpeter’s Innovation Theory, the Resource-Based View (RBV) and the Resource Dependence Theory (RDT), R&D investment persistence provides SRUN enterprises with the dual safeguards of “resource input + knowledge accumulation” through its two dimensions. On the one hand, R&D investment intensity cultivates heterogeneous technical resources (rooted in the RBV), which serve as the core foundation for enterprises to enhance market competitiveness and profitability. On the other hand, R&D investment stability enables the continuous accumulation of knowledge and technology (based on Schumpeter’s Innovation Theory) and reduces dependence on external technologies (in line with the RDT), thereby strengthening enterprises’ risk resilience. These two dimensions together drive the improvement of enterprises’ comprehensive growth quality across the dimensions of revenue growth, profitability and risk resilience.

H1a: 

R&D investment intensity has a significantly positive correlation with the growth quality of Chinese A-share SRUN listed companies.

Based on the RBV, R&D investment intensity guarantees sufficient human, material and financial input into SRUN enterprises’ R&D activities, laying the material foundation for core technology research and product innovation. Adequate R&D resource input helps enterprises build product differentiation advantages in segmented markets, increase market share and product added value and thus comprehensively improve growth quality in terms of revenue growth, profitability and risk resilience.

H1b: 

R&D investment stability has a significantly positive correlation with the growth quality of Chinese A-share SRUN listed companies.

Based on Schumpeter’s Innovation Theory and the RDT, R&D investment stability avoids the interruption of R&D projects and the loss of accumulated knowledge caused by investment fluctuations, facilitating the continuous accumulation of technological knowledge and the improvement of R&D efficiency. At the same time, stable R&D investment reduces enterprises’ reliance on external technologies, enhances their ability to withstand market risks, and releases positive market signals that help enterprises secure more resource support, thereby promoting the improvement of growth quality [29].

H1c: 

R&D investment intensity has a significantly positive correlation with the revenue growth rate of Chinese A-share SRUN listed companies.

H1d: 

R&D investment intensity has a significantly positive correlation with the profitability of Chinese A-share SRUN listed companies.

H1e: 

R&D investment intensity has a significantly positive correlation with the risk resilience of Chinese A-share SRUN listed companies.

H1f: 

R&D investment stability has a significantly positive correlation with the revenue growth rate of Chinese A-share SRUN listed companies.

H1g: 

R&D investment stability has a significantly positive correlation with the profitability of Chinese A-share SRUN listed companies.

H1h: 

R&D investment stability has a significantly positive correlation with the risk resilience of Chinese A-share SRUN listed companies.

2.3. Hypotheses on the Mediating Role of Technology Conversion Efficiency

The relationship between R&D investment persistence and firm growth quality is not direct, but is instead mediated by technology conversion efficiency [30,31]. Sustained R&D investment lays a solid material foundation and builds up technological reserves for technology conversion, yet R&D investment in itself does not translate into firm growth. Only when R&D achievements are effectively converted into practical productive forces—such as patents and new products—can the value of R&D investment be fully realized and enterprise growth quality be enhanced.

R&D investment persistence boosts technology conversion efficiency in two key ways. First, high R&D investment intensity ensures the depth and breadth of R&D projects, elevates the quality of R&D achievements, provides high-quality technical resources for conversion, and supports the pilot testing and industrialization of these achievements, thereby cutting conversion costs and mitigating related risks. Second, stable R&D investment strengthens the stability and professionalism of R&D teams, fosters a mature technology conversion process and enhances cross-departmental collaboration between R&D, production and marketing. This ensures that R&D achievements are aligned with market demand and further improves conversion efficiency.

Enhanced technology conversion efficiency is positively linked to the improvement of firm growth quality, and this manifests in three aspects: (1) It accelerates patent output, consolidates enterprises’ intellectual property advantages and technical barriers, safeguards their market positions, and creates additional revenue streams through patent transfer and licensing, thus boosting profitability; (2) It drives the rapid translation of R&D achievements into new products, enriches enterprises’ product portfolios, meets evolving market demand, expands market share and fuels revenue growth; (3) It converts technological advantages into market competitive edges, reduces enterprises’ reliance on external technologies, and strengthens their ability to respond to industrial competition and market fluctuations, thereby enhancing risk resilience.

For SRUN enterprises, technology conversion efficiency is the critical link connecting R&D investment and growth quality—only efficient technology conversion can turn sustained R&D investment into tangible growth momentum for enterprises. Based on this analysis, the following research hypotheses are proposed:

H2: 

Technology conversion efficiency plays a mediating role between R&D investment persistence and the growth quality of Chinese A-share SRUN listed companies.

Rooted in the innovation value chain theory, the innovation activities of SRUN enterprises form a complete value chain: R&D input → technological achievement output → market value conversion. R&D investment persistence serves as the starting point, providing ample resources and stable accumulation for the generation of technological achievements; technology conversion efficiency is the core link, realizing the transformation from technological achievements to market performance; and the improvement of firm growth quality is the ultimate end, embodying the actual realization of innovation value. In this way, R&D investment persistence indirectly drives the improvement of SRUN enterprises’ growth quality by enhancing technology conversion efficiency, forming the transmission path: R&D investment persistence → technology conversion efficiency → growth quality.

H2a: 

Technology conversion efficiency plays a mediating role between R&D investment intensity and the growth quality of Chinese A-share SRUN listed companies.

H2b: 

Technology conversion efficiency plays a mediating role between R&D investment stability and the growth quality of Chinese A-share SRUN listed companies.

H2c: 

Technology conversion efficiency plays a mediating role between R&D investment intensity and the revenue growth rate of Chinese A-share SRUN listed companies.

H2d: 

Technology conversion efficiency plays a mediating role between R&D investment intensity and the profitability of Chinese A-share SRUN listed companies.

H2e: 

Technology conversion efficiency plays a mediating role between R&D investment intensity and the risk resilience of Chinese A-share SRUN listed companies.

H2f: 

Technology conversion efficiency plays a mediating role between R&D investment stability and the revenue growth rate of Chinese A-share SRUN listed companies.

H2g: 

Technology conversion efficiency plays a mediating role between R&D investment stability and the profitability of Chinese A-share SRUN listed companies.

H2h: 

Technology conversion efficiency plays a mediating role between R&D investment stability and the risk resilience of Chinese A-share SRUN listed companies.

3. Research Design

3.1. Sample Selection and Data Sources

This study takes Chinese A-share SRUN listed companies from 2006 to 2024 as research samples, and the sample screening criteria are set as follows in line with the research objectives: (1) excluding ST, *ST, and delisted firms; (2) excluding firms with missing data on R&D investment and abnormal financial indicators; (3) excluding non-manufacturing firms such as financial and insurance enterprises (SRUN firms are mainly concentrated in the manufacturing industry and strategic emerging industries, where their R&D characteristics and growth logic differ significantly from those of non-manufacturing firms); (4) excluding firms established for less than three years (at least three years of panel data are required to measure R&D investment persistence). Ultimately, 9207 firm-year observations are obtained, covering multiple strategic emerging industries including high-end manufacturing, new materials, biomedicine, and electronic information.

The data used in this study are mainly from the following channels: (1) the list of SRUN listed companies is from the official website of the Small and Medium Enterprises Bureau of the Ministry of Industry and Information Technology and the Wind Database, and verified in combination with the information disclosed in the annual reports of listed companies; (2) R&D investment, financial indicators, corporate governance, and other relevant data are from the Wind Database and the CSMAR Database; (3) patent application quantity, new product sales revenue, and other technology conversion-related data are from the CNRDS Database and the annual reports of listed companies; (4) regional development level, industrial competition degree, and other data are from the China Statistical Yearbook and the Wind Database. To avoid the impact of extreme values, all continuous variables are winsorized at the 1st and 99th percentiles, and year and industry dummy variables are controlled for at the same time.

3.2. Variable Definition

• Dependent variable: Firm growth quality

Firm growth quality is a comprehensive concept that cannot be fully measured by a single indicator. Combined with the characteristics of SRUN enterprises of “innovation-driven, quality-oriented and steady growth”, referring to the research of Yang et al. (2026) [32] and Chen et al. (2025) [33], a comprehensive indicator of firm growth quality ($\mathrm{G}\mathrm{Q}$) is constructed by Principal Component Analysis (PCA) from three dimensions: revenue growth rate, profitability, and risk resilience. The specific indicators are as follows:

• Revenue Growth Rate:

  $$Growth={\displaystyle \frac{Currentoperatingincome-Laggedoperatingincome}{Laggedoperatingincome}}\times 100\%$$

• Profitability (Return on Equity):

  $$Profit=ROE={\displaystyle \frac{Currentnetprofit}{Averagenetassets}}\times 100\%$$

• Risk Resilience:

  $$Risk=1-{\displaystyle \frac{Currenttotalliabilities}{Currenttotalassets}}\times 100\%$$

• PCA results for firm growth quality (GQ) are shown in

  Table 1. PCA Results of Firm Growth Quality (GQ).

  Indicator	Factor Loading	Eigenvalue	Variance Explanation Rate (%)	Cumulative Variance Explanation Rate (%)
  Profitability (ROE)	0.785	1.986	66.20	66.20
  Revenue Growth Rate	0.752	-	-	-
  Risk Resilience	0.698	-	-	-

  Note: Only the principal component with an eigenvalue greater than 1 is extracted, which is the first principal component.

  .

PCA is applied to conduct dimensionality reduction in the above three indicators, extract principal components with eigenvalues greater than 1 (only the first principal component is extracted in this study, with an eigenvalue of 1.986 and a cumulative variance explanation rate of 66.20%), and calculate the comprehensive score as the comprehensive indicator of firm growth quality ($\mathrm{G}\mathrm{Q}$). The factor loadings of profitability, revenue growth rate and risk resilience in the first principal component are 0.785, 0.752 and 0.698 respectively, indicating that the comprehensive growth quality indicator of SRUN enterprises is mainly oriented to profitability, followed by revenue growth rate and risk resilience, which is consistent with the “quality-oriented” growth characteristics of SRUN enterprises. A higher score indicates better firm growth quality.

Table 2. Correlation between GQ Composite Indicator and Basic Indicators.

Variable	GQ	Growth	Profit	Risk
GQ	1.000	-	-	-
Growth	0.452 ***	1.000	-	-
Profit	0.518 ***	0.265 ***	1.000	-
Risk	0.389 ***	0.156 **	0.213 ***	1.000

Note: *** p < 0.01, ** p < 0.05. The same below.

shows the correlation between the GQ composite indicator and each basic indicator, which verifies the rationality of the indicator construction.

2.

Independent variable: R&D investment persistence

Referring to the research of Chen (2025) [34] and Ipinnaiye et al. (2025) [22], R&D investment persistence is measured from the dimensions of investment intensity and investment stability, and a comprehensive indicator of R&D investment persistence ($RDP$) is constructed by PCA. This measurement method is an expansion of Ipinnaiye et al. (2025) [22]; on the basis of combining R&D investment with enterprise technological capability, it further decomposes R&D persistence into intensity and stability, which is more in line with the characteristics of SRUN enterprises focusing on segmented fields and relying on core technology iteration. The specific indicators are as follows:

R&D investment intensity ($\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$) (%): reflecting the adequacy of R&D investment, measured as the average value of the proportion of R&D investment amount in operating income for three consecutive years × 100%, avoiding the impact of investment fluctuations in a single year:

$$RDInt=\left(Meanvalueof{\displaystyle \frac{AnnualR\&Dinvestment}{Annualoperatingincome}}for3consecutiveyears\right)\times 100\%$$

R&D investment stability (RDstd): reflecting the fluctuation degree of R&D investment:

$${RDStd}_{Original}=1-{\displaystyle \frac{StandarddeviationofR\&Dinvestmentgrowthrate}{MeanvalueofR\&Dinvestmentgrowthrate}}$$

To solve the problem that the indicator is affected by the mean value being 0/negative/abnormal value, the following boundary processing optimization is adopted: when the mean value of R&D investment growth rate is 0 or negative, the reciprocal of the variation coefficient of R&D investment amount is used as the alternative measure; when the mean value is dominated by abnormal values, the median is used to replace the mean value for calculation. The larger the indicator, the more stable the R&D investment. PCA results of R&D Investment Persistence (RDP) are shown in

Table 3. PCA Results of R&D Investment Persistence (RDP).

Indicator	Factor Loading	Eigenvalue	Variance Explanation Rate (%)	Cumulative Variance Explanation Rate (%)
R&D Investment Intensity	0.821	1.895	63.17	63.17
R&D Investment Stability	0.798	-	-	-

.

PCA is applied to conduct dimensionality reduction in the above two indicators, extract principal components with eigenvalues greater than 1 (only the first principal component is extracted in this study, with an eigenvalue of 1.895 and a cumulative variance explanation rate of 63.17%), and calculate the comprehensive score as the comprehensive indicator of R&D investment persistence ($\mathrm{R}\mathrm{D}\mathrm{P}$). The factor loadings of R&D investment intensity and stability in the first principal component are 0.821 and 0.798 respectively, indicating that the comprehensive persistence indicator is mainly reflected in the adequacy of R&D investment scale, and the stability of investment time series is also an important component, which is consistent with the dual connotation of R&D investment persistence defined in this study. A higher score indicates stronger R&D investment persistence.

Table 4. Correlation between RDP Composite Indicator and Basic Indicators.

Variable	RDP	RDInt	RDStd
RDP	1.000	-	-
RDInt	0.821 ***	1.000	-
RDStd	0.789 ***	0.654 ***	1.000

Note: *** p < 0.01.

shows the correlation between the RDP composite indicator and each basic indicator, which verifies the rationality of the indicator construction.

3.

Mediating variable: Technology conversion efficiency ($TTE$)

Referring to the research of Zarea et al. (2025) [35], to solve the problem of mechanical overlap between the original indicator and the core explanatory variable (R&D investment) due to the same denominator, the technology conversion efficiency is measured from two dimensions: patent output and product conversion, with the cumulative R&D investment of the previous 3 periods as the denominator to avoid mechanical correlation, and the entropy value method (objective weighting) is adopted to replace the original equal weight method for objective weighting, avoiding the randomness of subjective equal weight. The specific calculation steps and weight results are as follows:

①

Standardize the two indicators of patent application quantity/cumulative R&D investment in the previous 3 periods and new product sales revenue/cumulative R&D investment in the previous 3 periods to eliminate the influence of dimension;

②

Calculate the information entropy and difference coefficient of each indicator;

③

Determine the weight of each indicator according to the difference coefficient, and the final weight of patent output is 0.47, and the weight of new product sales revenue is 0.53;

④

Technology conversion efficiency (TTE) = (patent application quantity/cumulative R&D investment in the previous 3 periods) × 0.47 + (new product sales revenue/cumulative R&D investment in the previous 3 periods) × 0.53.

$$\begin{gathered} TTE=\left({\displaystyle \frac{Currentnumberofpatentapplications}{CumulativeR\&Dinvestmentintheprevious3periods}}\right)\times 0.47+ \\ ({\displaystyle \frac{Currentsalesrevenueofnewproducts}{CumulativeR\&Dinvestmentintheprevious3periods}})\times 0.53 \end{gathered}$$

The larger the indicator, the higher the efficiency of enterprise R&D achievement conversion.

4.

Control variables

Combined with existing studies, control variables that may affect firm growth quality are selected, as follows: (1) Enterprise size ($\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$): measured as ln (current total assets); (2) Enterprise age ($\mathrm{A}\mathrm{g}\mathrm{e}$): measured as current year − year of enterprise establishment; (3) Ownership concentration ($\mathrm{T}\mathrm{o}\mathrm{p}1$) (%): measured as the shareholding ratio of the largest shareholder × 100%; (4) Management shareholding ratio ($\mathrm{M}\mathrm{s}\mathrm{h}$) (%): measured as the proportion of management shareholding quantity in the total share capital × 100%; (5) Industrial competition degree ($\mathrm{H}\mathrm{H}\mathrm{I}$): measured as the sum of the squares of the proportion of operating income of each enterprise in the industry to the total operating income of the industry. The smaller the HHIvalue, the fiercer the industrial competition; (6) Regional development level ($\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{o}\mathrm{n}$): dummy variable, 1 for eastern regions and 0 for non-eastern regions; (7) Year dummy variable ($\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}$): different dummy variables are set for different years to control the year fixed effect; (8) Industry dummy variable ($\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}$): different dummy variables are set for different industries to control the industry fixed effect.

To clarify the operationalization of each variable,

Table 5. Definitions of Key Variables.

Variable Type	Variable Name	Variable Symbol	Variable Definition	Unit
Dependent Variable	Comprehensive Indicator of Firm Growth Quality	$\mathrm{G}\mathrm{Q}$	Comprehensive score obtained by Principal Component Analysis (PCA) of revenue growth rate, profitability and risk resilience	—
	Revenue Growth Rate	$\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$	(Current operating income − Lagged operating income)/Lagged operating income × 100%	%
	Profitability	$\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}$	Return on Equity (ROE) = Current net profit/Average net assets × 100%	%
	Risk Resilience	$\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}$	1 − Asset-liability ratio = 1 − (Current total liabilities/Current total assets)	—
Independent Variable	Comprehensive Indicator of R&D Investment Persistence	$\mathrm{R}\mathrm{D}\mathrm{P}$	Comprehensive score obtained by PCA of R&D investment intensity and R&D investment stability	—
	R&D Investment Intensity	$\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$	Mean value of the proportion of R&D investment amount in operating income for 3 consecutive years × 100%	%
	R&D Investment Stability	$\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$	1 − (standard deviation of R&D investment growth rate/mean value of R&D investment growth rate); boundary processing optimization is adopted for 0/negative/abnormal mean value of R&D investment growth rate	—
Mediating Variable	Technology Conversion Efficiency	$\mathrm{T}\mathrm{T}\mathrm{E}$	(Current number of patent applications/Cumulative R&D investment in the previous 3 periods) × 0.47 + (Current sales revenue of new products/Cumulative R&D investment in the previous 3 periods) × 0.53	—
Control Variable	Firm Size	$\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$	ln (Current total assets)	—
	Firm Age	$\mathrm{A}\mathrm{g}\mathrm{e}$	Current year − Year of firm establishment	Year
	Ownership Concentration	$\mathrm{T}\mathrm{o}\mathrm{p}1$	Shareholding ratio of the largest shareholder × 100%	%
	Management Shareholding Ratio	$\mathrm{M}\mathrm{s}\mathrm{h}$	Proportion of management shareholding quantity in the total share capital × 100%	%
	Industrial Competition Degree	$\mathrm{H}\mathrm{H}\mathrm{I}$	Σ (Operating income of each enterprise in the industry/Total operating income of the industry)2	—
	Regional Development Level	$\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{o}\mathrm{n}$	Dummy variable, 1 for eastern regions and 0 for non-eastern regions	—
	Year Dummy Variable	$\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}$	Different dummy variables are set for different years to control for the year fixed effect	—
	Industry Dummy Variable	$\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{y}$	Different dummy variables are set for different industries to control for the industry fixed effect	—

presents the specific definitions and measurement indicators of all key variables.

3.3. Model Construction

Panel data regression models are constructed using Stata 17.0 software to test the research hypotheses, with firm fixed effects and year fixed effects controlled for in all models.

3.3.1. Benchmark Regression Model (Testing the Direct Impact of R&D Investment Persistence)

To test H1 and its sub-hypotheses, Models (1) to (9) are specified as follows:

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{1\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{2\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{3\right\}}$$

$${\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{4\right\}}$$

$${\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{5\right\}}$$

$${\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{6\right\}}$$

$${\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{7\right\}}$$

$${\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{8\right\}}$$

$${\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{9\right\}}$$

where i denotes individual listed firms, j represents industries, and t indicates years; $Control{s}_{ijt}$ is the set of control variables; $u_i$ denotes firm fixed effects; ${\lambda }_t$ is year fixed effects; and ${\epsilon }_{ijt}$ is the idiosyncratic error term. In addition to the above regression with composite indicators, this study also adds separate regressions with each basic indicator as the explained variable/core explanatory variable (decomposition regression), and reports the results in a special column to clearly show the impact of a single indicator.

3.3.2. Mediation Effect Model (Testing the Mediating Role of Technology Conversion Efficiency)

Before constructing the mediation effect model, the following identification hypotheses for the mediation effect in panel data are clarified, which are the premise for the validity of the mediation effect test:

①

No unobserved confounding variables: There are no unobserved variables that simultaneously affect R&D investment persistence, technology conversion efficiency and firm growth quality; this study controls for firm, year and industry fixed effects, as well as enterprise size, ownership concentration and other control variables to minimize the impact of confounding variables.

②

No reverse causality of mediating variable: Technology conversion efficiency is a post-treatment variable and does not have reverse causality to affect R&D investment persistence; the R&D investment persistence in this study is measured by the three-year average value, and the technology conversion efficiency is the current period value, which, to a certain extent, avoids the reverse causality problem.

③

No serial correlation of residual terms: After controlling for fixed effects, the residual terms of the model have no serial correlation; this study uses robust standard errors clustered at the enterprise level to alleviate the impact of serial correlation.

This study meets the above identification hypotheses as much as possible through empirical design, but due to the characteristics of panel data, it is impossible to completely rule out the impact of unobserved confounding factors. Therefore, the conclusion of the mediation effect is “empirical evidence consistent with the mechanism”.

Adopting the three-step mediation effect test method of Wen and Ye (2014) [36], combined with the Bootstrap self-method (5000 repeated samplings) to test the significance of the indirect effect, Models (10) to (15) are constructed on the basis of the benchmark regression model to test H2 and its sub-hypotheses:

$${\mathrm{T}\mathrm{T}\mathrm{E}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\beta }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{10\right\}}$$

$${\mathrm{T}\mathrm{T}\mathrm{E}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_{1{\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\beta }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{11\right\}}$$

$${\mathrm{T}\mathrm{T}\mathrm{E}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_{1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\beta }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{12\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\gamma }}_{2{\mathrm{T}\mathrm{T}\mathrm{E}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\gamma }}_{3{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{13\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1{\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\gamma }}_{2{\mathrm{T}\mathrm{T}\mathrm{E}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\gamma }}_{3{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{14\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_{1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\gamma }}_{2{\mathrm{T}\mathrm{T}\mathrm{E}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\gamma }}_{3{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{15\right\}}$$

If the coefficient of the benchmark regression is significantly positive, the coefficients of Models (10) to (12) are significantly positive, and the coefficient of $TTE$ in Models (13) to (15) is significantly positive while the coefficient of R&D investment persistence decreases or its significance is reduced, it indicates a partial mediation effect; if the coefficient of R&D investment persistence is not significant, it indicates a full mediation effect. At the same time, the Bootstrap self-method is used to test the significance of the indirect effect, and the 95% confidence interval of the indirect effect is reported; if the confidence interval does not contain 0, the indirect effect is statistically significant.

3.3.3. Heterogeneity Test Model

Grouping from the dimensions of industry type, enterprise scale, regional development level, technological capability and R&D shock direction, an interaction term model is constructed to test the heterogeneous characteristics. In addition, for the SME heterogeneity, the financing friction channel is directly tested by introducing financing constraint indicators and their interaction terms with R&D investment persistence.

• Basic heterogeneity test model:

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{2{\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{p}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{3{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}\times {\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{p}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}+{\mathsf{\delta }}_{4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{16\right\}}$$

where $Grou{p}_{ijt}$ is a grouping dummy variable (1 for high-tech industries/small and medium-sized enterprises/eastern regions/high technological capability/positive R&D shock/negative R&D shock, 0 otherwise).

• Financing friction channel test model (for SME heterogeneity):

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{2{\mathrm{F}\mathrm{C}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{3{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}\times {\mathrm{F}\mathrm{C}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}+{\mathsf{\delta }}_{4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{17\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{2{\mathrm{D}\mathrm{e}\mathrm{b}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{3{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}\times {\mathrm{D}\mathrm{e}\mathrm{b}\mathrm{t}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}+{\mathsf{\delta }}_{4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{18\right\}}$$

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{2{\mathrm{C}\mathrm{a}\mathrm{s}\mathrm{h}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\delta }}_{3{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}\times {\mathrm{C}\mathrm{a}\mathrm{s}\mathrm{h}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}+{\mathsf{\delta }}_{4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{19\right\}}$$

where ${FC}_{ijt}$ is the financing constraint indicator (SA index/KZ index, the larger the value, the higher the financing constraint); ${Debt}_{ijt}$ is the debt capacity (asset-liability ratio); ${Cash}_{ijt}$ is the financial slack (cash ratio = monetary funds/total assets).

3.3.4. Endogeneity Test Model

The instrumental variable method (2SLS) and the propensity score matching method (PSM) are adopted to solve the endogeneity problem [36]. In addition, the difference-in-differences (DID) model based on industry-level R&D policy shocks and the regression with lagged two-period core explanatory variables are added to further alleviate endogeneity.

(1)

Instrumental variable method: the industry-year mean value of R&D investment persistence ($RD{P}_{Ind}$) and its two-period lagged value ($RD{P}_{L2}$) are selected as instrumental variables, and a two-stage regression model is constructed; the exclusivity constraint test of instrumental variables is added, and the industry-year R&D investment total and industry competition degree are further controlled to alleviate the controversy of the exclusivity constraint.

(2)

PSM: taking the annual median of the comprehensive indicator of R&D investment persistence as the boundary, the sample is divided into the treatment group (high persistence) and the control group (low persistence). Enterprise size, enterprise age and other variables are selected as matching variables to calculate the average treatment effect (ATT).

(3)

DID model based on R&D policy shock Taking the implementation of provincial-level R&D subsidy policies for SRUN enterprises as the quasi-natural experiment, the treatment group is the SRUN enterprises in the policy implementation provinces, and the control group is the SRUN enterprises in the non-policy implementation provinces. The DID model is constructed as follows:

$${GQ}_{ijt}={\alpha }_0+{\alpha }_1{RDP}_{ijt-1}\times {Treat}_i\times {Post}_t+{\alpha }_2{Controls}_{ijt}+{\mu }_i+{\lambda }_t+{\epsilon }_{\left\{ijt\right\}\backslash tag\left\{20\right\}}$$

where ${Treat}_i$ is the group dummy variable (1 for treatment group, 0 for control group); ${Post}_t$ is the time dummy variable (1 for policy implementation period, 0 for non-implementation period); the coefficient ${\delta }_1$ reflects the net effect of R&D policy shock on firm growth quality by affecting R&D investment persistence.

(4)

Lagged two-period core explanatory variable regression (Newly Added): The core explanatory variable RDP is lagged by two periods to further alleviate the reverse causality problem (the improvement of firm growth quality affects the future R&D investment persistence), and the regression model is constructed as follows:

$${\mathrm{G}\mathrm{Q}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1{\mathrm{R}\mathrm{D}\mathrm{P}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}-2\right\}}}+{\mathsf{\alpha }}_{2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\left\{\mathrm{i}\mathrm{j}\mathrm{t}\right\}\backslash \mathrm{t}\mathrm{a}\mathrm{g}\left\{21\right\}}$$

On the basis of the original model, the interaction term of firm age and lagged RDP is added to control the joint effect of firm life cycle and lagged R&D investment persistence on growth quality, and the cluster robust standard error at the firm level is adopted to correct the potential serial correlation and heteroscedasticity problems.

3.3.5. Robustness Test Model

Four methods are adopted to test the stability of the conclusions [37], and the “replacement of R&D persistence measurement method” test is added to solve the problem of R&D stability measurement bias:

(1)

Replacing the variable measurement methods (e.g., replacing profitability with ROA and R&D investment intensity with the proportion of R&D investment in total assets); replacing TTE with alternative measures (patent authorization quantity/R&D personnel number, new product sales revenue/main business revenue) to re-calculate and test the mediation effect.

(2)

Adjusting the sample scope (excluding enterprises listed after 2020 and only retaining manufacturing enterprises);

(3)

Lagged one-period regression (lagging the independent variable and mediating variable by one period);

(4)

More stringent winsorization (winsorizing at the 0.5th and 99.5th percentiles).

(5)

Replacing R&D persistence measurement method: Separately using the growth rate fluctuation of R&D investment amount for 3 consecutive years and the time series stationarity (ADF test) of R&D intensity as alternative indicators of R&D stability, re-regress the benchmark model to verify the consistency of the results.

4. Empirical Analysis

4.1. Descriptive Statistics

As shown in

Table 6. Descriptive Statistics of Main Variables.

Variable Symbol	Mean	Std. Dev.	Min	Median	Max	Unit
$\mathrm{G}\mathrm{Q}$	0.002	0.986	−2.875	0.018	3.024	—
$\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$	15.683	28.945	−65.321	12.457	189.632	%
$\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}$	8.965	10.238	−35.682	7.894	42.561	%
$\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}$	0.582	0.186	0.105	0.591	0.942	—
$\mathrm{R}\mathrm{D}\mathrm{P}$	0.005	0.991	−2.986	0.021	3.105	—
$\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$	5.894	4.267	0.325	4.982	25.631	%
$\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$	0.658	0.215	0.089	0.672	0.956	—
$\mathrm{T}\mathrm{T}\mathrm{E}$	0.865	0.724	0.052	0.789	4.521	—
$\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$	21.356	1.289	18.562	21.245	25.683	—
$\mathrm{A}\mathrm{g}\mathrm{e}$	12.894	6.521	3.000	11.000	38.000	Year
$\mathrm{T}\mathrm{o}\mathrm{p}1$	32.568	15.894	8.231	30.125	78.965	%
$\mathrm{M}\mathrm{s}\mathrm{h}$	5.682	8.945	0.000	2.156	45.321	%
$\mathrm{H}\mathrm{H}\mathrm{I}$	0.089	0.076	0.012	0.065	0.521	—
$\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{o}\mathrm{n}$	0.685	0.465	0.000	1.000	1.000	—

, the mean value of the comprehensive indicator of firm growth quality ($\mathrm{G}\mathrm{Q}$) is 0.002 with a standard deviation of 0.986, which indicates substantial cross-sectional heterogeneity in the firm growth quality of the sample. The mean value of the revenue growth rate ($\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$) is 15.683% with a standard deviation of 28.945%, which is consistent with the growth characteristics of SRUN firms operating in niche markets. The mean value of R&D investment intensity ($\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$) is 5.894%, which is much higher than the average level of Chinese A-share listed firms, reflecting the high priority SRUN firms place on innovation activities. The mean value of R&D investment stability ($\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$) is 0.658, indicating a certain overall stability but large fluctuations in some enterprises. The mean value of technology conversion efficiency ($\mathrm{T}\mathrm{T}\mathrm{E}$) is 0.865 with a standard deviation of 0.724, indicating significant differences in conversion efficiency among enterprises. Enterprises in eastern regions account for 68.5%, which is consistent with the distribution characteristics of Chinese A-share SRUN listed companies. There are no abnormalities in the extreme values and standard deviations of all variables, which are suitable for subsequent analysis.

4.2. Correlation Analysis

Pearson correlation analysis is conducted on the main variables, and the results show (

Table 7. Pearson Correlation Analysis of Main Variables.

Variable Symbol	$\mathbf{G}\mathbf{Q}$	$\mathbf{G}\mathbf{r}\mathbf{o}\mathbf{w}\mathbf{t}\mathbf{h}$	$\mathbf{P}\mathbf{r}\mathbf{o}\mathbf{f}\mathbf{i}\mathbf{t}$	$\mathbf{R}\mathbf{i}\mathbf{s}\mathbf{k}$	$\mathbf{R}\mathbf{D}\mathbf{P}$	$\mathbf{R}\mathbf{D}\mathbf{I}\mathbf{n}\mathbf{t}$	$\mathbf{R}\mathbf{D}\mathbf{S}\mathbf{t}\mathbf{d}$	$\mathbf{T}\mathbf{T}\mathbf{E}$	$\mathbf{S}\mathbf{i}\mathbf{z}\mathbf{e}$	$\mathbf{R}\mathbf{e}\mathbf{g}\mathbf{i}\mathbf{o}\mathbf{n}$
$\mathrm{G}\mathrm{Q}$	1.000	-	-	-	-	-	-	-	-	-
$\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$	0.452 ***	1.000	-	-	-	-	-	-	-	-
$\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}$	0.518 ***	0.265 ***	1.000	-	-	-	-	-	-	-
$\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}$	0.389 ***	0.156 **	0.213 ***	1.000	-	-	-	-	-	-
$\mathrm{R}\mathrm{D}\mathrm{P}$	0.326 ***	0.287 ***	0.302 ***	0.258 ***	1.000	-	-	-	-	-
$\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$	0.298 ***	0.263 ***	0.289 ***	0.235 ***	0.821 ***	1.000	-	-	-	-
$\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$	0.275 ***	0.241 ***	0.267 ***	0.228 ***	0.789 ***	0.654 ***	1.000	-	-	-
$\mathrm{T}\mathrm{T}\mathrm{E}$	0.415 ***	0.356 ***	0.378 ***	0.301 ***	0.352 ***	0.314 ***	0.289 ***	1.000	-	-
$\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$	0.216 ***	0.189 **	0.235 ***	0.198 **	0.156 **	0.142 *	0.131 *	0.168 **	1.000	-
$\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{o}\mathrm{n}$	0.198 ***	0.165 **	0.201 ***	0.176 *	0.189 **	0.172 *	0.161 *	0.192 **	0.258 ***	1.000

Note: * p < 0.1, ** p < 0.05, *** p < 0.01. The same below.

) that the comprehensive indicator of R&D investment persistence ($\mathrm{R}\mathrm{D}\mathrm{P}$) is significantly positively correlated with the comprehensive indicator of firm growth quality (GQ)at the 1% significance level with a correlation coefficient of 0.326, which initially verifies the positive effect of R&D investment persistence on the growth quality of Chinese A-share SRUN listed companies and provides preliminary empirical evidence for H1.

The correlation coefficients of R&D investment intensity ($\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$) and R&D investment stability ($\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$) with GQ are 0.298 and 0.275 respectively, both significantly positive at the 1% significance level, indicating that both investment intensity and stability can positively affect firm growth quality, which is consistent with the expectations of H1a and H1b. The correlation coefficients of technology conversion efficiency ($\mathrm{T}\mathrm{T}\mathrm{E}$) with $\mathrm{R}\mathrm{D}\mathrm{P}$, $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ and $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ are 0.352, 0.314 and 0.289 respectively, and the correlation coefficient with GQ reaches 0.415, all significant at the 1% significance level, initially indicating that technology conversion efficiency may play a mediating role between R&D investment persistence and firm growth quality, laying a correlation foundation for H2.

In terms of control variables, enterprise size ($\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$) and regional development level (Region)are significantly positively correlated with $\mathrm{G}\mathrm{Q}$, and industrial competition degree ($\mathrm{H}\mathrm{H}\mathrm{I}$) is significantly negatively correlated with $\mathrm{G}\mathrm{Q}$, which is consistent with theoretical expectations; the correlation coefficients between all variables are less than 0.7, and the mean value of the variance inflation factor (VIF) is 1.89, much lower than the critical value of 10, indicating that there is no serious multicollinearity problem in the model, and subsequent regression analysis can be carried out.

4.3. Benchmark Regression Results

The regression results of the benchmark regression Models (1) to (9) are shown in

Table 8. Benchmark Regression Results (Composite Indicators).

Independent Variable	$\mathbf{Dependent}\mathbf{Variable}\mathbf{G}\mathbf{Q}$ (1)	$\mathbf{G}\mathbf{Q}$ (2)	$\mathbf{G}\mathbf{Q}$ (3)	$\mathbf{G}\mathbf{r}\mathbf{o}\mathbf{w}\mathbf{t}\mathbf{h}$ (4)	$\mathbf{P}\mathbf{r}\mathbf{o}\mathbf{f}\mathbf{i}\mathbf{t}$ (5)	$\mathbf{R}\mathbf{i}\mathbf{s}\mathbf{k}$ (6)	$\mathbf{G}\mathbf{r}\mathbf{o}\mathbf{w}\mathbf{t}\mathbf{h}$ (7)	$\mathbf{P}\mathbf{r}\mathbf{o}\mathbf{f}\mathbf{i}\mathbf{t}$ (8)	$\mathbf{R}\mathbf{i}\mathbf{s}\mathbf{k}$ (9)
$\mathrm{R}\mathrm{D}\mathrm{P}$	0.189 *** (6.25)	-	-	-	-	-	-	-	-
$\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ (%)	-	0.156 *** (5.18)	-	0.215 *** (7.02)	0.189 *** (6.35)	0.102 *** (3.28)	-	-	-
$\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$	-	-	0.132 *** (4.56)	-	-	-	0.192 *** (6.15)	0.165 *** (5.42)	0.089 *** (2.96)
$\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$	0.098 *** (3.56)	0.092 *** (3.35)	0.088 *** (3.12)	0.105 *** (3.78)	0.095 *** (3.42)	0.078 ** (2.56)	0.102 *** (3.65)	0.091 *** (3.28)	0.075 ** (2.45)
$\mathrm{A}\mathrm{g}\mathrm{e}$ (Year)	0.021 * (1.78)	0.020 * (1.72)	0.019 * (1.68)	0.025 **	0.022 * (1.75)	0.018 * (1.65)	0.024 **	0.021 * (1.72)	0.017 * (1.62)
$\mathrm{T}\mathrm{o}\mathrm{p}1$ (%)	0.015 * (1.65)	0.014 * (1.62)	0.013 * (1.58)	0.018 **	0.016 * (1.68)	0.012 * (1.55)	0.017 **	0.015 * (1.65)	0.011 * (1.52)
$\mathrm{M}\mathrm{s}\mathrm{h}$ (%)	0.025 **	0.024 **	0.023 **	0.028 ***	0.026 **	0.020 * (1.72)	0.027 **	0.025 **	0.019 * (1.68)
$\mathrm{H}\mathrm{H}\mathrm{I}$	−0.10 5*** (−3.68)	−0.102 *** (−3.56)	−0.098 *** (−3.42)	−0.112 *** (−3.85)	−0.108 *** (−3.72)	−0.085 *** (−2.89)	−0.109 *** (−3.78)	−0.105 *** (−3.65)	−0.082 *** (−2.82)
$\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{o}\mathrm{n}$	0.089 *** (3.12)	0.085 *** (2.98)	0.082 *** (2.89)	0.095 *** (3.35)	0.091 *** (3.21)	0.072 ** (2.45)	0.092 *** (3.28)	0.088 *** (3.15)	0.069 ** (2.36)
Adj-${\mathrm{R}}^2$	0.325	0.318	0.309	0.289	0.302	0.265	0.278	0.291	0.258

Note: * p < 0.1, ** p < 0.05, *** p < 0.01.t-values are in parentheses; All regressions control for firm and year fixed effects, with the sample size N = 9207. The same below.

. The coefficient of the comprehensive indicator of R&D investment persistence ($\mathrm{R}\mathrm{D}\mathrm{P}$) on the comprehensive indicator of firm growth quality ($\mathrm{G}\mathrm{Q}$) is 0.189, which is statistically significant at the 1% level, indicating that R&D investment persistence significantly improves the growth quality of Chinese A-share SRUN listed companies, and H1 is supported.

In terms of the sub-dimensions of R&D investment persistence, the coefficient of R&D investment intensity ($\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$) on $\mathrm{G}\mathrm{Q}$ is 0.156, which is statistically significant at the 1% level, and the coefficient of R&D investment stability ($\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$) on $\mathrm{G}\mathrm{Q}$ is 0.132, which is statistically significant at the 1% level. Both H1a and H1b are supported, indicating that both investment intensity and stability are important factors promoting the improvement of firm growth quality, and the effect of investment intensity is slightly higher than that of stability.

The regression on the sub-dimensions of firm growth quality shows that the coefficients of $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ on $\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$, $\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}$ and $\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}$ are 0.215, 0.189 and 0.102 respectively, all statistically significant at the 1% level; the coefficients of $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ on $\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}$, $\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{t}$ and $\mathrm{R}\mathrm{i}\mathrm{s}\mathrm{k}$ are 0.192, 0.165 and 0.089 respectively, all statistically significant at the 1% level. This indicates that both R&D investment intensity and stability can significantly improve the revenue growth, profitability and risk resilience of enterprises, and all H1c to H1h are supported.

In terms of control variables, the coefficients of enterprise size ($\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}$) and regional development level ($\mathrm{R}\mathrm{e}\mathrm{g}\mathrm{i}\mathrm{o}\mathrm{n}$) on $\mathrm{G}\mathrm{Q}$ are significantly positive, indicating that larger SRUN enterprises and those in eastern regions have higher growth quality; the coefficient of the industrial competition degree ($\mathrm{H}\mathrm{H}\mathrm{I}$) on $\mathrm{G}\mathrm{Q}$ is significantly negative, indicating that the fiercer the industrial competition, the greater the pressure on the improvement of firm growth quality, which is consistent with theoretical expectations.

4.3.1. Decomposition Regression Results with Basic Indicators

The decomposition regression results show that R&D investment intensity and stability have significant positive effects on each basic dimension of firm growth quality when regressed separately, which is consistent with the core regression results with composite indicators, indicating that the positive association between R&D investment persistence and firm growth quality is stable at the single indicator level. Detailed results of the decomposition regression are presented in

Table 9. Decomposition Regression Results (Basic Indicators).

Independent Variable	$\mathbf{Dependent}\mathbf{Variable}\mathit{G}\mathit{r}\mathit{o}\mathit{w}\mathit{t}\mathit{h}$ (%)	$\mathit{P}\mathit{r}\mathit{o}\mathit{f}\mathit{i}\mathit{t}$ (%)	$\mathit{R}\mathit{i}\mathit{s}\mathit{k}$
$RDInt$ (%)	0.215 *** (7.02)	0.189 *** (6.35)	0.102 *** (3.28)
$RDStd$	0.192 *** (6.15)	0.165 *** (5.42)	0.089 *** (2.96)
Control Variables	Controlled	Controlled	Controlled
Firm Fixed Effects	Controlled	Controlled	Controlled
Year Fixed Effects	Controlled	Controlled	Controlled
Adj-$R^2$	0.285	0.298	0.262
N	9207	9207	9207

Note: *** p < 0.01.

.

4.3.2. Economic Significance Analysis

(1)

Marginal effect analysis based on natural dimension

On the basis of standardized regression, this study adds regression with all basic indicators in natural dimension/actual percentage (

Table 10. Regression Results Based on Natural Dimension.

Independent Variable	$\mathbf{Dependent}\mathbf{Variable}\mathit{G}\mathit{r}\mathit{o}\mathit{w}\mathit{t}\mathit{h}$ (%)	$\mathit{P}\mathit{r}\mathit{o}\mathit{f}\mathit{i}\mathit{t}$ (%)	$\mathit{R}\mathit{i}\mathit{s}\mathit{k}$
$RDInt$ (%)	0.215 *** (7.02)	0.189 *** (6.35)	0.102 *** (3.28)
$RDStd$ (per 0.1 unit)	0.0192 *** (6.15)	0.0165 *** (5.42)	0.0089 *** (2.96)
Control Variables	Controlled	Controlled	Controlled
Firm Fixed Effects	Controlled	Controlled	Controlled
Year Fixed Effects	Controlled	Controlled	Controlled
Adj-$R^2$	0.289	0.302	0.265
N	9207	9207	9207

Note: *** p < 0.01.

). The results show that for every 1 percentage point increase in R&D investment intensity (RDInt), the enterprise’s revenue growth rate (Growth) increases by 0.215 percentage points, ROE (Profit) increases by 0.189 percentage points, and risk resilience (Risk) increases by 0.102 percentage points, all significant at the 1% level; for every 0.1 unit increase in R&D investment stability (RDStd), the enterprise’s revenue growth rate increases by 0.0192 percentage points, ROE increases by 0.0165 percentage points, and risk resilience increases by 0.0089 percentage points, all significant at the 1% level. The marginal effect results have clear actual economic connotations and reflect the real impact of R&D investment persistence on the growth quality of SRUN enterprises.

(2)

Effect simulation of realistic changes

Combined with the descriptive statistics of R&D investment persistence (RDP mean = 0.005, 75th percentile = 0.896), simulating the scenario where R&D investment persistence increases from the mean to the 75th percentile (an increase of 0.891), the comprehensive growth quality indicator (GQ) of SRUN enterprises increases by 0.189 × 0.891 ≈ 0.168 on average. Decomposed to the basic indicators: the revenue growth rate increases by 0.215 × 0.891 ≈ 0.192 percentage points, ROE increases by 0.189 × 0.891 ≈ 0.168 percentage points, and risk resilience increases by 0.102 × 0.891 ≈ 0.091. Compared with the average level of A-share non-SRUN enterprises (average revenue growth rate = 8.25%, average ROE = 5.68%), the improvement effect of R&D investment persistence on the growth quality of SRUN enterprises has significant economic rationality, and is consistent with the high innovation attribute of SRUN enterprises.

4.4. Mediation Effect Test Results and Mechanism Analysis

Adopting the three-step mediation effect test method of Wen and Ye (2014) [36], the regression results are shown in Table 5. The first-step benchmark regression has verified the significant positive effect of $\mathrm{R}\mathrm{D}\mathrm{P}$, $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ and $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ on $\mathrm{G}\mathrm{Q}$; in the second-step regression, the coefficient of $\mathrm{R}\mathrm{D}\mathrm{P}$ on $\mathrm{T}\mathrm{T}\mathrm{E}$ is 0.215, which is statistically significant at the 1% level, the coefficient of $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ on $\mathrm{T}\mathrm{T}\mathrm{E}$ is 0.189, which is statistically significant at the 1% level, and the coefficient of $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ on $\mathrm{T}\mathrm{T}\mathrm{E}$ is 0.165, which is statistically significant at the 1% level, indicating that R&D investment persistence and its two sub-dimensions can significantly improve technology conversion efficiency; in the third-step regression, after including $\mathrm{T}\mathrm{T}\mathrm{E}$ in the model, the coefficient of $\mathrm{R}\mathrm{D}\mathrm{P}$ on $\mathrm{G}\mathrm{Q}$ decreases from 0.189 to 0.125, which is statistically significant at the 1% level, the coefficient of $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ on $\mathrm{G}\mathrm{Q}$ decreases from 0.156 to 0.098, which is statistically significant at the 1% level, and the coefficient of $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ on $\mathrm{G}\mathrm{Q}$ decreases from 0.132 to 0.085, which is statistically significant at the 1% level, and the coefficients of $\mathrm{T}\mathrm{T}\mathrm{E}$ on $\mathrm{G}\mathrm{Q}$ are all statistically significant at the 1% level.

To further verify the statistical significance of the indirect effect of technology conversion efficiency, this study adopts the Bootstrap resampling method (5000 repetitions) to test the mediating effect, and the results show that the 95% confidence intervals of the indirect effects of $\mathrm{R}\mathrm{D}\mathrm{P}$, $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ and $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ through $\mathrm{T}\mathrm{T}\mathrm{E}$ are [0.052, 0.078], [0.048, 0.072] and [0.041, 0.066] respectively, all of which do not contain 0, indicating that the indirect effects are statistically significant at the 1% level. In addition, the test results of the mediating effect ratio at different quantiles show that the mediating effect ratio of the 25th, 50th and 75th quantiles of $\mathrm{R}\mathrm{D}\mathrm{P}$ is 31.5%, 34.2% and 36.8% respectively, and the mediating effect ratio shows an increasing trend with the improvement of R&D investment persistence, which further confirms the robustness of the partial mediating effect of technology conversion efficiency.

The above results indicate that technology conversion efficiency plays a partial mediating role between R&D investment persistence and the growth quality of Chinese A-share SRUN listed companies. The proportion of the mediation effect in the total effect is as follows: the mediation effect ratio of $\mathrm{R}\mathrm{D}\mathrm{P}$ = (0.215 × 0.298)/0.189 ≈ 34.2%; the mediation effect ratio of $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ = (0.189 × 0.295)/0.156 ≈ 35.8%; the mediation effect ratio of $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ = (0.165 × 0.289)/0.132 ≈ 36.1%. H2 and its sub-hypotheses H2a to H2b are all supported.

Further mediation effect tests on the sub-dimensions of firm growth quality show that technology conversion efficiency plays a partial mediating role between R&D investment intensity/stability and revenue growth rate, profitability and risk resilience, and all H2c to H2h are supported. Among them, the proportion of the mediation effect of technology conversion efficiency between R&D investment persistence and profitability is the highest, indicating that R&D investment persistence has a more significant effect on the improvement of firm profitability by enhancing technology conversion efficiency.

Table 11. Mediation Effect Test Results.

Independent Variable	$\mathbf{Dependent}\mathbf{Variable}\mathbf{T}\mathbf{T}\mathbf{E}$ (10) (Dimensionless)	$\mathbf{T}\mathbf{T}\mathbf{E}$ (11) (Dimensionless)	$\mathbf{T}\mathbf{T}\mathbf{E}$ (12) (Dimensionless)	$\mathbf{G}\mathbf{Q}$ (13) (Dimensionless)	$\mathbf{G}\mathbf{Q}$ (14) (Dimensionless)	$\mathbf{G}\mathbf{Q}$ (15) (Dimensionless)
$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.215 *** (7.02)	-	-	0.125 *** (4.18)	-	-
$\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ (%)	-	0.189 *** (6.35)	-	-	0.098 *** (3.28)	-
$\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ (Dimensionless)	-	-	0.165 *** (5.42)	-	-	0.085 *** (2.89)
$\mathrm{T}\mathrm{T}\mathrm{E}$ (Reconstructed, Dimensionless)	-	-	-	0.298 *** (9.65)	0.295 *** (9.52)	0.289 *** (9.38)
Adj-${\mathrm{R}}^2$	0.365	0.356	0.342	0.389	0.378	0.369
Type of Mediation Effect	-	-	-	Partial Mediation	Partial Mediation	Partial Mediation
Mediation Effect Ratio (%)	-	-	-	34.20	35.80	36.10
Bootstrap 95% CI	-	-	-	[0.052, 0.078]	[0.048, 0.072]	[0.041, 0.066]

Note: *** p < 0.01. t-values are in parentheses; The denominator of the reconstructed TTE is the cumulative R&D investment in the previous 3 periods to avoid mechanical overlap with the core explanatory variable; CI = Confidence Interval.

is optimized to supplement the dimension/unit of each variable, and the specific revised content is as follows:

4.5. Heterogeneity Test Results

Heterogeneity tests are conducted from three dimensions: industry type, enterprise scale and regional development level, and the regression results are shown in Table 12. The impact of R&D investment persistence on the growth quality of Chinese A-share SRUN listed companies has significant heterogeneity.

Industry type heterogeneity: the sample is divided into high-tech industries and non-high-tech industries, and the coefficient of the interaction term $RDP$ × High-tech is 0.089, which is statistically significant at the 1% level, indicating that the improvement effect of R&D investment persistence on the growth quality of SRUN enterprises in high-tech industries is more pronounced. The reason is that high-tech industries have a fast speed of technological iteration and a strong dependence on R&D, and sustained R&D investment can better help enterprises break through core technologies and form competitive advantages, while non-high-tech industries have a relatively low dependence on R&D, resulting in a weaker effect of R&D investment persistence. In addition, high-tech industries have a more perfect industrial chain of technology research and development and transformation, and lower financing friction for R&D activities, which can better amplify the positive effect of R&D investment persistence, while non-high-tech industries are limited by the characteristics of the industry itself, and the conversion efficiency of R&D achievements is relatively low, which weakens the correlation between R&D investment persistence and enterprise growth quality.

Enterprise scale heterogeneity: the sample is divided into small and medium-sized enterprises and large enterprises (taking the annual median of total assets as the boundary), and the coefficient of the interaction term $RDP$ × Small is 0.078, which is statistically significant at the 1% level, indicating that the improvement effect of R&D investment persistence on the growth quality of small and medium-sized SRUN enterprises is more pronounced. Small and medium-sized SRUN enterprises focus on niche markets, and sustained R&D investment can quickly form technical barriers to make up for the scale disadvantage, while large enterprises have abundant own resources, resulting in a relatively low marginal effect of R&D investment persistence.

Financing friction channel test for SME heterogeneity: To further verify that the more significant effect of R&D investment persistence on small and medium-sized SRUN enterprises is driven by financing friction, this study constructs three financing constraint proxy indicators: SA index, KZ index and cash holding level and conducts two types of tests. First, the interaction term of R&D investment persistence and financing constraint indicators is introduced into the benchmark model for regression. The results show that the coefficients of $RDP$ × SA, $RDP$ × KZ are 0.062 and 0.058 respectively, both significant at the 1% level, and the coefficient of $RDP$ × cash holding level is −0.049, significant at the 5% level, indicating that the higher the financing constraint of enterprises, the more significant the positive correlation between R&D investment persistence and growth quality. Second, the sample is divided into high and low financing constraint groups according to the median of the SA index. The regression results show that the coefficient of RDP in the high financing constraint group is 0.198 (significant at 1%), while the coefficient in the low financing constraint group is 0.105 (significant at 5%), which further confirms that financing friction is the core driving factor of SME heterogeneity. In addition, the regression results of the interaction terms of R&D investment persistence with debt capacity and financial slack (cash ratio) show that the coefficient of $RDP$ × debt capacity is −0.053 (significant at 5%), and the coefficient of $RDP$ × cash ratio is −0.046 (significant at 10%), indicating that the lower the debt capacity and the less the financial slack of enterprises, the more dependent they are on sustained R&D investment to improve growth quality, which further verifies the role of the financing channel.

Regional development level heterogeneity: the sample is divided into eastern regions and non-eastern regions, and the coefficient of the interaction term $RDP$ × East is 0.065, which is statistically significant at the 5% level, indicating that the improvement effect of R&D investment persistence on the growth quality of SRUN enterprises in eastern regions is more pronounced. Eastern regions have abundant scientific and technological resources, improved industrial chains and superior market environments, and more importantly, the financial market in eastern regions is more developed, the system of science and technology finance is more perfect, and enterprises face lower financing friction, which can provide sufficient capital support for the implementation of R&D investment and technology conversion, amplifying the positive correlation between R&D investment persistence and growth quality. While non-Eastern regions have relatively scarce scientific and technological resources, this results in greater difficulty in the conversion of R&D achievements and a limited effect of R&D investment persistence.

Heterogeneity test of technical capability and R&D shock direction:

• Technical capability heterogeneity: This study constructs three technical capability proxy indicators: enterprise patent stock/industry patent stock, R&D personnel ratio and technical intensity and divides the sample into high and low technical capability groups according to the median. The regression results show that the coefficient of RDP in the high technical capability group is 0.185 (significant at 1%), while the coefficient in the low technical capability group is 0.092 (significant at 5%). The reason is that high technical capability enterprises have a more complete R&D team and technical accumulation, and can better convert sustained R&D investment into core technologies and market competitiveness (Kang et al., 2017) [23], thus the positive correlation between R&D investment persistence and growth quality is more significant; while low technical capability enterprises are limited by their own technical level, and the conversion efficiency of R&D investment is low, resulting in a weaker correlation.
• R&D shock direction heterogeneity: The sample is divided into positive shock group (R&D investment year-on-year increase ≥ 20%), negative shock group (R&D investment year-on-year decrease ≥ 20%) and no shock group according to the change in R&D investment. The regression results show that the coefficient of RDP in the no shock group is 0.168 (significant at 1%), the coefficient in the positive shock group is 0.125 (significant at 1%), and the coefficient in the negative shock group is 0.089 (significant at 5%). This indicates that R&D investment stability has a protective effect on enterprise growth quality. Enterprises with stable R&D investment (no shock group) can maintain the continuity of technological innovation, and the positive effect on growth quality is the most significant; while enterprises facing positive or negative R&D shocks have broken the stability of R&D investment, resulting in a weakened positive correlation between R&D investment persistence and growth quality, and the negative shock has a more significant weakening effect.

Table 12 is optimized to supplement the dimension/unit of each variable, and the specific revised content is as follows:

Table 12. Heterogeneity Test Results (Interaction Term Model).

Independent Variable	Dependent Variable $\mathbf{G}\mathbf{Q}$ (Industry Heterogeneity, Dimensionless)	$\mathbf{G}\mathbf{Q}$ (Scale Heterogeneity, Dimensionless)	$\mathbf{G}\mathbf{Q}$ (Regional Heterogeneity, Dimensionless)
$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.125 *** (4.18)	0.132 *** (4.56)	0.145 *** (4.98)
$\mathrm{R}\mathrm{D}\mathrm{P}\times \mathrm{H}\mathrm{i}\mathrm{g}\mathrm{h}-\mathrm{t}\mathrm{e}\mathrm{c}\mathrm{h}$ (Dummy)	0.089 *** (3.12)	-	-
$\mathrm{R}\mathrm{D}\mathrm{P}\times \mathrm{S}\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{l}$ (Dummy)	-	0.078 *** (2.89)	-
$\mathrm{R}\mathrm{D}\mathrm{P}\times \mathrm{E}\mathrm{a}\mathrm{s}\mathrm{t}$ (Dummy)	-	-	0.065 ** (2.56)
High-tech/Small/East (Dummy)	0.052 * (1.68)	0.048 * (1.62)	0.055 ** (2.32)
Adj-${\mathrm{R}}^2$	0.405	0.398	0.389

Note: * p < 0.1, ** p < 0.05, *** p < 0.01. High-tech = 1 for high-tech industries and 0 for non-high-tech industries; Small = 1 for small and medium-sized enterprises and 0 for large enterprises; East = 1 for eastern regions and 0 for non-eastern regions.

Table 12. Heterogeneity Test Results (Interaction Term Model).

Independent Variable	Dependent Variable $\mathbf{G}\mathbf{Q}$ (Industry Heterogeneity, Dimensionless)	$\mathbf{G}\mathbf{Q}$ (Scale Heterogeneity, Dimensionless)	$\mathbf{G}\mathbf{Q}$ (Regional Heterogeneity, Dimensionless)
$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.125 *** (4.18)	0.132 *** (4.56)	0.145 *** (4.98)
$\mathrm{R}\mathrm{D}\mathrm{P}\times \mathrm{H}\mathrm{i}\mathrm{g}\mathrm{h}-\mathrm{t}\mathrm{e}\mathrm{c}\mathrm{h}$ (Dummy)	0.089 *** (3.12)	-	-
$\mathrm{R}\mathrm{D}\mathrm{P}\times \mathrm{S}\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{l}$ (Dummy)	-	0.078 *** (2.89)	-
$\mathrm{R}\mathrm{D}\mathrm{P}\times \mathrm{E}\mathrm{a}\mathrm{s}\mathrm{t}$ (Dummy)	-	-	0.065 ** (2.56)
High-tech/Small/East (Dummy)	0.052 * (1.68)	0.048 * (1.62)	0.055 ** (2.32)
Adj-${\mathrm{R}}^2$	0.405	0.398	0.389

Note: * p < 0.1, ** p < 0.05, *** p < 0.01. High-tech = 1 for high-tech industries and 0 for non-high-tech industries; Small = 1 for small and medium-sized enterprises and 0 for large enterprises; East = 1 for eastern regions and 0 for non-eastern regions.

4.6. Endogeneity Test Results and Robustness Verification

To alleviate the potential endogeneity problems of the model (such as two-way causality and omitted variables), the instrumental variable method (2SLS) and the propensity score matching method (PSM) are adopted for testing [26], and the results are shown in Table 13.

Instrumental variable method: the industry-year mean value of R&D investment persistence ($\mathrm{R}\mathrm{D}{\mathrm{P}}_{\mathrm{I}\mathrm{n}\mathrm{d}}$) and its two-period lagged value ($\mathrm{R}\mathrm{D}{\mathrm{P}}_{\mathrm{L}2}$) are selected as instrumental variables. The first-stage regression shows that the instrumental variables are significantly positively correlated with $\mathrm{R}\mathrm{D}\mathrm{P}$, and the F-value is 45.28, much higher than the critical value of 10, passing the weak instrumental variable test; To test the exclusivity constraint of the instrumental variables, this study further controls the industry-year R&D investment total (%), industry competition degree (Herfindahl index) and other variables in the first-stage regression, and the results show that the F-value is still 38.65, much higher than 10, and the coefficient of the instrumental variables on $\mathrm{R}\mathrm{D}\mathrm{P}$ is still significantly positive at the 1% level. Although the industry-year mean value may be affected by industry demand shocks, the above processing has alleviated the dispute of the exclusivity constraint to a certain extent. In the second-stage regression, the coefficient of $\mathrm{R}\mathrm{D}\mathrm{P}$ on $\mathrm{G}\mathrm{Q}$ is 0.215, which is statistically significant at the 1% level, still significantly positive, indicating that after alleviating endogeneity, the positive impact of R&D investment persistence on firm growth quality is still valid.

Propensity score matching method: taking the annual median of the comprehensive indicator of R&D investment persistence as the boundary, the sample is divided into the treatment group (high persistence) and the control group (low persistence). Enterprise size, enterprise age, ownership concentration and other variables are selected as matching variables, and nearest neighbor matching (1:1) is adopted for matching. The standardized bias after matching is less than 10%, meeting the parallel trend assumption; the average treatment effect (ATT) after matching is 0.192, which is statistically significant at the 1% level, indicating that the growth quality of enterprises with high R&D investment persistence is significantly higher than that of enterprises with low persistence, further verifying the robustness of the benchmark regression results.

• Supplementary endogenous treatment methods:

• DID model based on R&D policy shock: This study takes the implementation of R&D subsidy policies for SRUN enterprises in some provinces as a quasi-natural experiment, constructs a DID model with the policy implementation year as the time dummy variable and the enterprises in the pilot provinces as the treatment group. The regression results show that the coefficient of the policy dummy variable × time dummy variable is 0.156 (significant at 1%), indicating that the implementation of the R&D subsidy policy has significantly improved the R&D investment persistence of the treatment group enterprises, and further significantly improved the growth quality, which further verifies the positive correlation between R&D investment persistence and growth quality by using the exogenous policy shock.
• Regression of core explanatory variables with two-period lag: To further alleviate the reverse causality problem (the improvement of enterprise growth quality affects the future R&D investment persistence), this study regresses the two-period lagged value of $\mathrm{R}\mathrm{D}\mathrm{P}$ on $\mathrm{G}\mathrm{Q}$. The results show that the coefficient of ${\mathrm{R}\mathrm{D}\mathrm{P}}_{\mathrm{L}2}$ is 0.142 (significant at 1%), and the Adj-R² is 0.315, indicating that the positive correlation between R&D investment persistence and enterprise growth quality is still significant after considering the two-period lag, which further alleviates the reverse causality problem.

Table 13 is optimized to supplement the dimension/unit of each variable, and the specific revised content is as follows:

Table 13. Endogeneity Test Results.

Test Method	Dependent Variable $\mathbf{G}\mathbf{Q}$	Core Independent Variable	Coefficient	t/Z-Value	N	Adj-${\mathbf{R}}^2$/ATT
Instrumental Variable Method (2SLS)	Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.215 ***	5.89	9207	0.365
Propensity Score Matching (PSM)	Dimensionless	ATT (Treatment group − Control group)	0.192 ***	6.15	8964	0.192
DID Model (R&D Policy Shock)	Dimensionless	$\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\times \mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}$ (Dummy)	0.156 ***	4.96	8542	0.332
Regression with Two-period Lagged Core Explanatory Variable	Dimensionless	$\mathrm{R}\mathrm{D}{\mathrm{P}}_{\mathrm{i}\mathrm{j}\mathrm{t}-2}$ (Dimensionless)	0.142 ***	4.68	7985	0.315

Note: *** p < 0.01; The instrumental variables for 2SLS are the industry-year mean value of R&D investment persistence and its two-period lagged value; PSM adopts the 1:1 nearest neighbor matching method.

Table 13. Endogeneity Test Results.

Test Method	Dependent Variable $\mathbf{G}\mathbf{Q}$	Core Independent Variable	Coefficient	t/Z-Value	N	Adj-${\mathbf{R}}^2$/ATT
Instrumental Variable Method (2SLS)	Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.215 ***	5.89	9207	0.365
Propensity Score Matching (PSM)	Dimensionless	ATT (Treatment group − Control group)	0.192 ***	6.15	8964	0.192
DID Model (R&D Policy Shock)	Dimensionless	$\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\times \mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}$ (Dummy)	0.156 ***	4.96	8542	0.332
Regression with Two-period Lagged Core Explanatory Variable	Dimensionless	$\mathrm{R}\mathrm{D}{\mathrm{P}}_{\mathrm{i}\mathrm{j}\mathrm{t}-2}$ (Dimensionless)	0.142 ***	4.68	7985	0.315

Note: *** p < 0.01; The instrumental variables for 2SLS are the industry-year mean value of R&D investment persistence and its two-period lagged value; PSM adopts the 1:1 nearest neighbor matching method.

4.7. Robustness Test Results and Core Conclusion Validation

Four methods are adopted for robustness tests, and the results all verify the stability of the benchmark conclusions [27] (Table 8):

Replacing variable measurement methods: Replacing profitability with return on assets (ROA) and R&D investment intensity with the proportion of R&D investment in total assets, recalculating $GQ$ and $RDP$, and the regression results show that the coefficient of $RDP$ for $GQ$ is 0.178, which is statistically significant at the 1% level, and is still significantly positive. In addition, this study adopts the ADF test of R&D intensity time series stability and the 3-year growth rate fluctuation of R&D investment amount as the alternative indicators of R&D stability to re-regress the benchmark model. The results show that the coefficients are 0.169 and 0.158 respectively, both significant at the 1% level, which verifies the consistency of the results.

Adjusting the sample scope: excluding enterprises listed after 2020 (to avoid the impact of the epidemic) and only retaining manufacturing enterprises, the coefficient is 0.182, which is statistically significant at the 1% level, and the result is robust.

Lagged one-period regression: the independent variable (R&D investment persistence, RDP) and the mediating variable (technology conversion efficiency, TTE) were lagged by one period to mitigate the problem of reverse causality. The regression results showed that the coefficient of RDP was 0.165, which was significantly positive at the 1% level (p < 0.01), with a t-value of 5.28, a sample size (N) of 8156, and an adjusted R-squared (Adj-R²) of 0.305.

More stringent winsorization: all continuous variables were winsorized at the 0.5th and 99.5th percentiles to further eliminate the impact of extreme values. The regression coefficient of RDP was 0.185, which was also significantly positive at the 1% level (p < 0.01), with a t-value of 6.12, a sample size (N) of 9207, and an adjusted R-squared (Adj-R²) of 0.321.

Supplementary robustness test of mediating effect with alternative TTE measurement: This study uses patent authorization amount/R&D personnel number and new product income/main business income (%) as alternative indicators to calculate TTE, and re-conducts the three-step mediating effect test and Bootstrap test. The results show that the mediation effect ratios of $\mathrm{R}\mathrm{D}\mathrm{P}$, $\mathrm{R}\mathrm{D}\mathrm{I}\mathrm{n}\mathrm{t}$ and $\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ are 32.5%, 34.1% and 35.2% respectively, and the 95% confidence intervals of the indirect effects do not contain 0, indicating that the partial mediating effect of technology conversion efficiency is still robust after replacing the TTE measurement method.

The above robustness test results indicate that the core conclusion of this study—that R&D investment persistence is significantly positively correlated with the growth quality of SRUN listed companies and that technology conversion efficiency plays a partial mediating role—is highly stable. This conclusion is not affected by differences in variable measurement methods, sample scope, or regression approaches, thus confirming its reliability and generalizability.

Table 14. Robustness Test Results.

Robustness Test Method	Dependent Variable $\mathbf{G}\mathbf{Q}$	Core Independent Variable	Coefficient	t-Value	N	Adj-${\mathbf{R}}^2$
Replacing Variable Measurement Methods	New Measurement, Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{P}$ (New Measurement, Dimensionless)	0.178 ***	5.92	9207	0.318
Replacing R&D Stability Measurement (ADF Test)	Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{t}\mathrm{d}$ (ADF, Dimensionless)	0.169 ***	5.78	9207	0.309
Adjusting Sample Scope (Excluding Firms Listed after 2020)	Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.182 ***	5.85	7896	0.322
Regression with One-period Lagged Independent and Mediating Variables	Dimensionless	$\mathrm{R}\mathrm{D}{\mathrm{P}}_{\mathrm{L}1}$ (Dimensionless)	0.165 ***	5.28	8156	0.305
More Stringent Winsorization (0.5%/99.5%)	Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.185 ***	6.12	9207	0.321
Replacing the TTE Measurement Method	Dimensionless	$\mathrm{R}\mathrm{D}\mathrm{P}$ (Dimensionless)	0.172 ***	5.65	9207	0.312

Note: *** p < 0.01; All regressions control for firm/year fixed effects and all control variables; $\mathrm{R}\mathrm{D}{\mathrm{P}}_{\mathrm{L}1}$ refers to the one-period lagged R&D investment persistence.

is optimized to supplement the dimension/unit of each variable, and the specific revised content is as follows:

5. Research Conclusions and Implications

Taking A-share specialized, refined, characteristic, and innovative (SRUN) listed companies from 2006 to 2024 as the research sample, this study systematically examines the impact of R&D investment persistence on corporate growth quality, explores the mediating role of technology conversion efficiency in depth, and analyzes the heterogeneous characteristics of this impact. Based on the above research results, this study summarizes the corresponding theoretical and practical implications.

5.1. Research Summary

• Core Conclusions

The study confirms three key core conclusions, with clear applicable boundaries:

First, there is a significantly positive correlation between R&D investment persistence and the growth quality of Chinese A-share manufacturing SRUN listed companies (excluding non-manufacturing, ST/*ST, delisted firms and those established for less than three years). This conclusion is not applicable to non-listed SRUN enterprises or non-manufacturing SRUN-related entities.

Second, the two core dimensions of R&D investment persistence—investment intensity and investment stability—can independently and positively drive the improvement of corporate growth quality. Both dimensions exert significant positive impacts on the three core dimensions of growth quality (revenue growth, profitability, risk resilience), and the positive explanatory power of investment intensity (regression coefficient: 0.156) is slightly stronger than that of investment stability (regression coefficient: 0.132).

Third, from a practical economic perspective, the improvement effect of R&D investment persistence on SRUN firms’ growth quality is substantial: a 1 percentage point increase in R&D investment intensity is associated with a 0.189 percentage point rise in ROE and a 0.215 percentage point growth in revenue growth rate; when R&D investment persistence is raised from the mean to the 75th percentile, firms achieve an average 0.192 percentage point increase in revenue growth rate, a 0.168 percentage point rise in ROE, and a 0.091 percentage point improvement in risk resilience (1 minus asset-liability ratio). These improvement amplitudes are 1.5 to 2 times higher than the average level of A-share non-SRUN firms, confirming that the sustained and healthy growth of SRUN firms depends on long-term and stable R&D investment rather than short-term utilitarian R&D behaviors.

2.

Mechanism Conclusions

Technology conversion efficiency plays a partial mediating role in the relationship between R&D investment persistence and the growth quality of A-share manufacturing SRUN listed companies. Specifically, R&D investment persistence indirectly enhances corporate growth quality by improving patent output and new product conversion efficiency.

The mediating effect ratio varies by the dimension of R&D investment persistence: 34.2% for overall R&D investment persistence, 35.8% for investment intensity, and 36.1% for investment stability. Notably, the mediating effect accounts for the highest proportion in the correlation between R&D investment persistence and corporate profitability, highlighting that technology conversion is a key bridging link between R&D investment and profit growth for SRUN enterprises. Bootstrap test results (5000 repetitions) show that the 95% confidence intervals of the indirect effects do not contain 0, and the mediating effect ratio presents an increasing trend with the improvement of R&D investment persistence, further verifying the robustness of this transmission mechanism.

3.

Heterogeneity Conclusions

The positive correlation between R&D investment persistence and the growth quality of A-share manufacturing SRUN listed companies exhibits significant heterogeneous characteristics, which are mainly reflected in three dimensions:

First, the effect is more pronounced in high-tech industries (interaction term coefficient: 0.089), driven by the high R&D dependence and mature technology transformation industrial chain of high-tech industries.

Second, the positive explanatory power is stronger for small and medium-sized SRUN enterprises (interaction term coefficient: 0.078), with financing friction being the core driving channel—enterprises with higher financing constraints or lower financial slack rely more on sustained R&D investment to improve growth quality.

Third, enterprises located in eastern China show a more significant correlation (interaction term coefficient: 0.065), benefiting from abundant scientific and technological resources, developed financial markets and lower financing friction in eastern regions.

Additionally, supplementary heterogeneity tests show that the positive effect is more prominent for enterprises with high technological capability, and stable R&D investment (without positive/negative R&D shocks) has a stronger protective effect on growth quality. All heterogeneous conclusions are consistent with the characteristics of A-share manufacturing SRUN listed companies and are not universally applicable to other types of enterprises.

It should be noted that although endogeneity issues are alleviated through instrumental variables, propensity score matching (PSM), difference-in-differences (DID) and other methods, this study is limited by data availability and cannot fully identify the strict causal effect between R&D investment persistence and SRUN enterprises’ growth quality. Thus, the conclusions are empirical evidence based on conditional correlation, and their generalization to non-listed SRUN enterprises or non-manufacturing sectors should be cautious.

5.2. Research Implications

Based on the above research findings, this study advances the following theoretical and practical implications.

5.2.1. Theoretical Implications

First, the study enriches the theoretical connotation and measurement system of R&D investment persistence in innovation and R&D investment theory. Breaking the limitation of existing research that mostly measures R&D persistence from a single perspective of R&D smoothing or time-series continuity, this study precisely defines R&D investment persistence from the dual dimensions of investment intensity and stability, and constructs a comprehensive measurement indicator through principal component analysis. This dual-dimensional deconstruction not only conforms to the R&D investment characteristics of SRUN enterprises that rely on core technology iteration and niche market competition but also supplements the multi-dimensional measurement perspective of R&D persistence in existing literature, further improving the theoretical definition and empirical measurement system of R&D investment persistence.

Second, the study improves the transmission chain theory of R&D input to firm growth and fills the research gap in the intrinsic mechanism of R&D persistence affecting high-quality firm growth. Existing studies have mostly verified the direct impact of R&D investment on firm performance, but the intermediate transmission link between R&D investment persistence and firm growth quality remains unclear. This study identifies technology conversion efficiency as the partial mediating variable between the two, and explicitly uncovers the transmission logic of R&D investment persistence (intensity/stability) → technology conversion efficiency (patent output/new product conversion) → firm growth quality (revenue growth/profitability/risk resilience). It quantifies the mediating effect ratio of different dimensions of R&D persistence, which perfects the theoretical chain of R&D input driving high-quality firm growth and provides a more detailed mechanism explanation for the relationship between R&D investment and firm growth in innovation economics.

Third, the study expands the heterogeneous research framework of R&D investment effects in firm growth theory and deepens the understanding of the boundary conditions of R&D persistence driving high-quality growth. On the basis of verifying the heterogeneous effects across industrial attributes, firm size and regional distribution, this study further conducts heterogeneity tests from technological capability and R&D shock direction, and more importantly, carries out direct empirical testing on the financing friction channel underlying the heterogeneity of SMEs. This research design not only systematically reveals the multi-dimensional heterogeneous characteristics of the effect of R&D investment persistence on SRUN enterprises’ growth quality, but also clarifies the core driving channel of heterogeneous effects, which makes the heterogeneous research of R&D investment effects more in-depth and hierarchical, and enriches the theoretical research on the boundary conditions and action channels of R&D investment driving firm growth.

Fourth, the study constructs an integrated theoretical analysis framework combining Schumpeter’s Innovation Theory, Resource-Based View and Resource Dependence Theory, and expands the application scenarios of classic theories in the research of SRUN enterprises’ high-quality growth. Different from the single theoretical application in existing studies, this study links the three classic theories with the dual dimensions of R&D investment persistence in a targeted manner: taking the Resource-Based View as the theoretical support for the investment intensity dimension, and Schumpeter’s Innovation Theory and Resource Dependence Theory as the dual theoretical support for the investment stability dimension. This targeted integration of theories not only provides a solid and detailed theoretical foundation for the research hypotheses, but also combines classic management and economics theories with the unique growth characteristics of SRUN enterprises in emerging markets, expanding the application scenarios and research boundaries of the above theories in the field of specialized, refined, unique and novel enterprise research.

Fifth, the study supplements the theoretical research on high-quality growth of SRUN enterprises and enriches the micro-foundation theory of manufacturing high-quality development. As the core micro subject of manufacturing transformation and upgrading, SRUN enterprises lack targeted theoretical research on the driving factors of their high-quality growth. This study takes Chinese A-share manufacturing SRUN listed companies as the research sample, systematically verifies the core driving effect of R&D investment persistence on their growth quality, and clarifies the mechanism and heterogeneous characteristics of this effect. It fills the theoretical gap of R&D investment driving SRUN enterprises’ high-quality growth, and provides micro-theoretical evidence for the research on the high-quality development of the real economy and manufacturing industry.

5.2.2. Practical Implications

The practical implications are proposed at two levels: the enterprise level and the government level, as detailed below.

• For SRUN Enterprises

First, enterprises must firmly establish the core philosophy of sustained R&D investment and construct a dual R&D investment system characterized by intensity plus stability. They should formulate long-term R&D plans aligned with their development strategies and market demand, ensure the adequacy and stability of R&D input, and prioritize increasing research efforts in core technology fields to lay a solid foundation for technological innovation.

Second, optimize the financing structure to alleviate financing constraints and provide financial safeguards for persistent R&D investment. Enterprises can expand financing channels through equity financing, sci-tech credit, and government R&D subsidies, increase their holdings of financial slack, and mitigate the impact of financing frictions on the continuity of R&D investment. For SMEs with severe financing constraints, strengthening cooperation with large enterprises and research institutions to share R&D costs is an effective way to reduce financing pressure.

Third, focus on enhancing technology conversion efficiency and forge a closed-loop mechanism of R&D investment → technology conversion → growth improvement. Enterprises should improve the technology conversion management system; strengthen cross-departmental coordination and collaboration; increase investment in pilot testing and industrialization; attach importance to the layout, operation and commercialization of patents; and accelerate the transformation of technological achievements into practical productivity.

Fourth, formulate differentiated R&D and growth strategies based on their own development characteristics: high-tech SRUN enterprises can focus on deploying research in cutting-edge technology fields; SMEs can deepen research into “bottleneck” technologies in segmented niche markets; enterprises with weak technological capabilities should strengthen the introduction and training of R&D personnel to accumulate internal technical capacity; and SRUN enterprises in non-eastern regions can enhance cross-regional technological cooperation and resource linkage to offset their inherent resource disadvantages.

• For Government Departments

First, introduce targeted support policies, increase financial subsidies and tax incentives for SRUN enterprises, and improve the sci-tech finance service system to effectively alleviate enterprises’ R&D funding pressure and provide policy and financial guarantees for their sustained R&D investment. Governments should improve the sci-tech finance system by launching special R&D credit products for SRUN SMEs, reducing their R&D financing costs, and increasing credit lines and loan discounts for R&D activities of enterprises with high financing constraints. Meanwhile, advance the development of the multi-level capital market, support qualified SRUN enterprises to list and raise capital, and expand their equity financing channels.

Second, build professional technology conversion service platforms, integrate various scientific and technological resources, promote in-depth industry-university-research collaboration, and provide full-process technology conversion services for enterprises to help them improve technology conversion efficiency. Specifically, establish regional technology transfer centers to offer services such as pilot testing, achievement evaluation, and market docking for SRUN enterprises; encourage universities and research institutes to set up special teams for cooperative R&D with SRUN enterprises, and share the costs of technology conversion; subsidize enterprises’ expenses in patent application, maintenance, and industrialization, and reward high-efficiency technology conversion projects.

Third, implement differentiated support strategies, prioritizing support for high-tech industries, SRUN SMEs, and SRUN enterprises in non-eastern regions, and strive to narrow the development gaps across regions, firm sizes and industries. For SRUN SMEs with severe financing frictions, set up special R&D funds and provide interest-free or low-interest loans; for non-eastern regions, increase the tilt of central financial transfer payments and guide the flow of scientific and technological resources to narrow the regional gap in R&D support; for high-tech industries, intensify subsidies for core technology R&D and establish industrial innovation alliances to promote collaborative innovation.

Fourth, optimize the distribution mechanism of public innovation support and avoid the “clustered allocation” of R&D subsidies. Work on competitive grant allocation notes that higher-intensity competition can lead to clustering or strategic behavior in proposal/topic selection (Mugerman et al., 2025) [38]. This reinforces the need for policy design that rewards long-term R&D persistence rather than short-term project targeting. Governments should adopt a performance-oriented and differentiated subsidy allocation model centered on R&D investment persistence and technology conversion efficiency, with specific policy designs as follows:

• Establish a multi-dimensional evaluation index system for subsidy allocation: Take R&D investment persistence (weight 40%, measured by the 3-year average R&D intensity and investment stability), technology conversion efficiency (weight 40%, measured by patent output per unit R&D investment and new product sales revenue ratio), and growth quality (weight 20%, measured by profitability, revenue growth rate, and risk resilience) as core evaluation indicators, and score enterprises annually to determine subsidy qualification and quota.
• Implement phased and sustained subsidy support: For enterprises with stable R&D investment (R&D intensity fluctuation within ±10% for three consecutive years) and continuously improving technology conversion efficiency, adopt a “3-year continuous subsidy + dynamic adjustment” model. The initial subsidy amount is determined based on the evaluation score, and the subsidy intensity is increased by 10–15% annually if the indicators meet the standards, to guide enterprises to carry out long-term R&D.
• Set up special subsidies for segmented niche markets: Allocate no less than 30% of SRUN enterprise R&D subsidies to enterprises focusing on core technology gaps in segmented fields (such as key components, special materials, and professional equipment), and prohibit the use of subsidies for hot-field utilitarian R&D projects to avoid redundant construction.
• Establish a post-subsidy supervision and feedback mechanism: Require enterprises to submit quarterly R&D progress reports and annual subsidy use efficiency reports; conduct regular audits on the use of subsidy funds, and recover subsidies for enterprises that falsify R&D data or fail to meet the evaluation indicators for two consecutive years; link subsidy allocation with the feedback results to form a closed-loop management of “allocation-usage-evaluation-adjustment”.

Fifth, continuously improve the innovation ecosystem by strengthening intellectual property protection, perfecting talent introduction, training and incentive mechanisms, optimizing the market competition environment, and fully stimulating the innovation vitality and development momentum of SRUN enterprises. Specifically, strengthen the law enforcement of intellectual property rights, establish a rapid dispute resolution mechanism for SRUN enterprises’ core patents; introduce targeted talent policies such as housing subsidies and R&D incentives to attract high-end R&D personnel; optimize the market access environment for SRUN enterprises, and give priority to purchasing products and services of SRUN enterprises in government procurement projects; establish an innovation credit evaluation system to link R&D performance with enterprise credit ratings, and provide more convenient policy support for high-credit innovation enterprises.

5.3. Limitations and Future Research Directions

This study has several limitations, which also point out directions for future research:

• The research sample is limited to A-share SRUN listed companies, excluding non-listed SRUN enterprises. Future research can expand the sample scope to include non-listed enterprises to explore the differences in the correlation between R&D investment persistence and the growth quality of SRUN enterprises with different listing statuses.
• The measurement of technology conversion efficiency only covers two dimensions: patent output and product conversion, without considering sub-dimensions such as the efficiency, effect, and speed of technology conversion. Future research can construct a more comprehensive evaluation system for technology conversion efficiency, and conduct in-depth exploration on the differences in the mediating role of different dimensions of technology conversion between R&D investment persistence and corporate growth quality.
• This study does not explore the role of moderating variables. Future research can introduce moderating variables such as corporate governance level, innovation culture, and policy support intensity to examine their moderating effects on the relationship between R&D investment persistence and corporate growth quality, thus further enriching the research framework.
• The correlation between R&D investment persistence and corporate growth quality is only explored from a static perspective. Future research can adopt dynamic panel models, threshold regression models and other methods to explore the threshold effect of R&D investment persistence on corporate growth quality from a dynamic perspective and clarify the optimal interval of R&D investment persistence for SRUN enterprises.
• The study fails to identify a completely exogenous shock of R&D investment persistence. Although instrumental variables, PSM, DID, and other methods alleviate partial endogeneity problems, strict causal identification is not achieved. Future research can combine quasi-natural experiments (such as the gradual implementation of R&D tax preferential policies and the pilot of the Science and Technology Innovation Board) to further verify the causal effect between R&D investment persistence and the growth quality of SRUN enterprises.

In summary, persistent R&D investment is the core driver for elevating the growth quality of SRUN listed companies, with technology conversion efficiency serving as the critical transmission link in this process. Against the backdrop of the national strategy to achieve self-reliance and self-improvement in science and technology, SRUN enterprises must commit to sustained R&D investment and strive to enhance their technology conversion efficiency. For government authorities, it is imperative to deliver targeted policy support and advance the development of a robust innovation ecosystem. These joint efforts will effectively boost the high-quality growth of SRUN enterprises, thereby furnishing a solid underpinning for the transformation and upgrading of China’s manufacturing sector, as well as the independence and controllability of its industrial and supply chains.