State Capital Refocusing and Innovation-Driven Development: Evidence from Beijing Municipal SOEs

Abstract

This study examines whether state capital refocusing promotes innovation-driven development by reshaping the scale and technological direction of innovation in state-owned enterprises (SOEs). Using 499 firm-year observations for 42 Beijing municipal SOEs from 2013 to 2024, we distinguish innovation output from innovation allocation across invention-oriented, green, and internationally oriented patents. Two-way fixed-effects estimates show that a one-standard-deviation increase in core business focus is associated with increases of 0.610, 0.468, 0.383, and 0.090 log points in overall, invention-oriented, green, and internationally oriented innovation output, respectively. Refocusing also raises the shares of invention-oriented and green innovation, while leaving the international share unchanged. The main patterns remain broadly robust to alternative measures, additional controls, alternative specifications, and a continuous-treatment difference-in-differences design based on the 2020–2022 SOE Reform Three-Year Action Plan. Mechanism analyses provide evidence consistent with industrial-chain control as an important transmission channel, while heterogeneity analyses show that functional missions, industrial-chain positions, and strategic-sector status condition the effects. The findings indicate that state capital refocusing influences not only how much SOEs innovate, but also where innovation resources are directed, providing firm-level evidence on how public-capital reallocation can support mission-oriented innovation.

1. Introduction

For emerging economies, innovation-driven development has become central to industrial upgrading, technological catch-up, and the search for more resilient growth. State-owned enterprises (SOEs) occupy a distinctive institutional position in this process. Foundational research on privatization and ownership emphasizes how state control, governance arrangements, and incentive structures shape firm behavior, while more recent work conceptualizes SOEs as hybrid organizations operating under both market and political logics (Megginson & Netter, 2001; Bruton et al., 2015; Musacchio et al., 2015; Tihanyi et al., 2019). In China, state ownership, political connections, and government objectives beyond commercial returns can shape firms’ strategies, innovation incentives, and internationalization in ways that differ from private-sector behavior (Li et al., 2018; Cuervo-Cazurra et al., 2023; Zhong & Zheng, 2025). SOEs therefore provide a useful setting for examining how shifts in state capital allocation shape firm behavior and the direction of innovation-led industrial upgrading.

Yet public ownership does not automatically make state capital a developmental force. When SOEs expand into too many weakly related business lines, capital, technical personnel, managerial attention, and organizational resources can become over-dispersed. Such diversification may enlarge firm scale in the short run, but it can also weaken technological depth, reduce innovation efficiency, and dilute the developmental function of state capital. Research on diversification and strategic fit has long shown that business scope affects firms’ innovation search and capability deployment (Kim, 2013), while recent evidence suggests that specialization can redirect innovation activity toward domains in which firms hold comparative advantages (Han et al., 2025). Related work further shows that firms differ not only in how much they invent, but also in where they are positioned in technological space and how differentiated their technology portfolios are (Arts et al., 2023, 2025). In this sense, business-scope adjustment is not merely a managerial decision; it may reshape the allocation of innovation resources across technological domains (Ocasio, 1997; Kaul, 2012; Vidal & Mitchell, 2018).

China’s SOE reform has evolved through several related but distinct stages. Earlier reforms emphasized corporatization, selective privatization and exit, the consolidation of large state-controlled groups, and the introduction of mixed ownership. These changes improved governance and operating performance, but also preserved substantial heterogeneity in state mandates, resource allocation, and productivity (Aivazian et al., 2005; Hsieh & Song, 2015; Berkowitz et al., 2017; Y. He & Yang, 2021). More recent reforms have shifted the emphasis from changing ownership alone toward managing state capital and clarifying where it should be deployed. They reorient state capital toward principal businesses, key industries, public-service functions, and strategic sectors. We use state capital refocusing to describe this broader policy-driven process. At the enterprise level, it is implemented through core-business refocusing and observed empirically through core business focus (CBF), the share of operating revenue generated by officially approved principal-business segments. This policy construct–organizational process–empirical measure hierarchy links the reform logic to a measurable firm-level variable (SASAC, 2023; Aghion et al., 2015; Mazzucato, 2018; Karo, 2018).

If state capital refocusing is to advance innovation-driven development, its effects should be assessed not only by the scale of patenting, but also by the technological composition of SOE innovation portfolios. Recent work on innovation measurement similarly emphasizes that innovation is multidimensional and that single indicators capture only selected inputs, stages, outputs, or impacts (Todorov et al., 2024). Aggregate patent counts capture the volume of observable inventive activity, but they do not show whether innovation is deepening core technologies, supporting green transformation, or extending firms’ technological capabilities into global markets. We therefore distinguish three domains of patent-based innovation. Invention-oriented patents reflect technological depth and cumulative capability building; green patents capture the alignment between innovation and low-carbon transformation; and internationally oriented patents indicate whether domestic technological capabilities are being extended into overseas patent systems and global technology markets. This distinction matters because innovation domains are not interchangeable. Recent research on green patenting, environmental innovation, and technological catch-up shows that institutional conditions and policy environments shape not only how much firms innovate, but also which technologies they pursue (L. H. Cohen et al., 2026; Gebhardt et al., 2026; Zhu & Li, 2025; Acemoglu et al., 2012; Aghion et al., 2016).

This perspective exposes a gap in the literature. One strand of SOE reform research examines how privatization, corporatization, and restructuring affect governance, efficiency, and firm performance (Megginson & Netter, 2001; Aivazian et al., 2005; Hsieh & Song, 2015; Berkowitz et al., 2017). A more recent strand studies how ownership change, mixed-ownership reform, and partial privatization influence aggregate innovation through interest alignment, agency-cost reduction, and governance incentives (W. He et al., 2020; X. Zhang et al., 2020; Pan et al., 2022). These studies establish that reform can affect whether and how much firms innovate, but leave open whether a policy-induced redefinition of business scope changes where SOE innovation is directed. At the same time, innovation research increasingly shows that patent counts, innovation portfolios, positioning in technology space, project selection, and influential innovation capture different dimensions of firm innovation rather than a single aggregate outcome (Arts et al., 2023, 2025; Böttcher & Klingebiel, 2024; Shi et al., 2025). Bringing these two strands of research together leads to a more specific question: does state capital refocusing merely increase innovation output, or does it also redirect SOE innovation portfolios toward development-oriented technological domains?

The answer is not theoretically obvious. On the one hand, refocusing may promote innovation by reducing resource dispersion, strengthening capability accumulation, improving project selection, and enhancing coordination within core industrial chains. These mechanisms are especially relevant for SOEs because their principal businesses are often tied to functional missions and policy priorities. On the other hand, excessive concentration may narrow exploratory search, reinforce path dependence, or crowd out emerging activities that fall outside the current definition of the core business. The effect of refocusing should therefore depend on how well a firm’s principal business aligns with its functional mission, industrial-chain position, and external technological environment. We incorporate these competing considerations by examining both average effects and heterogeneous effects across SOE missions, chain positions, and strategic-sector status (March, 1991; Feldman, 2014; Vidal & Mitchell, 2018).

To answer this question, we use a firm-level panel of Beijing municipal SOEs from 2013 to 2024, covering 42 firms and 499 firm-year observations. Beijing provides a useful empirical setting because its municipal SOEs underwent a clearly documented process of core-business refocusing during the 2020–2022 SOE Reform Three-Year Action Plan, when the number of approved principal businesses was reduced from 117 to 59. These firms are also formally classified into competitive, special-function, and urban public-service groups, operate across energy, transportation, construction, information technology, culture, and urban public services, and occupy different positions along industrial chains. While this setting does not allow for universal claims about all SOEs or all developing economies, it offers a coherent institutional environment in which a clearly defined refocusing policy can be linked to firm-level innovation behavior. Empirically, we measure core business focus as the share of operating revenue generated by principal business segments, distinguish between innovation output and the technological composition of patent portfolios, and combine two-way fixed-effects models with robustness checks and reform-based continuous-treatment difference-in-differences evidence.

We find that stronger core business focus is associated with higher innovation output, especially in invention-oriented and green innovation. More importantly, refocusing also reshapes SOE innovation portfolios: green innovation accounts for a larger share, invention-oriented innovation rises more modestly, and the share of internationally oriented innovation does not change significantly. Mechanism analyses suggest that industrial-chain control helps translate refocusing into innovation gains and portfolio adjustment, while heterogeneity analyses show that these effects vary across functional missions, industrial-chain positions, and strategic-sector status.

This study makes three contributions. First, it advances research on SOE reform and innovation by shifting attention from the quantity of innovation to its technological composition. Existing studies have mainly examined how ownership change, privatization, and mixed-ownership arrangements affect aggregate innovation output or firm performance (W. He et al., 2020; X. Zhang et al., 2020; Pan et al., 2022). By distinguishing innovation output from innovation allocation, this study shows that state capital refocusing is not only associated with more patenting, but also with a selective reorientation of SOE innovation portfolios toward development-oriented domains, especially technological upgrading and green transition.

Second, the study clarifies a transmission channel through which business-scope adjustment is linked to innovation outcomes. Existing explanations often emphasize internal resource concentration and strategic fit. We show that these internal mechanisms operate together with industrial-chain control, which captures a firm’s capacity to coordinate upstream and downstream actors, standards, and application scenarios around its principal business. This perspective links internal resource reallocation with external chain coordination and provides a more complete account of how refocusing can support innovation in complex institutional environments (Adner, 2017; Jacobides et al., 2018; E. Liu, 2019).

Third, the study highlights the boundary conditions under which refocusing translates into innovation outcomes. The developmental effect of refocusing is not uniform across SOEs; it depends on the alignment between a firm’s narrowed business scope and its institutional mandate, industrial-chain position, and strategic-sector status. By showing how innovation resources are reallocated differently across competitive, special-function, and urban public-service SOEs, as well as across different chain positions, the study provides firm-level evidence on how macro-level state-capital policies are translated into heterogeneous innovation responses.

2. State Capital Refocusing and Innovation-Driven Development: Institutional Logic and Hypotheses

This section develops the institutional logic linking state capital refocusing to the scale and direction of SOE innovation. We argue that refocusing is not simply a managerial move toward specialization. Rather, it represents a development-oriented reallocation of state capital toward domains in which SOEs have clearer mandates, accumulated capabilities, and industrial-chain responsibilities. Core business focus is the firm-level manifestation of this broader policy process. This framing complements SOE reform research centered on ownership and governance by shifting attention to the redefinition of business boundaries and the internal allocation of resources (W. He et al., 2020; X. Zhang et al., 2020; Pan et al., 2022). The theoretical argument centers on three mechanisms: resource concentration, capability accumulation, and industrial-chain coordination. These mechanisms help explain why refocusing may affect not only innovation output, but also the technological domains toward which innovation resources are directed (Bruton et al., 2015; Musacchio et al., 2015; Mazzucato, 2018).

2.1. State Capital Refocusing in China’s SOE Reform

2.1.1. From Diversified Expansion to Development-Oriented State Capital Refocusing

To understand how core business focus may shape SOE innovation, it is necessary to situate refocusing within the broader trajectory of China’s state-sector reform. China has not followed a single path of wholesale privatization. Its reforms have combined corporatization without complete privatization, selective exit of smaller SOEs, consolidation of large state-controlled groups, and mixed-ownership arrangements (Aivazian et al., 2005; Hsieh & Song, 2015; Berkowitz et al., 2017; Y. He & Yang, 2021). During the earlier decades of market transition, diversification was not merely a result of managerial drift. For many SOEs, it was a rational response to policy uncertainty, incomplete factor markets, and the need to stabilize revenue while market institutions were still developing. Expanding into multiple business lines helped firms spread risk and, in some cases, gain access to scarce resources (Bruton et al., 2015; Musacchio et al., 2015).

Over time, however, diversified expansion created a structural constraint. Business portfolios became broader than the technological and organizational capabilities needed to sustain them. Capital, engineering talent, managerial attention, and investment projects were spread across units with limited strategic or technological relatedness. The result was not only weaker operational focus, but also a thinner foundation for cumulative innovation in core technologies (Ocasio, 1997; Kim, 2013; Kaul, 2012).

Against this reform background, the turn toward core business focus reflects a deeper change in the logic of state-capital allocation. Earlier reform waves primarily altered ownership structures, governance arrangements, or the legal form of SOEs. More recent reforms increasingly define the substantive boundaries within which state capital should operate by linking business scope to the functional missions that SOEs are expected to perform. They encourage state capital to concentrate in principal businesses, key industries, public-service fields, and strategic sectors. This shift is better understood as state capital refocusing: a policy-driven process that redirects public capital from weakly related expansion toward core domains with stronger developmental value (Hsieh & Song, 2015; Berkowitz et al., 2017; Y. He & Yang, 2021; Mazzucato, 2018; Aghion et al., 2015).

Core business focus is therefore conceptually distinct from state capital refocusing. State capital refocusing refers to the broader policy process through which public capital is redirected toward development-oriented domains. Core business focus is its observable firm-level expression, capturing the extent to which a firm’s revenue-generating activities are anchored in its approved principal businesses.

This policy logic is consistent with recent evidence that institutions and industrial policy shape innovation by affecting both firm incentives and the fit between policy priorities and firm capabilities (Donges et al., 2023; Hua et al., 2025; H. Liu & Zhou, 2025). It also gives refocusing a more precise theoretical meaning. The key issue is not whether a firm is diversified or specialized in a generic strategic sense. Rather, it is whether state capital is concentrated in domains where the firm has a public mandate, accumulated capabilities, and a clearer role in national or urban development. The move from diversification to refocusing therefore marks a shift from horizontal expansion across industries to vertical deepening within core domains (Aghion et al., 2015; E. Liu, 2019; Mazzucato, 2018).

2.1.2. Functional Missions and the Direction of Innovation Resources

Core business focus helps explain why innovation resources may become more concentrated. Functional classification helps explain where those resources are likely to flow. This distinction matters because Chinese SOEs do not form a homogeneous ownership category. Research on state capitalism shows that SOEs differ in the configuration of state control, market discipline, political objectives, and organizational autonomy (Bruton et al., 2015; Musacchio et al., 2015; Tihanyi et al., 2019). Recent evidence from China further indicates that SOEs’ responses to innovation policy are shaped by state-related institutional logics and may differ systematically from those of private firms (Gong et al., 2023). Since the mid-2010s, China’s SOE reform agenda has increasingly emphasized differentiated development, supervision, responsibility, and assessment according to firms’ functional positions. In Beijing’s municipal SOE system, this logic is institutionalized through a three-way classification of competitive, special-function, and urban public-service SOEs.

Functional missions matter because state-guided innovation does not operate through national priorities alone. It depends on how those priorities are translated into action through government coordination, enterprise mandates, and industrial networks. Recent evidence from China shows that state coordination can catalyze innovation networks, that coordinated governance arrangements can improve innovation performance when supported by sufficient information capacity, and that SOEs may respond substantively to government innovation priorities when these are embedded in organizational mandates (Dai et al., 2024; Feng & Jiang, 2024; Gong et al., 2023; Mazzucato, 2018; Karo, 2018; E. Liu, 2019; Adner, 2017).

Competitive SOEs face more direct market pressure and therefore retain greater discretion in technological choices, commercialization, and business selection. Special-function SOEs undertake investment, operational, and strategic tasks in key sectors, making technological upgrading and industrial-chain support central to their mandates. Urban public-service SOEs are more closely tied to city operations, service quality, environmental upgrading, and public welfare. Once refocusing concentrates resources in principal businesses, these mission differences should shape not only whether innovation increases, but also the technological domains toward which innovation resources are redirected.

2.1.3. Mechanisms and Boundary Conditions

State capital refocusing links business-scope adjustment to innovation by reorganizing resources around development-oriented tasks. This reorganization unfolds along three dimensions. First, it changes internal resource allocation by channeling capital, technical talent, managerial attention, and organizational effort toward principal businesses. Second, it lengthens the time horizon of innovation, making sustained investment and cumulative learning in core domains more feasible. Third, it reshapes the external coordination network of innovation by giving SOEs a clearer basis for coordinating upstream and downstream actors, research organizations, standards, and application scenarios around core technological tasks (Ocasio, 1997; Dierickx & Cool, 1989; Teece et al., 1997; Adner, 2017; Jacobides et al., 2018).

Refocusing is therefore not expected to promote innovation simply because firms become more specialized. It matters when a narrower business scope is translated into a stronger organizational and industrial foundation for technological upgrading. The innovation effect of refocusing is consequently conditional rather than automatic. A formal narrowing of business scope may have limited consequences if resources remain dispersed, if principal businesses are defined only for administrative compliance, or if the firm lacks the capabilities needed to deepen innovation in its approved core domains. Conversely, refocusing is more likely to generate innovation gains when principal businesses align with functional missions, accumulated capabilities, industrial-chain positions, and policy priorities (March, 1991; Feldman, 2014; Vidal & Mitchell, 2018).

The theoretical question is therefore not whether specialization is generally beneficial, but under what conditions state capital refocusing becomes a substantive reallocation of resources toward development-oriented innovation. The hypotheses below build on this logic by examining innovation output, domain-specific innovation, and the allocation of innovation across technological domains.

2.2. Hypothesis Development

2.2.1. Hypothesized Effects on Innovation Output

Prior SOE reform research shows that changes in ownership and governance can improve innovation by aligning interests, reducing agency costs, and strengthening organizational control (W. He et al., 2020; X. Zhang et al., 2020; Pan et al., 2022). State capital refocusing addresses a different but complementary organizational friction: the fragmentation of capital, technical personnel, and managerial attention across weakly related business lines. Under diversified structures, dispersed business units tend to prioritize immediate operational returns over long-term and uncertain technological exploration. Refocusing may ease this structural constraint through the joint operation of resource concentration and capability accumulation (Ocasio, 1997; Kaul, 2012; Shi et al., 2025).

By reducing noncore operations, refocusing brings dispersed resources back into principal business domains and creates a more concentrated base of key R&D inputs. More importantly, it lengthens the time horizon for innovation. This logic is consistent with the resource-based view, which emphasizes that strategic capabilities cannot be built through short bursts of investment, but require sustained and path-dependent learning (Dierickx & Cool, 1989). For SOEs, refocusing aligns public capital with approved core domains and can turn it into a form of patient capital. This institutional buffer helps protect long-term R&D from the pull of short-lived commercial opportunities (Barney, 1991; Manso, 2011; Howell, 2017; Belenzon & Cioaca, 2025).

Clearer principal-business boundaries can also reduce internal coordination costs. They make it easier for firms to organize internal resources and mobilize external complementary assets around specific technological bottlenecks, a central concern in the dynamic-capabilities perspective (Teece et al., 1997). By providing more stable resource support, a longer horizon for learning, and clearer coordination boundaries, state capital refocusing should increase overall innovation output (Adner, 2017; Jacobides et al., 2018).

Hypothesis 1 (H1).

State capital refocusing increases the overall innovation output of SOEs.

2.2.2. State Capital Refocusing and Domain-Specific Innovation Output

Invention-oriented, green, and internationally oriented innovation face different structural constraints. State capital refocusing may therefore promote domain-specific innovation through related but distinct mechanisms.

Invention-Oriented Innovation

Invention-oriented innovation is constrained above all by the need for sustained capability accumulation. Because it pushes firms beyond established technological trajectories, it requires tolerance for uncertainty, repeated experimentation, and the ability to absorb long periods of potential failure. Under diversified structures, such deep exploration is often interrupted by shifting business priorities (March, 1991; Manso, 2011; Shi et al., 2025).

For invention-oriented innovation, refocusing operates primarily through the capability accumulation mechanism. By concentrating financial resources, technical talent, and experimental infrastructure within core technological fields, refocusing aligns R&D investment with long-term developmental mandates. This concentration can turn state capital in core domains into patient capital, providing a stable platform for path-dependent learning and protecting deep technological exploration from operational disruptions. Recent evidence confirms that sustained in-house scientific research remains an important foundation for fundamental technological breakthroughs (Arora et al., 2024). Refocusing provides the institutional continuity needed to sustain this long-term and uncertain exploration (Howell, 2017; Belenzon & Cioaca, 2025; Arora et al., 2024).

Hypothesis 2a (H2a).

State capital refocusing increases the invention-oriented innovation output of SOEs.

Green Innovation

For SOEs, a central barrier to green innovation is not necessarily an absolute shortage of resources. It is that environmental technologies may be marginalized under diversified structures. When business portfolios span unrelated areas, green upgrading is often treated as a peripheral compliance cost, disconnected from core revenue-generating activities and supported only by intermittent residual budgets (Acemoglu et al., 2012; Aghion et al., 2016; Wang & Jiang, 2021).

Refocusing changes this dynamic mainly through resource concentration. By reducing noncore operations, refocusing frees capital and technical talent that can be reallocated to core operational tasks. For SOEs in energy, infrastructure, and urban public services, environmental upgrading is not external to the business. It is embedded in core production systems and service provision. Refocusing aligns green objectives with principal businesses and can transform environmental upgrading from a passive regulatory response into a substantive technological capability (Wang & Jiang, 2021; Mazzucato, 2018).

Institutional environments shape the direction of environmental innovation (Park et al., 2024; Gebhardt et al., 2026). Concentrated resource allocation provides the organizational and material support needed to sustain green R&D within the firm’s strategic core. By anchoring green technologies in principal businesses, refocusing helps internalize environmental priorities and should increase green innovation output (Acemoglu et al., 2012; Aghion et al., 2016).

Hypothesis 2b (H2b).

State capital refocusing increases the green innovation output of SOEs.

Internationally Oriented Innovation

Internationally oriented innovation reflects the extension of domestic technological capabilities into global markets. It may benefit from refocusing because resource concentration provides a more coherent domestic technological base for overseas patenting and cross-border technology deployment (Chang et al., 2024; Cuervo-Cazurra et al., 2023).

However, internationally oriented innovation is more exposed to external constraints than domestic invention or green upgrading. It depends not only on internal technological capacity, but also on market access, global intellectual-property systems, international collaboration networks, and geopolitical conditions. These constraints are especially salient in the context of technological dependence and U.S.–China technology decoupling (Jiang et al., 2020; Han et al., 2024; Zhu & Li, 2025). For SOEs, international innovation also involves balancing commercial objectives with government mandates in internationalization (Cuervo-Cazurra et al., 2023).

Refocusing can build a stronger domestic technological foundation, but it cannot by itself remove these external institutional and geopolitical barriers. It therefore provides a necessary but insufficient basis for internationally oriented innovation. Its effect is expected to be weaker, and more dependent on external conditions, than its effects on invention-oriented and green innovation.

Hypothesis 2c (H2c).

State capital refocusing increases the internationally oriented innovation output of SOEs.

2.2.3. Hypothesized Effects on Innovation Allocation

Beyond increasing the scale of innovation, state capital refocusing may also change how innovation is distributed across technological domains. The distinction between scale and structure is central to this study. If refocusing raises all types of patenting proportionally, its role is mainly to expand innovation activity. If it changes the relative weight of different domains, then it reshapes the strategic direction of SOE innovation (Klingebiel & Rammer, 2014; Böttcher & Klingebiel, 2024).

Under diversified structures, innovation portfolios often emerge as the cumulative outcome of separate business units responding to their own operational needs. The resulting portfolio is fragmented and reactive, reflecting past expansion more than a deliberate development strategy. Research shows that innovation performance depends critically on how resources are allocated across projects (Klingebiel & Rammer, 2014), and that firms’ technological portfolios can have strategic consequences for subsequent competitive behavior (Arts et al., 2023, 2025; Böttcher & Klingebiel, 2024; Ocasio, 1997).

State capital refocusing changes this allocation logic. As noncore activities are reduced, innovation investment becomes more closely tied to the technological demands embedded in core areas. For SOEs, these demands are shaped by functional missions, industrial-chain responsibilities, and policy priorities. Refocusing therefore redirects innovation resources away from dispersed business expansion and toward technologies with clearer developmental significance. This composition effect is analytically distinct from a simple output effect (Mazzucato, 2018; E. Liu, 2019; Adner, 2017).

The direction of this reallocation should vary by firm mandate. SOEs in basic materials and strategic infrastructure are more likely to increase the relative weight of invention-oriented innovation. Firms responsible for urban operations and public services are more likely to shift a larger share of innovation effort toward green technologies. Firms with overseas market-development needs may place greater emphasis on internationally oriented innovation, although its relative share may respond less strongly because of external constraints. This directional logic is consistent with evidence that SOEs translate government innovation priorities through state-related institutional mandates rather than through market incentives alone (Gong et al., 2023). If this allocation mechanism operates, the empirical pattern should not appear only as an increase in total patenting. It should also appear as a structural realignment of SOE innovation portfolios around mandated technological needs (Mazzucato, 2018; Karo, 2018; Wang & Jiang, 2021; Chang et al., 2024).

Hypothesis 3 (H3).

State capital refocusing alters the relative allocation of SOE innovation across invention-oriented, green, and internationally oriented domains.

3. Data, Variables, and Research Design

3.1. Sample and Data

This study constructs a firm-level panel of Beijing municipal SOEs for the period 2013–2024. This period covers the 2020–2022 SOE Reform Three-Year Action Plan, during which the number of approved principal businesses in the Beijing municipal SOE system was reduced from 117 to 59. The reform therefore generated observable variation in the extent to which firms adjusted their business portfolios. Beijing municipal SOEs are also formally classified into competitive, special-function, and urban public-service categories, which provides an institutional basis for examining mission-based heterogeneity without relying on ex post statistical grouping. The sample covers energy, transportation, construction, information technology, culture, and urban public services.

These institutional features shape the empirical setting of the study. The single-city sample reduces variation in regulatory rules, ownership supervision, performance assessment, and reform timing. The empirical design therefore relies primarily on within-firm changes in core business focus under a relatively homogeneous institutional environment. The sample is not statistically representative of all Chinese SOEs. Its value lies instead in allowing state capital refocusing to be examined in a setting where functional missions, approved principal businesses, and firm-level business-scope adjustments are observable with unusual clarity.

The data are assembled from annual reports, credit-rating reports, patent databases, and disclosures by the Beijing State-owned Assets Supervision and Administration Commission. Segment revenues and approved principal-business scopes were hand-collected. Patent data were obtained from the China National Intellectual Property Administration and global intellectual-property databases, and were matched to firms using standardized names and historical name changes. Firm-year observations are retained only when segment revenue, total operating revenue, patent outcomes, and key financial controls can be consistently matched. Continuous variables are winsorized at the 1st and 99th percentiles. The final sample consists of 499 firm-year observations from 42 firms and forms a mildly unbalanced panel.

3.2. Variable Definitions and Measurement

3.2.1. Innovation Output and Innovation Allocation

Consistent with the theoretical argument, the dependent variables capture two dimensions of firm innovation: innovation output and innovation allocation. Innovation output measures the scale of patenting, whereas innovation allocation measures the composition of patenting across technological domains. This distinction separates changes in innovation volume from changes in technological orientation.

Patent applications are used as the baseline measure because they provide firm-level, technology-specific information that is consistently observable over the sample period. Prior research shows that patent counts, patent value, and firms’ positions in technology space capture different dimensions of innovation (Kogan et al., 2017; Arts et al., 2023; Zhu & Li, 2025). In this study, patent outcomes serve as consistent, technology-specific proxies for observable innovation activity (Hall et al., 2005).

Innovation-output variables are constructed as the natural logarithm of one plus the number of patent applications filed by a firm in a given year. Overall innovation, denoted as Innova, is measured by total patent applications. Invention-oriented innovation, denoted as Ori_Innova, is measured by invention patent applications, which are examined for novelty, inventiveness, and practical applicability under China’s patent system (CNIPA, 2022). Green innovation, denoted as Gre_Innova, is measured by green patent applications identified using the WIPO IPC Green Inventory (WIPO, 2024). Internationally oriented innovation, denoted as Inter_Innova, is measured by international patent applications and captures the positioning of firm technologies in overseas patent systems.

Innovation-allocation variables capture the relative salience of each technological domain within a firm’s patent portfolio. The three share variables, Ori_Share, Gre_Share, and Inter_Share, are defined respectively as invention, green, and international patent applications divided by total patent applications. For firm-years with no patent applications, the share variables are set to zero to preserve the panel structure. The three categories are not mutually exclusive. An invention patent may also be green or international. The share variables therefore measure domain salience rather than exhaustive portfolio components that sum to one.

Together, these variables allow the empirical models to distinguish expansion in patent output from changes in the relative allocation of patenting across technological domains.

3.2.2. Core Explanatory Variable

The core explanatory variable is core business focus (CBF), the firm-level measure of core-business refocusing defined conceptually in Section 2.1.1. It captures the extent to which a firm’s realized operating revenue is concentrated in officially approved principal-business segments rather than noncore activities. It is defined as follows:

$${\mathrm{CBF}}_{it}={\displaystyle \frac{{\mathrm{Revenue}\mathrm{from}\mathrm{core}\mathrm{business}\mathrm{segments}}_{it}}{{\mathrm{Total}\mathrm{operating}\mathrm{revenue}}_{it}}}$$

(1)

The construction of CBF proceeds in three steps. First, annual business-segment revenue is hand-collected from corporate disclosures. Second, approved principal-business segments are identified using formal institutional sources, including the firm’s functional classification, Beijing SASAC approvals, annual reports, and substantive business descriptions. A segment is classified as a principal business because it falls within the officially approved functional and business scope, not because of contemporaneous profitability, revenue growth, or managerial preference. When segment labels change over time, they are harmonized according to the approved scope and underlying business activities. Third, revenue generated by the identified principal-business segments is aggregated and divided by total operating revenue. CBF ranges from 0 to 1, with higher values indicating greater concentration in approved principal businesses.

This revenue-weighted measure captures the realized economic significance of approved principal businesses rather than the formal number of business lines. A simple segment count would assign equal weight to a small peripheral activity and a business line responsible for most of the firm’s operations. Expressed as a revenue share, CBF is comparable across firms with different sizes and reporting structures. General revenue changes affect CBF only when they alter the relative contribution of approved principal businesses and noncore activities. Robustness analyses use lagged CBF and an ordered CBF measure to assess sensitivity to the operational definition.

The policy interpretation of CBF rests on the formally approved business boundaries described in Section 3.1 and the differential pre-reform exposure used in Section 4.5. Because principal-business status is institutionally determined rather than inferred from profitability or revenue growth, CBF captures the realized business-structure outcome associated with refocusing, not a generic indicator of operating efficiency. The DID design separately uses firms’ pre-reform noncore-business burden to examine whether stronger streamlining pressure is followed by larger changes in innovation outcomes.

3.2.3. Control Variables

The baseline models include firm-level controls with established links to firms’ innovation capacity. Firm age, denoted as Age, captures life-cycle differences in organizational learning, accumulated routines, technological experience, and innovation propensity (Huergo & Jaumandreu, 2004). Firm size, denoted as Size, captures the scale of the resource base and the ability to spread the fixed costs of R&D across a larger volume of activity (W. M. Cohen & Klepper, 1996). Sales growth, denoted as Growth, captures changes in market expansion and demand conditions that may affect firms’ incentives and opportunities to innovate (Piva & Vivarelli, 2007).

ROA and operating cash flow capture related but distinct dimensions of firms’ capacity to support innovation. ROA reflects operating performance and profit-generating capacity, whereas CFO more directly measures the internally generated funds available to finance R&D. This distinction matters because innovation investment is risky, long-term, and largely intangible, making it particularly dependent on internal finance (Himmelberg & Petersen, 1994; Brown et al., 2009; Hall & Lerner, 2010). Leverage, denoted as Lev, captures debt pressure and financial flexibility. Listed status, denoted as Listed, reflects access to public equity financing and exposure to capital-market monitoring, both of which may shape managerial incentives to undertake risky, long-horizon innovation (Brown et al., 2009; Aghion et al., 2013).

Ownership structure, denoted as Es, and ownership hierarchy, denoted as Eh, capture differences in state control, ownership concentration, monitoring arrangements, managerial incentives, and access to strategic resources. These factors are especially relevant in China, where ownership structure affects both firms’ access to innovation inputs and the efficiency with which those inputs are converted into innovation output (Choi et al., 2011; Zhou et al., 2017). Higher values of Es indicate a stronger degree of direct state ownership.

Table 1. Variable Definitions.

Category	Variable	Symbol	Definition
Dependent variables: Innovation output	Overall innovation	Innova	Natural logarithm of one plus the total number of patent applications filed by a firm in a given year
Dependent variables: Innovation output	Invention-oriented innovation	Ori_Innova	Natural logarithm of one plus the number of invention patent applications filed by a firm in a given year
Dependent variables: Innovation output	Green innovation	Gre_Innova	Natural logarithm of one plus the number of green patent applications filed by a firm in a given year
Dependent variables: Innovation output	Internationally oriented innovation	Inter_Innova	Natural logarithm of one plus the number of international patent applications filed by a firm in a given year
Dependent variables: Innovation allocation	Share of invention-oriented innovation	Ori_Share	Invention patent applications divided by total patent applications
Dependent variables: Innovation allocation	Share of green innovation	Gre_Share	Green patent applications divided by total patent applications
Dependent variables: Innovation allocation	Share of internationally oriented innovation	Inter_Share	International patent applications divided by total patent applications
Core explanatory variable	Core business focus	CBF	Revenue from approved principal-business segments divided by total operating revenue
Control variables	Firm age	Age	Natural logarithm of firm age
Control variables	Firm size	Size	Natural logarithm of total assets
Control variables	Sales growth	Growth	Change in operating revenue relative to operating revenue in the previous year
Control variables	Return on assets	ROA	Net profit divided by total assets
Control variables	Operating cash flow	CFO	Net cash flow from operating activities divided by total assets
Control variables	Leverage	Lev	Total liabilities divided by total assets
Control variables	Listed status	Listed	Equals 1 if the firm is listed and 0 otherwise
Control variables	Ownership structure	Es	Equals 1 for state-controlled firms and 2 for wholly state-owned firms
Control variables	Ownership hierarchy	Eh	Number of layers in the control chain

Notes: Higher values of Es indicate a stronger degree of direct state ownership.

summarizes the definitions and measurement of all variables used in the empirical analysis.

3.3. Baseline Model Specification

The baseline relationship is estimated using a two-way fixed-effects model:

$$Y_{it}=\alpha +\beta {\mathrm{CBF}}_{it}+\gamma X_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

where i and t index firms and years, respectively. Y_it denotes an innovation-output or innovation-allocation measure. CBF_it is core business focus, measured as the share of operating revenue generated by approved principal-business segments. X_it is the vector of firm-level controls. μ_i and λ_t denote firm and year fixed effects, respectively. Standard errors are clustered at the firm level.

The fixed-effects structure matches the empirical setting described above. Firm fixed effects absorb time-invariant differences across firms, while year fixed effects absorb shocks common to all firms in a given year. The coefficient of interest, β, is identified from within-firm changes in core business focus over time. In the innovation-output models, β captures the association between core business focus and patenting scale. In the innovation-allocation models, β captures whether core business focus changes the relative weight of a given technological domain in the patent portfolio.

3.4. Identification Strategy

The baseline model identifies the relationship between within-firm changes in CBF and changes in innovation outcomes after accounting for time-invariant firm heterogeneity, common annual shocks, and observable changes in firms’ financial and governance conditions. Because contemporaneous fixed-effects estimates alone do not establish causal ordering, the analysis uses a layered empirical strategy. Lagged CBF improves temporal sequencing, while the 2020–2022 SOE Reform Three-Year Action Plan provides a policy-based source of variation for reassessing the relationship.

Using the reform setting detailed in Section 4.5, firms’ standardized 2019 noncore-business burdens provide predetermined variation in exposure to streamlining pressure. This exposure is used to construct a continuous-treatment difference-in-differences design. A continuous-treatment event study examines differential pre-reform innovation trends and the timing of post-reform changes, while placebo tests assess whether comparable estimates arise when treatment intensity is randomly reassigned across firms. Firm-specific linear trends and extended time-varying controls provide additional checks on gradual firm trajectories and observable changes in firms’ operating environments. Together, these analyses strengthen the credibility of the empirical identification and allow the interpretation of the estimates to be calibrated to the assumptions supported by the research design (Baker et al., 2022; Callaway et al., 2024).

4. Empirical Results and Robustness Checks

4.1. Sample Characteristics and Descriptive Evidence

Table 2. Descriptive Statistics.

Variable	Mean	Standard Deviation	Minimum	Maximum
Innova	1.111	1.609	0.000	5.958
Ori_Innova	0.743	1.257	0.000	5.220
Gre_Innova	0.587	1.003	0.000	4.407
Inter_Innova	0.130	0.515	0.000	4.174
Ori_Share	0.153	0.284	0.000	1.000
Gre_Share	0.063	0.204	0.000	1.000
Inter_Share	0.018	0.108	0.000	0.786
CBF	0.899	0.080	0.720	1.000
Age	3.257	0.483	1.792	4.317
Size	6.585	1.883	3.523	10.657
Growth	0.140	0.248	−0.200	0.789
ROA	0.020	0.063	−0.080	0.150
CFO	0.027	0.078	−0.100	0.154
Lev	0.638	0.150	0.242	0.912
Listed	0.241	0.428	0.000	1.000
Es	1.783	0.412	1.000	2.000
Eh	2.540	0.499	2.000	3.000

reports the descriptive statistics. Core business focus (CBF) has a mean of 0.899 and a standard deviation of 0.080, with values ranging from 0.720 to 1.000. The high mean reflects the long-standing efforts of Beijing municipal SOEs to streamline business operations and concentrate on approved principal businesses. At the same time, the observed range and dispersion indicate that the sample retains meaningful variation in core business focus. The fixed-effects estimates below use within-firm changes over time to identify the relationship between refocusing and innovation.

The innovation-output variables show substantial dispersion across firms and years. The mean of overall innovation is 1.111, while the means of invention-oriented, green, and internationally oriented innovation are 0.743, 0.587, and 0.130, respectively. Patent activity is therefore concentrated mainly in invention-oriented and green technologies, whereas international patenting remains limited for the average firm. The innovation-allocation variables show a similar pattern. The mean shares of invention-oriented and green innovation are 0.153 and 0.063, respectively, compared with 0.018 for internationally oriented innovation. The maximum value of Inter_Share reaches 0.786, however, indicating that a small number of firms have developed a substantial international patent presence. Beijing municipal SOEs therefore differ not only in the scale of patenting, but also in the technological composition of their innovation portfolios.

4.2. Baseline Results and Economic Magnitudes

4.2.1. State Capital Refocusing and Innovation Output

Table 3. Baseline Regression Results for Innovation Output.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Innova	Ori_Innova	Gre_Innova	Inter_Innova	Innova	Ori_Innova	Gre_Innova	Inter_Innova
CBF	7.932 *** (0.782)	5.972 *** (0.681)	5.364 *** (0.531)	1.186 *** (0.297)	7.621 *** (0.926)	5.853 *** (0.723)	4.790 *** (0.577)	1.129 *** (0.339)
Age					0.704 *** (0.145)	0.499 *** (0.113)	0.264 *** (0.090)	0.117 ** (0.053)
Size					0.002 (0.044)	0.048 (0.035)	0.018 (0.028)	0.001 (0.016)
Growth					0.489 * (0.255)	0.241 (0.199)	0.372 ** (0.159)	0.053 (0.093)
ROA					1.329 (1.093)	1.192 (0.854)	−0.959 (0.681)	0.027 (0.400)
CFO					0.741 (0.787)	0.244 (0.615)	0.216 (0.491)	0.451 (0.288)
Lev					−1.737 *** (0.507)	−0.505 (0.396)	−0.993 *** (0.316)	−0.097 (0.185)
Listed					1.086 *** (0.173)	0.839 *** (0.135)	0.620 *** (0.108)	0.042 (0.063)
Es					0.185 (0.182)	0.309 (0.193)	−0.290 ** (0.113)	0.092 (0.067)
Eh					0.106 (0.148)	0.297 ** (0.116)	0.145 (0.092)	0.221 *** (0.054)
Controls	No	No	No	No	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499	499	499	499
Adjusted R2	0.313	0.238	0.253	0.045	0.352	0.293	0.288	0.062

Notes: Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. The same convention applies to subsequent tables.

reports the baseline fixed-effects estimates. Columns (1)–(4) include firm and year fixed effects, while columns (5)–(8) add firm-level controls. Across all specifications, the coefficient on CBF is positive and statistically significant. Since CBF measures the share of operating revenue generated by approved principal-business segments, the estimates indicate that stronger state capital refocusing is associated with higher overall, invention-oriented, green, and internationally oriented patent output.

The estimated effects are economically meaningful. In the controlled specifications, a one-standard-deviation increase in CBF corresponds to increases of 0.610, 0.468, 0.383, and 0.090 log points in overall, invention-oriented, green, and internationally oriented innovation, respectively. These magnitudes indicate sizable within-firm changes in observable innovation activity.

The ordering of the estimates is also informative. Invention-oriented innovation shows the strongest response, consistent with its reliance on sustained investment and cumulative learning in core technological fields. Green innovation also responds strongly, reflecting its close connection with production processes, equipment upgrading, and operational transformation within principal businesses. The coefficient for internationally oriented innovation is positive but smaller, consistent with the view that overseas patenting depends not only on internal resource concentration, but also on market access, international intellectual-property systems, and geopolitical conditions. After firm-level controls are included, the CBF coefficients decline slightly but remain statistically significant, supporting H1 and H2a–H2c.

4.2.2. State Capital Refocusing and Innovation Allocation

Table 4. Baseline Regression Results for Innovation Allocation.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	Ori_Share	Gre_Share	Inter_Share	Ori_Share	Gre_Share	Inter_Share
CBF	0.278 ** (0.118)	0.326 *** (0.104)	0.041 (0.043)	0.236 * (0.128)	0.285 ** (0.113)	0.032 (0.047)
Age				0.048 * (0.028)	0.035 (0.025)	0.012 (0.010)
Size				0.018 ** (0.009)	0.012 (0.008)	0.004 (0.003)
Growth				0.043 (0.052)	0.052 (0.046)	0.007 (0.019)
ROA				−0.067 (0.223)	−0.154 (0.196)	−0.066 (0.081)
CFO				0.177 (0.160)	0.142 (0.141)	0.042 (0.058)
Lev				−0.067 (0.103)	−0.089 (0.091)	−0.044 (0.038)
Listed				0.165 *** (0.035)	0.069 ** (0.031)	0.008 (0.013)
Es				0.018 (0.037)	−0.046 (0.033)	0.004 (0.014)
Eh				0.061 ** (0.030)	0.046 * (0.026)	0.026 ** (0.011)
Controls	No	No	No	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499	499
Adjusted R2	0.137	0.159	0.049	0.142	0.163	0.053

Notes: Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

turns to innovation allocation by examining whether state capital refocusing changes the technological composition of SOE patent portfolios. The dependent variables are the shares of invention-oriented, green, and internationally oriented patent applications in total patent applications. This specification distinguishes proportional growth in patenting from a compositional shift across technological domains.

The results show a selective change in innovation allocation. In the controlled models, the CBF coefficient is 0.236 for the invention-oriented share and is marginally significant at the 10% level. The coefficient is 0.285 for the green innovation share and is significant at the 5% level. By contrast, the coefficient for the internationally oriented share is small and statistically insignificant. In economic terms, a one-standard-deviation increase in CBF is associated with a 1.9 percentage-point increase in the invention-oriented share and a 2.3 percentage-point increase in the green innovation share, with no detectable change in the international share.

Taken together, Table 3 and Table 4 show that state capital refocusing is associated with both higher patent output and a selective reallocation of patenting across technological domains. The compositional shift is most evident for green innovation, weaker for invention-oriented innovation, and absent for internationally oriented innovation. The findings therefore provide qualified support for H3: refocusing reshapes innovation allocation, but the shift is concentrated in green and, to a lesser extent, invention-oriented technologies. This pattern motivates the mechanism and heterogeneity analyses that follow.

4.3. Extended Control Robustness Checks

R&D intensity, executive turnover, policy support, and industry conditions may affect both core business focus and innovation outcomes.

Table 5. Robustness to Additional Controls.

	Panel A: Innovation Output							Panel B: Innovation Allocation
Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
	Innova	Innova	Innova	Innova	Ori_Innova	Gre_Innova	Inter_Innova	Ori_Share	Gre_Share	Inter_Share
CBF	6.184 *** (0.897)	5.937 *** (0.891)	4.852 *** (0.842)	4.516 *** (0.836)	3.872 *** (0.654)	3.408 *** (0.571)	0.813 ** (0.327)	0.208 * (0.122)	0.254 ** (0.108)	0.027 (0.045)
RD_Intensity	3.246 *** (0.782)	3.189 *** (0.778)	2.864 *** (0.741)	2.752 *** (0.735)	2.146 *** (0.575)	1.623 *** (0.503)	0.487 * (0.289)	0.086 ** (0.042)	0.052 (0.037)	0.011 (0.015)
CEO_Change		−0.187 (0.142)	−0.174 (0.140)	−0.168 (0.139)	−0.126 (0.109)	−0.098 (0.095)	−0.031 (0.055)	−0.012 (0.015)	−0.008 (0.013)	−0.003 (0.005)
Policy_Sub			0.098 *** (0.018)	0.122 *** (0.018)	0.167 *** (0.014)	0.094 *** (0.012)	0.018 ** (0.007)	0.007 ** (0.003)	0.005 * (0.003)	0.001 (0.001)
Ind_Boom				0.426 * (0.253)	0.318 (0.198)	0.287 * (0.173)	0.062 (0.099)	0.018 (0.027)	0.024 (0.024)	0.004 (0.010)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499	499	499	499	499	499
Adjusted R2	0.358	0.360	0.364	0.366	0.293	0.292	0.072	0.142	0.163	0.053

Notes: Standard errors clustered at the firm level are reported in parentheses. Baseline controls include Age, Size, Growth, ROA, CFO, Lev, Listed, Es, and Eh. Additional controls are RD_Intensity, CEO_Change, Policy_Sub, and Ind_Boom, as defined in Section 4.3. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

therefore introduces four extended controls sequentially. RD_Intensity is measured as R&D expenditure divided by operating revenue and captures observable innovation input. CEO_Change equals 1 when the firm’s chief executive changes in a given year and captures changes in strategic leadership that may alter R&D priorities and the continuity of innovation programs (Ahmad et al., 2024). Policy_Sub is measured as government subsidies divided by total assets and captures policy-based financial support that may ease financing constraints (Howell, 2017). Recent evidence on Chinese industrial enterprises also shows that government industrial funds can promote innovation by narrowing equity-financing gaps and guiding patient social capital toward early-stage and long-term investment (Y. Zhang & Zhou, 2026). Ind_Boom is the annual revenue growth rate of the firm’s industry and captures changes in industry demand and technological opportunities (Piva & Vivarelli, 2007; Malerba & Orsenigo, 1996). Because R&D expenditure and government subsidies may themselves adjust as firms refocus their business scope, they are introduced sequentially as extended controls to assess coefficient stability rather than being treated as predetermined baseline covariates.

After observable differences in technological input, public support, and industry demand are taken into account, the CBF coefficient declines but remains positive and statistically significant in all innovation-output models. In the innovation-allocation models, the positive associations with the invention-oriented and green innovation shares persist, while the coefficient for the internationally oriented share remains insignificant. The baseline findings are therefore not driven by these observable differences in technological input, policy support, or industry conditions.

4.4. Alternative Measures and Model Specifications

Table 6. Robustness to Alternative Measures.

Panel A: Innovation Outcomes Measured by Granted Patents
Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Innova1	Ori_Innova1	Gre_Innova1	Inter_Innova1	Ori_Share1	Gre_Share1	Inter_Share1
CBF	6.838 *** (0.859)	4.214 *** (0.646)	3.415 *** (0.478)	1.220 *** (0.301)	0.221 * (0.131)	0.268 ** (0.116)	0.029 (0.049)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499	499	499
Adjusted R2	0.321	0.222	0.249	0.080	0.088	0.109	0.039
Panel B: Core Business Focus Lagged By One Year
Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Innova	Ori_Innova	Gre_Innova	Inter_Innova	Ori_Share	Gre_Share	Inter_Share
CBF_lag	7.148 *** (0.952)	5.427 *** (0.744)	4.536 *** (0.594)	1.063 *** (0.349)	0.224 * (0.132)	0.271 ** (0.117)	0.028 (0.048)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	457	457	457	457	457	457	457
Adjusted R2	0.274	0.271	0.274	0.063	0.098	0.126	0.043

Notes: Panel A replaces patent applications with granted patents. Panel B uses a one-year lag of CBF, the firm-level measure of state capital refocusing. Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

examines whether the results are sensitive to alternative measurement choices. Panel A replaces patent applications with granted patents. Because granted patents have passed substantive examination, they provide a more conservative measure of patent output. The main pattern remains unchanged: CBF is positively associated with all four innovation-output variables. In the allocation models, the coefficient remains positive for the green innovation share, marginally positive for the invention-oriented share, and insignificant for the internationally oriented share (Hall et al., 2005).

Panel B uses a one-year lag of CBF. This specification allows business-scope adjustment to be reflected in patenting outcomes with a delay and reduces concerns about simultaneity between current revenue structure and current patenting. The signs and significance patterns remain similar to the baseline estimates.

Table 7. Robustness to an Ordered Measure of Core Business Focus.

Variable	(1)	(2)	(3)	(4)
	Innova	Ori_Innova	Gre_Innova	Inter_Innova
CBF1	0.840 *** (0.106)	0.596 *** (0.084)	0.423 *** (0.068)	0.089 ** (0.039)
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Observations	499	499	499	499
Adjusted R2	0.268	0.217	0.236	0.053

Notes: The cutoff points are 0.85 and 0.95. CBF1 is coded 1 for low focus (CBF < 0.85), 2 for medium focus (0.85 ≤ CBF < 0.95), and 3 for high focus (CBF ≥ 0.95). ** and *** denote statistical significance at the 5% and 1% levels, respectively.

further recodes continuous CBF into an ordered categorical measure using cutoffs of 0.85 and 0.95. The ordered measure is positively associated with all four innovation-output variables. These checks indicate that the core findings do not depend on a single measurement approach for state capital refocusing.

4.5. Reform-Based Difference-in-Differences Evidence

The 2020–2022 SOE Reform Three-Year Action Plan provides a reform-based check on the main results. During this period, the number of approved principal businesses in the Beijing municipal SOE system was reduced from 117 to 59. Rather than treating the reform as a clean natural experiment, we use it as a policy-driven source of variation in refocusing pressure. Firms entered the reform with different pre-reform noncore-business burdens. Those with broader peripheral business portfolios faced greater pressure to streamline noncore activities and had more room to increase core business focus.

Treatment intensity, denoted as TreatIntensity_i, is defined as the standardized number of noncore business segments held by each firm in 2019, the year immediately before the reform. Noncore segments are identified using the same criteria as in the construction of CBF. This variable captures pre-reform exposure to streamlining pressure rather than realized post-reform adjustment. If some firms had already reduced noncore activities before 2019, the resulting measurement error would likely attenuate the estimates rather than exaggerate the reform effect.

The continuous-treatment DID specification is:

$$Y_{it}=\alpha +\theta ({\mathrm{Post}}_t\times {\mathrm{TreatIntensity}}_i)+\gamma X_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

where Post_t equals 1 for 2020 and later years and 0 otherwise. The coefficient of interest is θ, which captures whether firms with heavier pre-reform noncore-business burdens experienced larger post-reform changes in innovation outcomes.

The DID interpretation requires that firms with different treatment intensities would have followed parallel innovation trends in the absence of the reform. Figure 1 reports event-study estimates using 2019 as the reference year. The pre-reform coefficients for 2013–2018 fluctuate around zero, and their confidence intervals include zero. A joint F-test yields a p-value of 0.463, so the null of no differential pre-trends cannot be rejected. After 2020, the coefficients become positive and rise gradually, consistent with a delayed innovation response after noncore-business streamlining and resource reallocation.

Table 8. Difference-in-Differences Estimates.

Variable	Panel A: Innovation Output				Panel B: Innovation Allocation
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Innova	Ori_Innova	Gre_Innova	Inter_Innova	Ori_Share	Gre_Share	Inter_Share
Postt × TreatIntensityi	4.283 ** (1.876)	3.147 ** (1.392)	2.651 ** (1.218)	0.714 * (0.398)	0.148 * (0.081)	0.193 ** (0.088)	0.019 (0.033)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499	499	499
Adjusted R2	0.366	0.293	0.292	0.072	0.142	0.163	0.053

Notes: TreatIntensity_i is the standardized number of noncore business segments held by each firm in 2019. Post_t equals 1 for 2020 and later years and 0 otherwise. The event-study specification uses 2019 as the omitted reference year and jointly tests the pre-reform coefficients from 2013 to 2018. Standard errors clustered at the firm level are reported in parentheses. * and ** denote statistical significance at the 10% and 5% levels, respectively.

reports the DID estimates. In Panel A, the interaction term is positive and statistically significant for all four innovation-output measures. In Panel B, the interaction term is positive and significant for the invention-oriented and green innovation shares, but insignificant for the internationally oriented share. This distinction is important. Reform-driven streamlining increases the absolute volume of international patenting, but it does not increase the relative weight of internationally oriented patents in the portfolio. The DID results therefore mirror the baseline pattern: firms with heavier pre-reform noncore-business burdens experience larger post-2020 increases in patent output and a selective shift toward invention-oriented and green patenting.

Overall, the fixed-effects estimates, additional-control tests, alternative-measure checks, and reform-based DID evidence point to the same main pattern. Appendix A 

Table A1. Diagnostic Tests.

Diagnostic Test	Statistic	p-Value	Interpretation
Mean VIF	2.41	—	No serious multicollinearity
Maximum VIF	4.38	—	Below conventional threshold
Wooldridge test for serial correlation	21.76	<0.001	Serial correlation present
Modified Wald test for heteroskedasticity	352.40	<0.001	Heteroskedasticity present
Pesaran CD test	1.28	0.201	No strong cross-sectional dependence

Notes: The presence of serial correlation and heteroskedasticity supports the use of firm-clustered standard errors in the main regressions.

shows serial correlation and heteroskedasticity but no strong cross-sectional dependence, supporting the use of firm-clustered standard errors. Appendix A 

Table A2. Alternative Model Specifications for Innovation Output.

Model	Innova	Ori_Innova	Gre_Innova	Inter_Innova
Firm-specific trends	5.218 ***	4.332 ***	3.701 ***	0.735 *
	(1.214)	(0.936)	(0.812)	(0.421)
Poisson FE	2.480 ***	2.315 ***	2.046 ***	0.644 *
	(0.516)	(0.472)	(0.438)	(0.351)
Negative binomial with firm/year indicators	2.102 ***	1.934 ***	1.788 ***	0.512
	(0.612)	(0.548)	(0.506)	(0.382)
Baseline controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Observations	499	499	499	499

Notes: The table reports the coefficient on CBF. The firm-specific-trend specification retains the log(1 + patent applications) outcomes used in the baseline model, whereas the conditional Poisson fixed-effects and negative-binomial specifications with firm and year indicators use the corresponding untransformed patent counts. Standard errors clustered at the firm level are reported in parentheses. * and *** denote significance at the 10% and 1% levels, respectively.

further shows that the positive associations with overall, invention-oriented, and green innovation remain statistically significant under firm-specific trends, Poisson fixed effects, and negative-binomial models with firm and year indicators. The internationally oriented innovation coefficient remains positive and is significant under the firm-trend and Poisson specifications, but not under the negative-binomial specification, indicating that this comparatively small effect is less stable. Appendix A 

Table A3. Fractional Response Models for Innovation Allocation.

Variable	Ori_Share	Gre_Share	Inter_Share
CBF	0.018 *	0.022 **	0.003
	(0.010)	(0.009)	(0.004)
Baseline controls	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Firm-level controls	Yes	Yes	Yes
Observations	499	499	499

Notes: The table reports average marginal effects associated with a one-standard-deviation increase in CBF from fractional-response models. The models include the baseline firm-level controls and year indicators, with standard errors clustered at the firm level. They are used as functional-form robustness checks and do not reproduce the within-firm estimator of the baseline two-way fixed-effects specification. Because the share variables include boundary values of 0 and 1, fractional-response models are more appropriate than standard beta regression (Papke & Wooldridge, 1996). * and ** denote statistical significance at the 10% and 5% levels, respectively.

confirms positive marginal effects for the invention-oriented and green innovation shares, but not for the internationally oriented share. Appendix A 

Table A4. Placebo Tests for the Continuous-Treatment DID Design.

Outcome	Actual DID Coefficient	Mean Placebo Coefficient	Placebo SD	Empirical p-Value
Innova	4.283	0.042	1.305	0.028
Ori_Innova	3.147	0.036	1.102	0.034
Gre_Innova	2.651	0.029	0.968	0.041
Inter_Innova	0.714	0.011	0.406	0.086
Ori_Share	0.148	0.004	0.075	0.071
Gre_Share	0.193	0.006	0.081	0.044
Inter_Share	0.019	0.002	0.031	0.318

Notes: Treatment intensity is randomly reassigned across firms 1000 times. The empirical p-value is calculated as the share of placebo estimates that are larger than the actual estimate in absolute value.

shows that the placebo estimates are centered close to zero and that the empirical p-values reproduce the selective DID pattern. Taken together, the evidence is strongest and most consistent for overall, invention-oriented, and green innovation, while the international-output result should be interpreted more cautiously. The next section examines the mechanisms behind these patterns.

5. Mechanism, Heterogeneity, and Recognized Application-Oriented Outcomes

Section 4 shows that state capital refocusing is associated with higher SOE innovation output and a selective shift in innovation allocation. These findings raise three further questions. First, through what organizational channel does a narrower business scope translate into stronger innovation? Second, do the effects vary across SOEs with different mandates, chain positions, and strategic roles? Third, does refocusing remain confined to patenting, or is it also associated with formally recognized application-oriented outcomes? This section addresses these questions by examining industrial-chain control, structural heterogeneity, and recognized application-oriented outcomes.

5.1. Industrial-Chain Control as a Transmission Mechanism

The baseline estimates establish the relationship between core business focus and innovation but do not reveal the organizational channel through which it operates. Section 2.1.3 identifies industrial-chain coordination as the directly observable link in the proposed mechanism. Industrial-chain control captures the firm’s capacity to coordinate upstream suppliers, downstream users, strategic resources, technical standards, and application scenarios around its principal business (Adner, 2017; E. Liu, 2019).

This channel matters because concentrated resources generate innovation only when they are organized around interdependent suppliers, users, research institutions, and standards. Industrial-chain control enables the firm to act as a coordinating node in this wider innovation system. Recent research accordingly emphasizes ecosystem interdependence and the importance of alignment between policy priorities, firm capabilities, and industrial positions (C. Y. Baldwin et al., 2024; Dai et al., 2024; Hua et al., 2025; Jacobides et al., 2018).

This channel is also relevant to innovation allocation. Industrial-chain control places SOEs closer to task-specific technological needs. When principal businesses involve bottleneck technologies, equipment reliability, or core process improvement, stronger chain control is likely to direct resources toward invention-oriented patents. When principal businesses involve energy use, infrastructure operation, or environmental upgrading, it is more likely to support green patents. By contrast, internationally oriented patenting is less directly governed by domestic chain control because it depends more heavily on overseas market access, foreign intellectual-property systems, and external regulatory conditions. Industrial-chain control should therefore be most visible in the allocation of innovation toward invention-oriented and green domains (E. Liu, 2019; Mazzucato, 2018).

The channel is measured using a composite Industrial-Chain Control Index (CCI). The index captures four dimensions of coordination and influence: market position, control over strategic resources, supply-chain influence, and policy or standard-setting capacity. To improve measurement reliability, five independent experts evaluated the indicators. The intraclass correlation coefficient is 0.81, suggesting a high level of agreement across raters.

A sequential fixed-effects approach is used to evaluate this channel.

Table 9. Mechanism Analysis: Industrial-Chain Control.

		Panel A: Innovation Output				Panel B: Innovation Allocation
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	CCI	Innova	Ori_Innova	Gre_Innova	Inter_Innova	Ori_Share	Gre_Share	Inter_Share
CBF	0.969 *** (0.114)	4.010 *** (0.888)	4.079 *** (0.748)	3.209 *** (0.590)	0.771 ** (0.377)	0.163 (0.126)	0.176 (0.111)	0.025 (0.046)
CCI		3.729 *** (0.343)	1.832 *** (0.289)	1.632 *** (0.228)	0.370 ** (0.146)	0.075 * (0.040)	0.113 *** (0.035)	0.007 (0.015)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499	499	499	499
Adjusted R2	0.334	0.435	0.357	0.372	0.089	0.118	0.152	0.048

Notes: Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

reports that CBF is positively associated with the Industrial-Chain Control Index (β = 0.969, p < 0.01). After CCI is added to the innovation-allocation equations, the CBF coefficient declines from 0.236 to 0.163 for the invention-oriented share and from 0.285 to 0.176 for the green share, while CCI remains positively associated with both outcomes. This pattern is interpreted as evidence consistent with a transmission channel rather than as a fully identified causal indirect effect. KHB is designed primarily for coefficient-rescaling problems in nested nonlinear models, and a structural equation model would not provide an additional source of identifying variation in the present 42-firm panel. The sequential fixed-effects approach therefore provides the clearest channel test supported by the available data.

5.2. Heterogeneity Analyses

The mechanism based on industrial-chain control implies that the innovation effects of refocusing should vary across institutional and industrial contexts. A firm’s principal business is not only a revenue category. It also carries a functional mandate, a position in the industrial chain, and, in some cases, strategic significance. The following analyses examine these three dimensions as related conditions under which refocusing is more or less likely to translate into innovation.

5.2.1. Functional Missions

China’s differentiated SOE classification system embeds principal businesses in distinct mandate structures. Competitive SOEs operate under stronger market discipline. Special-function SOEs undertake strategic investment and policy-guided industrial development. Urban public-service SOEs support city operations and public welfare. This classification is consistent with research showing that SOE behavior reflects the interaction of state objectives, market constraints, and organizational autonomy, rather than ownership alone (Li et al., 2018; Tihanyi et al., 2019; Gong et al., 2023). A similar increase in core business focus may therefore be associated with different innovation responses because the technological demands embedded in these mandates differ.

Panel A of

Table 10. Heterogeneity by Functional Mission.

Panel A. Innovation Output

Variable

Competitive SOEs

Special-Function SOEs

Urban Public-Service SOEs

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Innova

Ori_ Innova

Gre_ Innova

Inter_ Innova

Innova

Ori_ Innova

Gre_ Innova

Inter_ Innova

Innova

Ori_ Innova

Gre_ Innova

Inter_ Innova

CBF

7.750 *** (1.702)

6.288 *** (1.391)

3.608 *** (1.046)

1.699 *** (0.571)

9.367 *** (1.836)

8.503 *** (1.462)

5.946 *** (1.028)

0.682 (0.705)

4.714 * (2.694)

0.935 (2.130)

5.539 *** (1.833)

0.600 (2.679)

Controls

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations

176

176

176

176

176

176

176

176

147

147

147

147

Adjusted R2

0.341

0.302

0.269

0.081

0.391

0.372

0.349

0.054

0.284

0.171

0.323

0.047

Panel B. Innovation Allocation

Variable

Competitive SOEs

Special-Function SOEs

Urban Public-Service SOEs

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Ori_Share

Gre_Share

Inter_Share

Ori_Share

Gre_Share

Inter_Share

Ori_Share

Gre_Share

Inter_Share

CBF

0.250 (0.185)

0.135 (0.163)

0.063 (0.068)

0.496 ** (0.196)

0.340 ** (0.173)

−0.040 (0.072)

−0.186 * (0.113)

0.519 *** (0.148)

0.075 (0.062)

Controls

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations

176

176

176

176

176

176

147

147

147

Adjusted R2

0.131

0.118

0.047

0.169

0.184

0.049

0.095

0.201

0.041

Notes: Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

shows that competitive SOEs experience broad patenting growth across all four innovation-output measures, consistent with market-oriented expansion. Special-function SOEs exhibit significant increases in overall, invention-oriented, and green innovation, whereas the internationally oriented coefficient is not statistically significant; the domain-specific gains are therefore concentrated in invention-oriented and green technologies, in line with their mandates for technological upgrading and strategic investment. Urban public-service SOEs display a narrower but more targeted response: overall and green innovation increase, while the invention-oriented and internationally oriented coefficients are not significant, and the largest domain-specific coefficient appears for green innovation. This pattern is consistent with their responsibilities in urban environmental infrastructure, public services, and low-carbon operations.

Panel B reports the corresponding changes in portfolio composition. For competitive SOEs, refocusing expands patent output without substantially altering the structure of the patent portfolio. For special-function SOEs, refocusing significantly increases the shares of invention-oriented and green patents. For urban public-service SOEs, the green innovation share rises, while the invention-oriented share declines. This does not necessarily indicate weaker innovation capacity. Rather, it suggests a portfolio shift toward green and service-related technologies that are more closely tied to public-service mandates. Functional missions therefore shape the direction in which refocusing is translated into innovation.

5.2.2. Positions Along the Industrial Chain

The innovation effects of refocusing also depend on the firm’s position along the industrial chain. Upstream firms are closer to basic inputs and foundational bottlenecks, so their innovation is more likely to involve process improvement, materials development, and core technical upgrading. Midstream firms connect upstream inputs with downstream applications and often innovate through engineering integration and system adaptation. Downstream firms are closer to users and application scenarios, making their innovation more responsive to demand differentiation and external market opportunities (Alfaro et al., 2019; E. Liu, 2019).

Panel A of

Table 11. Heterogeneity by Position Along the Industrial Chain.

Panel A. Innovation Output

Variable

Upstream

Midstream

Downstream

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Innova

Ori_ Innova

Gre_ Innova

Inter_ Innova

Innova

Ori_ Innova

Gre_ Innova

Inter_ Innova

Innova

Ori_ Innova

Gre_ Innova

Inter_ Innova

CBF

9.847 *** (1.726)

7.623 *** (1.348)

6.218 *** (1.075)

0.576 (0.632)

7.293 *** (1.358)

5.416 *** (1.061)

4.872 *** (0.847)

1.687 *** (0.497)

4.571 ** (1.894)

2.148 (1.479)

3.946 *** (1.180)

2.023 *** (0.694)

Controls

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations

133

133

133

133

207

207

207

207

159

159

159

159

Adjusted R2

0.392

0.361

0.336

0.052

0.377

0.331

0.314

0.083

0.286

0.195

0.273

0.071

Panel B. Innovation Allocation

Variable

Upstream

Midstream

Downstream

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Ori_Share

Gre_Share

Inter_Share

Ori_Share

Gre_Share

Inter_Share

Ori_Share

Gre_Share

Inter_Share

CBF

0.386 ** (0.178)

0.214 (0.131)

−0.032 (0.054)

0.412 *** (0.149)

0.168 (0.132)

0.028 (0.055)

−0.153 (0.197)

0.246 * (0.148)

0.183 *** (0.062)

Controls

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations

133

133

133

207

207

207

159

159

159

Adjusted R2

0.167

0.143

0.040

0.181

0.148

0.052

0.103

0.162

0.057

Notes: Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

shows these differences. Upstream firms exhibit significant gains in overall, invention-oriented, and green innovation, while the internationally oriented coefficient is not significant, reflecting their focus on foundational technologies and process upgrading. Midstream firms increase across all four innovation-output measures, consistent with their role as technical integrators. Downstream firms show significant increases in overall, green, and internationally oriented innovation, but not in invention-oriented innovation; among the domain-specific outcomes, the gains are therefore concentrated in green and internationally oriented technologies, consistent with their proximity to users, application scenarios, and external markets.

Panel B further shows that chain position shapes portfolio reorientation. For upstream firms, refocusing is associated with technical deepening and a significant increase in the invention-oriented share. For midstream firms, refocusing is accompanied by a moderate increase in the invention-oriented share, consistent with technology scaling and system integration. For downstream firms, refocusing shifts the portfolio toward green solutions and outward-facing technologies. The developmental value of state capital refocusing therefore depends not only on whether firms focus, but also on where the focused activities are located along the industrial chain.

5.2.3. Key Strategic Sectors

Key strategic sectors provide a direct test of whether the innovation effect of refocusing depends on the developmental content of the firm’s principal business. In these sectors, principal businesses are more closely tied to national security, public service, strategic emerging industries, or other policy priorities. Refocusing in such sectors is therefore not only a financial or organizational adjustment. It concentrates state capital in activities where innovation is closely linked to strategic tasks.

Panel A of

Table 12. Heterogeneity by Exposure to Key Strategic Sectors.

Panel A. Innovation Output
Variable	Innova		Ori_Innova		Gre_Innova		Inter_Innova
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Fka = 1	Fka = 0	Fka = 1	Fka = 0	Fka = 1	Fka = 0	Fka = 1	Fka = 0
CBF	10.326 *** (1.324)	1.673 * (0.901)	8.168 *** (1.110)	0.763 (0.781)	6.527 *** (0.780)	0.971 (0.859)	1.991 *** (0.669)	−0.766 (0.477)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	343	156	343	156	343	156	343	156
Adjusted R2	0.398	0.247	0.349	0.184	0.321	0.201	0.081	0.044
Panel B. innovation allocation
Variable	Ori_Share			Gre_Share			Inter_Share
	(1)	(2)		(3)		(4)	(5)	(6)
	Fka = 1	Fka = 0		Fka = 1		Fka = 0	Fka = 1	Fka = 0
CBF	0.324 ** (0.148)	0.043 (0.167)		0.287 ** (0.131)		0.281 * (0.147)	0.019 (0.054)	0.038 (0.061)
Controls	Yes	Yes		Yes		Yes	Yes	Yes
Year FE	Yes	Yes		Yes		Yes	Yes	Yes
Firm FE	Yes	Yes		Yes		Yes	Yes	Yes
Observations	343	156		343		156	343	156
Adjusted R2	0.158	0.093		0.171		0.141	0.052	0.049

Notes: Standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

shows a clear contrast. Among firms in key strategic sectors, CBF is positively and significantly associated with all four innovation-output measures. Among firms outside these sectors, only the coefficient for overall innovation is marginally significant at the 10% level; the invention-oriented, green, and internationally oriented coefficients are not statistically significant. This pattern suggests that narrowing business scope alone is not sufficient to generate broad domain-specific innovation gains. Such gains are most likely to appear when the strengthened core business carries substantive technological mandates.

Panel B further supports this interpretation. Among firms in key strategic sectors, refocusing significantly increases the shares of invention-oriented and green innovation. Outside these sectors, the invention-oriented and internationally oriented share coefficients are not significant, while the green-share coefficient is only marginally significant at the 10% level. Strategic-sector status therefore strengthens the translation of refocusing into a clearer directional shift toward mission-relevant technologies, rather than leaving it as a byproduct of administrative streamlining.

Taken together, the heterogeneity results reveal a coherent pattern. The innovation effect of state capital refocusing is strongest when the principal business carries substantive technological mandates, occupies a coordinating position in the industrial chain, and aligns with the firm’s functional mission. This pattern reinforces the mechanism identified above: refocusing is most likely to matter when the focused domain gives the firm both the capacity and the mandate to coordinate innovation.

5.3. Beyond Patents: Recognized Application-Oriented Outcomes

Patent output captures technological generation, but it does not by itself show whether technologies move toward application. For China’s innovation-driven development, the connection between R&D and application is a central policy concern. We therefore examine recognized application-oriented outcomes using the natural logarithm of one plus the number of provincial- or ministerial-level and above application-oriented science and technology awards. These awards explicitly evaluate application, diffusion, and economic or social benefits, making them a conservative proxy for formally recognized applied outcomes rather than a direct measure of market commercialization (Hall et al., 2005; Teece, 1986).

State capital refocusing may support recognized application-oriented outcomes by anchoring technological activity in concrete business contexts. When SOEs concentrate resources in principal businesses, R&D is more likely to be organized around defined production tasks, service needs, and infrastructure operations. Technologies developed in these settings are closer to actual use conditions and can benefit from feedback from suppliers, users, and operational processes (Teece, 1986; Adner, 2017; Jacobides et al., 2018).

Table 13. State Capital Refocusing and Recognized Application-Oriented Outcomes.

Variable	Application-Oriented Outcomes
	(1)	(2)	(3)	(4)	(5)
CBF	3.295 *** (0.879)	1.465 (0.962)	2.112 ** (0.973)	1.132 (0.903)	2.316 ** (0.911)
Innova		0.240 *** (0.044)
Ori_Innova			0.202 *** (0.064)
Gre_Innova				0.452 *** (0.069)
Inter_Innova					0.680 *** (0.121)
Controls	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes
Observations	499	499	499	499	499
Adjusted R2	0.321	0.378	0.341	0.412	0.326

Notes: Standard errors clustered at the firm level are reported in parentheses. ** and *** denote statistical significance at the 5% and 1% levels, respectively.

reports the results. Column (1) shows that CBF is positively associated with recognized application-oriented outcomes. Columns (2)–(5) introduce the innovation-output variables. When overall innovation or green innovation is included, the coefficient on CBF becomes statistically insignificant. This attenuation is consistent with overall and green patenting accounting for part of the association between refocusing and these outcomes, although the sequential regressions should not be interpreted as a formal mediation test. By contrast, when invention-oriented or internationally oriented innovation is included, the coefficient on CBF remains significant. This pattern indicates that these patent outcomes do not fully account for the association between refocusing and recognized application-oriented outcomes. Refocusing may therefore also operate through other channels, including more concentrated capital allocation, closer supply-demand coordination, and stronger alignment between R&D activity and operational constraints.

These findings extend the analysis beyond patent generation by showing that business-boundary adjustment is also associated with formally recognized application-oriented outcomes. They complement prior research on ownership change and mixed-ownership governance by identifying industrial-chain coordination as an additional route through which state capital can connect technological effort with production needs, application settings, and diffusion (W. He et al., 2020; X. Zhang et al., 2020; Pan et al., 2022; E. Liu, 2019; Teece, 1986). The evidence therefore supports a more specific interpretation: purposeful resource concentration is most consequential when it strengthens the organizational capacity to absorb technological feedback and move innovation closer to application.

6. Conclusions and Implications

6.1. Main Conclusions

State capital refocusing is better understood as a strategic re-anchoring of state capital within innovation networks rather than as a mere administrative retreat from excessive diversification. Within the SOE system, approved principal businesses are not simply revenue-generating segments; they are organizational vehicles through which state capital fulfills technological, industrial, and public-policy mandates. Refocusing can promote innovation by reallocating dispersed capital, managerial attention, and technical personnel toward these mandate-bearing domains. The central theoretical issue is therefore not whether organizational scope narrows per se, but whether greater concentration provides a clearer organizational and strategic anchor for cumulative, mission-oriented innovation.

The empirical evidence supports this interpretation. Among Beijing municipal SOEs, stronger core business focus is associated with higher levels of overall, invention-oriented, green, and internationally oriented patenting. More importantly, refocusing produces a selective shift in the composition of innovation: the shares of invention-oriented and green innovation increase, whereas the share of internationally oriented patents remains unchanged. This pattern suggests that state capital refocusing initially activates innovation domains most closely connected to domestic technological bottlenecks and green-transition mandates. The mechanism analysis provides evidence consistent with industrial-chain control as an important transmission channel. The results suggest that refocusing is more likely to be effective when internal resource concentration strengthens firms’ capacity to coordinate upstream suppliers, downstream users, and application scenarios.

The innovation effects of refocusing are also structurally contingent. Their direction and magnitude depend on the substantive content of the principal businesses being strengthened. An SOE’s functional mission—whether competitive, special-function, or urban public-service oriented—and its position within the industrial chain both shape its innovation response. The more pronounced effects observed in strategically important sectors indicate that narrowing organizational scope alone is insufficient to generate innovation. Innovation gains are more likely to emerge when state capital is concentrated in domains that carry explicit technological, strategic, and public-development responsibilities.

Finally, extending the analysis from patent production to recognized application-oriented outcomes clarifies the broader economic relevance of refocusing. Innovation policy depends not only on selecting priority sectors or providing subsidies and financial support, but also on firms capable of sustaining technological investment, accumulating capabilities, and coordinating complementary resources. SOEs are organizational vehicles through which public priorities can be translated into business commitments, resource-allocation decisions, and long-term technological development. The study therefore shifts attention from how much state capital is invested and where it is allocated to how it is governed within enterprises.

6.2. Policy Implications for SOE Managers and Policymakers

The implications of these findings extend beyond China’s specific institutional setting. For SOE managers, refocusing should be implemented as business-portfolio optimization combined with substantive resource redeployment, rather than as a simple reduction in business lines. For policymakers, strategic responsibilities should be matched with clear functional mandates, coherent business boundaries, differentiated performance evaluation, sufficient managerial discretion, and complementary innovation-support arrangements. For countries beyond China, the transferable lesson lies not in replicating China’s principal-business regulation, but in adapting this alignment among enterprise functions, business portfolios, managerial authority, performance evaluation, and innovation support to local institutional conditions (Musacchio et al., 2015; Mazzucato, 2018).

6.2.1. Aligning Capital Allocation with Core Strategic Missions

The central governance value of state capital refocusing lies in correcting the structural mismatch between dispersed capital allocation and concentrated strategic missions. Managers should treat refocusing as business-portfolio optimization linked directly to capital budgeting, R&D planning, talent allocation, and managerial attention. A narrowing of business scope should not be treated merely as asset disposal or financial downsizing. Resources released through the adjustment of noncore activities should support foundational research, the resolution of critical technological bottlenecks, the preservation of complementary capabilities, and the accumulation of long-term technological capacity (Mazzucato, 2018; Howell, 2017; Belenzon & Cioaca, 2025).

Whether capital allocation has genuinely been aligned with strategic missions cannot be determined from static concentration measures alone. It should be assessed through an integrated process linking business-boundary adjustment, resource reallocation, innovation upgrading, and recognized application-oriented outcomes. Refocusing becomes strategically meaningful only when resource concentration improves the composition of mission-relevant innovation, particularly invention-oriented and green innovation, and enables technological achievements to generate practical and industrial value. From this perspective, state capital refocusing is not simply a form of capital restructuring, but a mechanism for rebuilding strategic capabilities.

6.2.2. Differentiating Governance by Functional Mission

A uniform evaluation system is poorly suited to the heterogeneous mandates assigned to state capital. Policymakers should accompany strategic responsibilities with clear functional mandates, coherent principal-business boundaries, differentiated performance evaluation, sufficient managerial discretion, and complementary innovation-support measures. Competitive SOEs should be evaluated primarily in terms of market-oriented technological efficiency and commercialization performance. Special-function SOEs should be assessed on their contributions to strategic technological breakthroughs and industrial- and supply-chain security. Urban public-service SOEs should place greater emphasis on green technology adoption, service quality, and public-welfare outcomes (Bruton et al., 2015; Musacchio et al., 2015).

Differentiated governance does not imply weakening common standards of accountability. Rather, it means translating the shared objective of innovation-led development into performance criteria that reflect the distinct responsibilities of different types of SOEs. Such an approach can reduce the distortions created by one-size-fits-all evaluation systems and preserve the strategic and public functions that state-owned enterprises are expected to perform.

6.2.3. Turning Internal Resource Concentration into Coordination Across the Industrial Chain

The innovation effects of refocusing cannot be fully realized through internal resource concentration alone. They also depend on whether firms can translate stronger core-business capabilities into effective coordination across the wider industrial ecosystem. Managers should therefore treat refocusing as an active process of organizational and relational reconfiguration rather than as passive asset disposal. Its broader value emerges when R&D expenditure, technical personnel, managerial attention, and collaborative networks are reorganized around core businesses and connected more closely with suppliers, downstream users, and application contexts (Adner, 2017; Jacobides et al., 2018; E. Liu, 2019).

Policymakers should link the clarification of core-business boundaries with the development of chain-leading and system-integration capabilities. SOEs should be encouraged to establish technical standards, coordinate supplier networks, connect technological development with downstream demand, and open real-world application scenarios for new technologies. Through this process, firm-level resource concentration can be translated into coordination across the industrial chain, wider technological diffusion, and more systemic forms of innovation.

6.2.4. Adapting the Logic of Strategic Re-Anchoring to Emerging Economies

For other emerging economies, the transferable lesson lies not in the institutional form of China’s principal-business regulation, but in the underlying governance alignment. The design of refocusing policies should reflect domestic institutional capacity, industrial structure, and development priorities. State ownership is more likely to support innovation when enterprise functions, business portfolios, managerial authority, performance evaluation, and innovation-support measures are mutually aligned. This principle should be adapted to local institutions rather than used to justify direct replication of China’s regulatory arrangements (Aghion et al., 2015; Karo, 2018; E. Liu, 2019).

In resource-dependent economies, this logic may support downstream value-chain upgrading and the development of green-transition technologies. In economies pursuing industrial catch-up, it may justify concentrating state resources on upstream technological bottlenecks and supply-chain resilience. Where state-linked enterprises compete extensively in international markets, refocusing may need to be accompanied by stronger market discipline, commercialization incentives, and global intellectual-property capabilities (R. Baldwin & Freeman, 2022; E. Liu, 2019).

The broader implication is that the effectiveness of state capital should not be judged solely by the breadth of its market presence. It should also be assessed by its capacity to coordinate strategic resources, overcome systemic technological constraints, and undertake development tasks that private actors may be unwilling or unable to perform independently.

6.3. Research Boundaries and Future Directions

First, by establishing a systematic empirical relationship between state capital refocusing and enterprise innovation, this study also identifies several scope conditions. The empirical analysis draws on a relatively compact panel of 42 Beijing municipal SOEs and 499 firm-year observations. Beijing offers a comparatively coherent ownership and supervision environment, a clearly documented refocusing process, and a high concentration of universities, research institutes, skilled personnel, policy resources, and strategic industries. These features support detailed within-system analysis but may also affect firms’ capacity to convert refocusing into innovation. The sample size limits the precision of highly disaggregated subgroup comparisons, and the specific effect sizes and innovation patterns should not be generalized directly to SOEs operating under different administrative, regional, or national institutional conditions. The heterogeneity results are therefore interpreted as systematic differences across broad institutional categories rather than as precise estimates for narrowly defined subgroups.

Second, the measures employed capture important but bounded dimensions of refocusing and innovation. CBF provides an institutionally anchored measure of firms’ realized business concentration, but it does not fully capture the allocation of assets, personnel, R&D expenditure, managerial attention, decision authority, or organizational capabilities. Patent indicators capture observable technological activity and portfolio composition, while provincial- or ministerial-level and above application-oriented awards provide complementary evidence on formally recognized application, diffusion, and economic or social value. Neither set of indicators fully represents technological quality, knowledge diffusion, actual adoption, commercialization, or realized market returns. Future research could combine segment-level financial data, project-level R&D records, personnel mobility, patent citations and patent-family information, technical-standard participation, technology-adoption records, and organizational process evidence.

Third, the temporal interpretation of the findings rests on combined evidence from within-firm variation, lagged CBF, the reform-based continuous-treatment design, the event study, placebo reassignments, firm-specific trends, and extended controls. Year fixed effects absorb macroeconomic and policy shocks common to all firms, and the extended specifications account for observable differences in government support, industry conditions, R&D intensity, and executive turnover. Firms may nevertheless differ in their exposure to municipal and industry policies implemented during the reform period. The DID estimates are therefore interpreted as reform-related effects arising within a broader policy environment rather than as the isolated effect of a single policy instrument. Sharper policy discontinuities and more granular administrative data could distinguish related policy channels and temporal sequences more precisely.

A first research direction is to test this conditional governance relationship through cross-regional comparisons, comparisons between central and local SOEs, and cross-country institutional analysis, rather than directly extrapolating the numerical estimates from the Beijing sample. A second research direction is to move beyond the focal enterprise and examine the systemic effects of refocusing across industrial chains and innovation networks. The finding that industrial-chain control constitutes a key transmission mechanism suggests that refocusing may alter the relationships among suppliers, downstream users, private firms, research organizations, and innovation platforms. Future studies should therefore investigate the potentially dual effects of capital concentration. On the one hand, it may stimulate complementary innovation, accelerate knowledge diffusion, and strengthen supply-chain resilience. On the other hand, it may crowd out alternative technological trajectories, reinforce dependence on dominant state-owned actors, or generate new coordination costs. Shifting the unit of analysis from the firm to the broader industrial ecosystem would clarify when state capital concentration produces positive spillovers and when it constrains technological diversity and decentralized experimentation (Adner, 2017; Jacobides et al., 2018; E. Liu, 2019).

A third direction concerns the relationship between state capital refocusing and the restructuring of global value chains. As geopolitical tensions and technological competition reshape cross-border production and innovation networks, the reallocation of state capital may have consequences beyond national borders. Future research could examine how refocusing affects international supplier relationships, cross-border knowledge flows, technological standards, and the geographical distribution of strategic capabilities. A particularly important question is whether concentrating state resources in critical technologies reduces dependence on vulnerable nodes and strengthens strategic resilience, or whether it contributes to investment duplication, technological fragmentation, and new forms of structural dependence. Addressing these questions would connect research on state-owned enterprises and state capital with wider debates on technological sovereignty, strategic autonomy, and geoeconomic fragmentation (Alfaro et al., 2019; R. Baldwin & Freeman, 2022).

Finally, the digital and green transitions call for a more dynamic understanding of refocusing. These transformations are not simply creating new strategic industries; they are also redefining the technological boundaries of established core businesses. Future research could investigate how state-linked enterprises reconfigure their business portfolios and resource allocation around artificial intelligence, advanced manufacturing, renewable energy systems, and circular-economy technologies. Particular attention should be paid to how refocusing balances the concentration of resources in established strategic domains with experimentation in adjacent and emerging technologies. Cross-country comparisons could further reveal how differences in technological capabilities, industrial structures, and public missions shape distinct trajectories of strategic re-anchoring. This would help determine whether refocusing enhances long-term adaptability and mission-oriented innovation or creates new forms of path dependence (Acemoglu et al., 2012; Aghion et al., 2016; Mazzucato, 2018).

Together, these directions frame state capital refocusing as an ongoing governance problem involving the joint adjustment of business boundaries, capital allocation, organizational capabilities, industrial coordination, and public missions. Comparative and network-level research can further identify the institutional conditions under which this governance process supports mission-oriented innovation and structural transformation.