Does flattening the government hierarchy improve corporate innovation? Evidence from China

ABSTRACT We examine the impact of flattening the government hierarchy on corporate innovation. Leveraging the change from province–city–county to province–managing–county (PMC) administrative structures in China, we use a different-in-differences research design to show that the PMC reform enhances corporate innovation. The results are robust to a battery of checks. We report that political connection, state-ownership of firms and improved government services mediate the effects of the PMC reform on corporate innovation. Moreover, in the cross-sectional analysis, we find that the impact of the PMC reform is more salient for non-state-owned firms or when a province has new leadership.


INTRODUCTION
Innovation is essential for a firm and a country's economic development. However, innovation involves greater risk and uncertainty than other corporate activities (Egorova & Dubkov, 2016;Galasso & Schankerman, 2015;Holmstrom, 1989). Many firms hesitate to engage in innovation because of the risk of failure. Hence, from a policy perspective, it is imperative to explore the determinants of corporate innovation so as to promote such an endeavour. Among the various determinants, policymakers consider government effort, especially those from various local governments, to be one of them. For instance, in the context of smart specialization strategies in a region, Ruhrmann et al. (2022) articulate that innovation success depends on the decentralization of regional and local governments in Germany, whereas Kristensen et al. (2022) argue that the empowerment of change agents (local and regional governments) is critical to regional innovation. Consequently, most countries have some form of pro-innovation government policy at the regional or local level for the same reason. It is not surprising that local governments of the United States and many other countries have set up technology transfer offices and science parks to help firms engage in innovation. Similarly, the UK set up a non-departmental public body in 2018 -UK Research and Innovation (UKRI)to foster research and development (R&D) within UK localities to create positive impacts on human knowledge, economy and society. In addition, the Chinese government began the 'mass entrepreneurship' and 'grassroots entrepreneurship' campaigns in 2014 to ask local governments to provide corporatefriendly platforms to encourage corporate innovation.
Pro-innovation government policies are typically revealed through tax reliefs, subsidies and infrastructure provisions to stimulate corporate innovation. 1 The implementation of these policies requires the various regional and local governments to administer their specific steps and procedures. Thus, regional and local governments play an essential role in administering pro-innovation government policies. We contend that the effective implementation of these policies and distribution of government resources partly depends on the local government structure and how different local governments interact with individual firms.
While several studies have examined the role of government in helping (Charumilind et al., 2006;D'Ingiullo & Evangelista, 2020) or hurting a firm's innovation (Kong, 2020), few have examined the effect of the government's administrative hierarchy on corporate innovation. Typically, most local governments have a three-layer organizational structureprovince-city-county (or state-city-county). When the three-layer system is reduced to the two-layer design of province-county, the government hierarchy is flattened.
However, this flattening of the government hierarchy has mixed effects on economic outcomes. Some studies have found that after flattening, county governments enjoy increases in revenue (Li et al., 2016), while others have documented that county officials divert public funds for personal use (Bo et al., 2020). Hence, the research question is whether a flattening government hierarchy enhances corporate innovation.
This study uses a sample of Chinese firms to examine the effect a flattened government hierarchy has on corporate innovation. Leveraging several recent exogenous changes in China, from a three-layer (province-citycounty) to two-layer (province-county) administrative structure, we use a staggered difference-in-differences (DID) research design to analyse the effect that flattening the government hierarchy has on corporate innovation. In addition to the natural experiments on flattening the government hierarchy in China, we use a sample of Chinese firms to examine the research question due to administrative inefficiencies among local Chinese governments.
We argue that a flattened government hierarchy undoubtedly reduces the red tape in the structure, which lowers the transaction costs between firms and the government units that administer pro-innovation policies. According to Williamson (1981) and North (1992), transaction costs are the total costs of making a transaction. Hence, these costs significantly affect business operations and management. When there are many layers of government (as before the provincemanaging-county (PMC) reform), officials in different layers of the government have complex interactions among themselves, leading to the suboptimal execution of government policy (Rodriguez-Ward et al., 2018). This inter-governmental agency conflict adversely affects the desired pro-innovation policy goals. Further, officials in different layers of the government may interpret a specific pro-innovation policy differently. Thus, firms must navigate different sets of interpreted rules to receive pro-innovation resources. Reducing the layers of government minimizes the potential confusion in executing proinnovation policies caused by different layers of governments, resulting in a smoother allocation of resources for corporate innovation.
However, innovation engagement is expensive. Often, innovation projects fail because of insufficient funding. Firms unavoidably need external financing to pursue innovation. Zhang and Zheng (2020) use a sample of Chinese firms to document that financial constraints impede innovation. Similarly, Deng et al. (2021) report that corporate innovation increases when banks diversify their operations geographically. They theorize that geographically diversified banks often provide greater financial operational flexibility to borrowing firms. Thus, an increase in bank loans eases firms' financial constraints, helping them engage in more corporate innovation. Firms have better access to funds, leading to better innovation. Hence, firms have more external resources (from the government) available to enhance innovation (Bertello et al., 2022). By contrast, a flattened government means lower monitoring by government officials, which may translate into severe rent extraction from government officials, thus decreasing corporate innovation. As China shares similar characteristics with other emerging markets and represents a suitable business environment for examining our research question, our findings offer valuable insights for other emerging markets.
After the flattening of government hierarchies, the structure becomes that of a province directly managing a county or PMC. We identify firms in PMC locations (treatment or PMC firms) and firms in non-PMC locations (control or non-PMC firms) for the period 2003-17. We then document that PMC firms exhibit better innovation than non-PMC firms in terms of the total and inventive patent applications and patents held. The results are robust to a subsample that excludes major cities, alternative innovation metrics, placebo tests, an alternative estimation method and a propensity score matching (PSM) sample.
In the cross-sectional analysis, we find that the PMC reform's positive impact on corporate innovation is more salient for non-state-owned firms (enterprises) (non-SOEs) or when a province has new leadership. Moreover, consistent with intuition, we report that political connections, state ownership of firms and improved government services mediate the effects of the PMC reform on corporate innovation. These findings corroborate the logic of reduced red tape because of a flattened government leading to improved governance.
This study makes three contributions to the literature and policies. First, we advance corporate innovation literature by documenting a new determinant of corporate innovation in the flattening of hierarchies in local governments. A simplified local government structure contributes to better corporate innovation. Thus, our findings complement the literature stating that social capital (Kobeissi et al., 2022), decentralization of regional and local governments (Ruhrmann et al., 2022), and empowerment of local change agents (Kristensen et al., 2022), among others, contribute to corporate innovation.
Second, we contribute to the broad literature on public administration and government structure by affirming the positive effect of a flattened government hierarchy on corporate activities. We demonstrate that a more efficient government is helpful beyond government operations. Hence, these findings motivate the government to improve its services and simplify its structure.
Third, in terms of government policy, we show that government plays a role in corporate innovation by simplifying its local hierarchy. Thus, a government can go beyond tax cuts, subsidies, or other pro-innovation policies, to promote corporate innovation.

Background
China began its economic reforms in the early 1980s. In addition to these reforms, the Chinese government has privatized its SOEs (Li et al., 2004). In public administration, it has adopted a province-city-county model, where each province is charged with various economic targets such as the provincial-level gross domestic product growth rate, job creation, and tax revenue. The threelayer model unavoidably creates agency conflicts owing to layering and unclear accountability within various governments. Different government officials pursue their own personal benefits (Rodriguez-Ward et al., 2018). Since 2002, the Chinese central government has started experimenting with transforming the province-city-county model to a province-county model (Bo et al., 2020;Li et al., 2016). This marks the beginning of the PMC reforms. Some provinces, such as Zhejiang Province, initiated PMC reforms in select counties in late 2002. Other provinces followed. The differences in the timing of the PMC reform in different provinces suggests a staggered event setting.
2.2. Literature review 2.2.1. Government structure reform There are three major conceptual underpinnings related to government structural reform in a public administration setting.
First, it is associated with the effective use of public funds. Studies have typically examined conceptual frameworks around differences in the efficiency of providing public goods between lower and higher level governments. When the differences between economic, cultural, geographical and natural resources are significant across regions, decentralizing the power to provide public goods is more efficient (Martinez-Vazquez & McNab, 2003).
Second, it relates to the costs and benefits of different government structures. While the literature recognizes the benefits of decentralizing a government, coordination and administrative expenses increase after decentralization. Thus, it becomes an optimization problem to balance the costs and benefits of decentralization (Breton & Scott, 1978).
Third, the conceptual framework emphasizes that different regions or levels of government compete for benefits. Tiebout (1956) has proposed that constituents 'vote with their feet' to find their preferred administrative areas to live or settle their businesses in. Breton (1996) suggests that local governments compete for labour and other resources. Hence, the government engages in horizontal (other same-level) and vertical (higher and lower level) competition to garner more labour and other resources. Consequently, governments provide the best public goods to draw resources from their constituents.
In the practice of government structure, it is common to have a three-layer province-city-county setting. Previous studies on the PMC reform suggest that there are advantages and disadvantages to flattening the government structure. In terms of benefits, the literature supports that the PMC reform enhances economic growth (Liu & Alm, 2016;Ma & Mao, 2018), increases county governments' revenues (Li et al., 2016), and expands provincial and county government education inputs (Huang et al., 2017), among others. Underneath these benefits of flattening the government is the reduction in transaction costs and easing of the financial constraints of firms participating in the government's pro-innovation policies. By contrast, the PMC reform also means fewer checks and balances for government officials. Such reform hinders governmental financial reform (Wang et al., 2012) and provides county officials with opportunities to divert public funds for personal use (Bo et al., 2020). 2

The role of government in corporate innovation
This strand of literature contains two clusters. The first cluster suggests that government inhibits corporate innovation. For instance, Kong (2020) documents that firms significantly lower their innovation when the government increases general spending. Kong suggests that resource diversion is induced by increased government spending. Kong and Qin (2021) find that the anti-corruption campaign in China has helped corporate innovation, implying that government corruption is detrimental to corporate innovation.
The second cluster stipulates that the government plays the role of a 'helping hand'. For instance, a government can help firms obtain bank loans (Charumilind et al., 2006), provide subsidies and tax relief (Adhikari et al., 2006;Deng et al., 2019Deng et al., , 2020Zhao et al., 2019), ease industry regulation (Fan et al., 2009), protect intellectual property rights (Brown et al., 2013), and obtain crucial R&D resources (Zhou et al., 2017), among others. Other studies suggest that state ownership, although not equally beneficial, enhances a firm's R&D impact on innovation (Yi et al., 2017). D'Ingiullo and Evangelista (2020) suggest that provincial government effectiveness, regulatory quality, voice and accountability contribute to corporate innovation. Overall, the government can command resources to help firms engage in corporate innovation. required to run an effective county government in the province are communicated well. For instance, after the PMC reform, firms apply for tax reliefs, subsidies, and infrastructure provisions to two levels of government instead of three. The time and effort (transaction costs) saved in interacting with one less government unit become significant. Thus, a simplified government structure leads to reduced transaction costs and improved firm performance, including innovation.
Second, the PMC reforms reduce agency conflict between different levels of government, leading to the better execution of governments' pro-innovation policies. When there are many layers of government, officials in each layer pay attention to their own areas. The silo effect suggests that officials may use different criteria to execute pro-innovation policies. Rodriguez-Ward et al. (2018) present an interesting case study of environmental protection in Peru. Through their interviews, they report that different government agencies and major players have differing attitudes towards forest sustainability aimed at reducing emissions from deforestation and forest degradation (a global initiative). With no agreement among officials, no substantial and effective regional regulations have been put in place, leading to more illegal gold mining activities. Their study shows that many layers of government create agency conflicts, resulting in them not meeting the desired government goal. In our context, suppose the central government has several pro-innovation policies. More layers of government indicate a higher possibility of agency conflicts resulting in different government rules being imposed. Firms then find it more challenging to materialize resources from pro-innovation policies. Hence, the PMC reform provides advantages for corporate innovation and channels government resources to ease the financial constraints for firms to engage in innovation. 3 However, the PMC reform also has disadvantages. There are fewer oversights on firms that receive resources from flattened governments. While they are not certain, some studies suggest that less oversight translates into more opportunities for firms to engage in rent-seeking activities (Liu et al., 2018) and other corruption opportunities (Wang et al., 2020). We argue that after the PMC reform, firms face fewer officials allowing them to conduct rent-seeking or making it easier to corrupt officials, which counters the positive effect of the PMC reform on corporate innovation.
The reform also suggests lesser monitoring by the government officials. Accordingly, county officials may waste resources (Bo et al., 2020). Additionally, firms may rely overly on local county governments (Sydow & Koch, 2009). As a result, executives may focus on developing political connections with local officials to secure their benefits. Consequently, firms will spend resources to secure county officials' help without using them for corporate innovation. Thus, corporate innovation may reduce.
Having said that, with the ongoing anti-corruption campaign in China (Tong, 2022), we expect the collective disadvantages of rent-seeking or corruption on innovation to be minor, if any. Additionally, the central inspection team of the Chinese government has aggressively conducted inspections of provincial, city and county governments as part of its anti-corruption campaign (Xu et al., 2022). These inspections have shaped government officials in a positive way. Thus, the disadvantages of PMC reform, if any, should be moderate. Therefore, we expect a net positive effect on corporate innovation. As a result, we present the following testable hypothesis: Hypothesis 1: PMC firms engage in more corporate innovations than non-PMC firms.
Our hypothesis suggests that the PMC reform streamlines administration to facilitate PMC firms' engagement in innovation. However, the mechanism underlying this positive effect of the PMC reform on corporate innovation remains unclear. We propose two potential transmission mechanisms.
First, the PMC reform reduces government bureaucracy, implying that the number of roadblocks due to excessive government procedures is reduced for firms. If these arguments are valid, we expect a lower likelihood of a firm developing political connections. In emerging markets such as China, firms spend resources to develop political connections and leverage them to get access to government services or receive favourable treatment, such as a lower tax rate (Adhikari et al., 2006), a competitive advantage to access target firms in mergers , among others. We argue that after the PMC reform, there is a lesser need to develop political connections. Thus, PMC firms can save resources to better finance their innovation activities. Further, with a lesser need for political connections, resource allocation at the local government level becomes more efficient. Accordingly, political connections mediate the impact of the PMC reform on innovation metrics.
Second, a flattened government can provide better services due to the reduced red tape. As a result, firms enhance their operating efficiency thanks to the better government services that aid their innovation endeavours. Therefore, firms have better corporate innovation and operating efficiency. For instance, the public-private partnership (PPP) literature suggests that PPP improves the operating efficiency of a local government in providing public goods (Anwar et al., 2018). Similarly, in the PPP setting of high-speed rail operations, government regulations inhibit the efficiency of private operators of the rail system (Bugalia et al., 2021). While the PPP literature does not directly suggest that a flattened government enhances government services, we deduce that an excessively burgeoning government structure neither improves a government's operational efficiency nor the quality of government services. In the context of the PMC reform, it certainly restrains government bureaucracy. Thus, we contend that it improves government operating efficiency, leading to lower operating costs for a firm and/or improved ability of the county to enhance its services. We state the second testable hypothesis as follows: Hypothesis 2: Political connections and improved government services mediate the PMC's impact on corporate innovation.

Data
We examine all provincial and county governments in China to identify if a county has a PMC structure as of 2015. If a county has had a PMC reform, we identify the year in which it switched from non-PMC to PMC. We then identify the headquarter locations of public firms for the period 2003-17 to classify firms into a treatment group (those located in a PMC) and control group (those located in a non-PMC). Firm-level innovation and financial data are extracted from the China Stock Market and Accounting Research Database (CSMAR). We delete firms that (1) are listed in the growth enterprise market; (2) belong to the financial industry; (3) are financially distressed; or (4) have missing accounting and finance information. We exclude firms in the growth enterprise market because these firms were allowed to be publicly listed in 2009, while our sample period is from 2003 to 2017. To avoid mismatches between the period and firms listed on the main board, we do not use them.
We exclude financial firms because they are subject to heavy regulations as a result of their sensitivity to the stability of the economy. Additionally, the output of financial firms differs from that of other industrial firms. The objective of financially distressed firms is survival. Thus, other activities such as innovation are not their focus. For firms with missing accounting and financial information, we do not have sufficient information to control for specific accounting and financial characteristics that may impact their innovation.
The final sample comprises 7635 firm-years with 1769 firm-years from the PMC group. We winsorize all the continuous variables at the 1% and 99% levels.

Multiple regression model
We use the following multiple regression model to examine the impact of the PMC reform on a firm's innovation: where Y i,k,t+1 is the innovation metric from firm i in location k at time t + 1. DIRECT i,k,t is a binary (1, 0) indicator variable at time t with a value of 1 if a firm is located in a PMC location, and 0 otherwise. Equation (1) is a staggered difference-in-differences (DID) research setting to compare the effect of PMC on a firm's innovation at different times in different counties, making the PMC switch. DIRECT i,k,t is the interaction term in the DID Natural logarithm of 1 plus the number of inventive patent applications in a year UTILITY_APP Natural logarithm of 1 plus the number of utility patent applications in a year PATENT_HELD Natural logarithm of 1 plus the number of all patents held from applications in a year INVENT_HELD Natural logarithm of 1 plus the number of inventive patents held from inventive patent applications in a year UTILITY_HELD Natural logarithm of 1 plus the number of utility patents held from utility patent applications in a year

DIRECT
Dummy variable with a value of 1 if a firm is located in a county with any modes of province-managingcounty in a year, and 0 otherwise

SIZE
Natural logarithm of total assets DEBT Ratio of total liabilities to total assets LANGE Natural logarithm of a firm's listing year Q Ratio of market capitalization to a firm's asset replacement value STOCKPORT Annual stock return LNEMPLOYEE Natural logarithm of total number of employees

TOP
Ownership percentage of the largest shareholder DUALITY If a firm's chief executive officer and chairperson is the same individual, the value is 1, and 0 otherwise SEPARATION Degree of separation of control and ownership right GDP GROWTH Gross domestic product growth rate in the location of a firm PROPERTY If a firm is state-owned, the value is 1, and 0 otherwise Does flattening the government hierarchy improve corporate innovation? Evidence from China model. It captures the PMC and time effect. This is the same as TREAT*POST (if TREAT is defined as a dummy variable with a value of 1 if a firm is located in a PMC county, and POST is a dummy variable with a value of 1 if a firm is in the year after the county becomes a PMC county). We follow Hoynes et al. (2011) and use only DIRECT to capture the DID effect of the model. For innovation, we use six different metrics to capture different dimensions of innovationtotal patents applied and held (PATENT_APP and PATENT_HELD), inventive patents applied and held (INVENT_APP and INVENT_HELD), and utility patents applied and held (UTILITY_APP and UTILITY_HELD). Inventive patents are high-quality innovations that require substantial resources and time to develop, whereas utility patents are easier and quicker to develop than inventive patents. Inventive patents provide more long-term benefits to firms than utility patents. Hence, we gauge corporate innovation performance and strategy by studying the inventive and utility patents applied for and held.
Following Hsu et al. (2014), we include a set of control variables in equation (1) Table 1 presents the definitions of all the variables. We control for year and firmfixed effects in equation (1) and cluster the standard error of the estimate at the firm level. Table 2 displays the sample's frequency distribution by year, industry, and province/autonomous city. In panel A, the early years have fewer samples than later years, reflecting the PMC reform trend. Panel B shows that the manufacturing industry has the largest number of PMC firms. Panel C shows that the eastern and southern provinces/autonomous cities have more samples than other parts of China. Table 3 presents the summary statistics of the sample. For patent applications, the means of PATENT_APP, INVENT_APP and UTILITY_APP (before the natural logarithm) are 68.448, 28.069 and 39.864, respectively. For the patents held, the means of PATENT_HELD, INVENT_HELD and UTILITY_HELD (before the natural logarithm) are 52.031, 10.549 and 39.737, respectively. The sum of inventive and utility patent applications/held does not precisely equal the total patent applications/ held because of the winsorization of the variables.

Summary statistics
Two interesting results are observed. First, the standard deviation of each innovation metric is two to three times larger than the corresponding mean. The range (maximum minus minimum patent metric) is generally large. Thus, corporate innovation varies significantly  (1) in panel A of Table 4. Across the six columns, the coefficients of DIRECT are positive and significant at the 1%, 5% or 10% levels of the total and inventive patents columns in columns (1), (2), (4) and (5). By contrast, the same coefficients are insignificant for the UTILITY_APP and UTILITY_HELD regression equations in columns (3) and (6). The findings suggest that the PMC reform is effective in driving the total patent applications and patents held, especially for inventive patents. Moreover, from the results in columns (2) and (5), the PMC reform encourages firms to spend resources on long-term inventive innovation, leading to more inventive patents. Thus, a flattened government provides a sound environment that encourages firms to engage in long-term innovation strategies. The insignificance of the coefficients of DIRECT in columns (3) and (6) suggests that the PMC reform did not help utility patent applications and patents held. We conjecture that as utility patents are easy innovations, the PMC reform does not contribute to the firm's decisions to engage in utility-type innovation.        Note: We use county land area, population, economic power, budget, county scale, county fixed assets, primary industry, and secondary industry to conduct a one-to-one match of a treatment firm (PMC firm) with a control firm (non-PMC firm) by a propensity score difference < 0.01. Table 1 shows the definitions of variables. Standard errors of estimates are clustered at the firm level. T-statistics are shown in the parentheses. ***, ** and *Significance at the 1%, 5% and 10% levels, respectively.

Does flattening the government hierarchy improve corporate innovation? Evidence from China
Where significant, the control variables carry the expected signs. For instance, the coefficients of LNSIZE, LNAGE and LNEMPLOYEE are positively significant, consistent with the intuition that a firm with more resources (larger or more employees) or experience (listed longer) has more innovations. 4

Propensity score matching (PSM)
A firm may decide to be located in a PMC reform county. To mitigate this potential selectivity bias, we follow Heckman et al. (1998) and conduct PSM to match the treatment firm with a control firm. 5 Following the literature (Huang et al., 2017;Ma & Mao, 2018), we use county land area (INLAND, the natural logarithm of land in a county), population (LNPOPULATION, the natural logarithm of a county's population), economic power (LNMGDP, the natural logarithm of per capita GDP), budget (GENBUDRE, county revenue to GDP ratio), county scale (LNTOWN-SHIP, the natural logarithm of the number of villages), county fixed assets (LNFIXED, the natural logarithm of a county's fixed assets), primary industry (PRIMARY, primary industry to GDP ratio), and secondary industry (SECOND-ARY, secondary industry to GDP ratio). The set of PSM variables is based on the logic that the Chinese government loosely follows certain criteria in the PMC reform. According to Huang et al. (2017), the PMC reform prioritizes counties with heavy financial burdens, poverty-stricken counties, and counties with large agricultural production as pilot projects. At the same time, each province may choose pilot counties according to their development characteristics, fully considering factors such as economic and social development conditions and development prospects. Appendix A in the supplemental data online presents an explanation of the PSM variables.
We use a one-to-one match with a propensity score difference of <0.01. The PSM sample is reduced to 1682. For the PSM to be valid, we examine the parallel trend assumption behind equation (1). We use an event study approach in equation (2): where DIRECT(N) is a dummy variable with a value of 1 if the time is N years before or after when the county in which a firm is located has undergone the PMC reform. We present the findings in Appendix C in the supplemental data online. As expected, all the coefficients of DIRECT(-5), DIRECT(-4), DIRECT(-3) and DIRECT(-2) are insignificant, thus supporting the parallel trend assumption. However, the coefficients of DIRECT(-1) are positively significant at the 5% and 10% levels. We attribute these results to firms responding to the anticipated inclusion of a county in the PMC reform in the final stage of the decision process. The results in panel B of Table 4 for the PSM show that the results are qualitatively similar to those in panel A.

Robustness checks 4.3.1. Alternative samples
The results from the full sample in Table 4 do not consider that some counties are located in well-developed regions, some industries are more innovative than others, or that some newer firms are more innovative. For robustness, we delete the firms located in the districts of well-developed cities (equivalent to counties) such as Beijing, Shanghai, Guangzhou and Shenzhen, firms that are not from the manufacturing industry, and firms with initial public offerings after 2010. The reduced sample has 2218 firm-year observations. Panel A of Table 5 presents these findings. For robustness, we do not include the coefficients of the control variables. The coefficients of DIRECT remain positively significant at the 5% or 10% level for total and inventive patents applied and held. These values are qualitatively similar to those listed in Table 4.

Alternative metrics of innovation
We use the total number of patent applications or patents held to gauge a firm's innovation level. Alternatively, we use the patent success rate (the ratio of patents held to patents applied). Specifically, we define INNO1 (INNO2) as the ratio of patents ultimately approved by the patent office in a year to the number of patents (inventive patents) applied for in the same year. Additionally, patent applications require time to obtain approvals. Hence, we gauge a firm's innovation efficiency using the ratio of patents approved within 12 months to the total number of patents applied for (INNO3). Finally, when approving a patent, the patent office provides the length of time to approve the specific patent. Therefore, we use the average time required to approve all the patent applications of a firm to assess its innovation efficiency  Note: Panel A excludes firms that are located in districts of autonomous cities (equivalent to counties) of Beijing, Shanghai, Guangzhou and Shenzhen, firms that are not from the manufacturing industry, and initial public offering firms after 2010. Panel B uses alternative metrics for corporate innovation. INNO1 (INNO2) is the ratio of patents ultimately approved by the patent office in a year to the number of patents (inventive patents) applied in the same year. INNO3 is the ratio of patents approved within 12 months to the total number of patents applied. INNO4 is the average length of time required to approve all the patent applications in a firm. Panel C presents the placebo test results. We move up three years for every PMC occurrence time to define the DIRECT variable (denoted as DIRECT_PLACEBO). Panel D presents the results using Tobit. For brevity, we do not present the coefficients of control variables. Table 1 shows the definitions of variables. The standard errors of estimates are clustered at the firm level. T-statistics are shown in the parentheses. ***, ** and *Significance at the 1%, 5% and 10% levels, respectively.
variable (denoted as DIRECT_PLACEBO). Despite this, the placebo test should not yield significant results. The findings in panel C of Table 5 confirm that the coefficients of DIRECT_PLACEBO are insignificant. Hence, our baseline findings in Table 3 are not due to the natural improvement of the government or societal progress.

Alternative estimation method
It is a fact that some firms do not engage in any innovation. Therefore, their patent application number zero. For robustness, we estimate equation (1) again using Tobit to account for zero patent applications and patents held. The coefficients of DIRECT remain positive and significant at the 5% and 10% levels in columns (1), (2), (4) and (5) in panel D of Table 5. These findings are qualitatively similar to those presented in Table 3.

Cross-sectional analysis: state ownership and provincial new leadership
We contend that a firm's motivation and the provincial government matter in corporate innovation. Specifically, we consider if the firm is state-owned and if the province has new provincial government leadership.   Note: Panel A presents the results using state ownership. We define PROPERTY as a (1, 0) indicator variable with a value of 1 if a firm is state-owned, and 0 otherwise. Panel B presents the findings using the new leadership of the provincial government. PARTY is a (1, 0) variable with a value of 1 if the provincial government is new at t -1 and t (t is the PMC year), and 0 otherwise. For brevity, we do not present the coefficients of control variables. Table 1 shows the definitions of variables. The standard errors of estimates are clustered at the firm level. T-statistics are shown in the parentheses. ***, ** and *Significance at the 1%, 5% and 10% levels, respectively.
Does flattening the government hierarchy improve corporate innovation? Evidence from China 1571 SOEs have government ownership and support. Hence, they face fewer financial constraints and receive ample funding from banks (Huang et al., 2016) and other sources. These loosened financial constraints facilitate innovation (Zhan & Zhu, 2021) and other activities. In general, Cull et al. (2015) suggest that SOEs have Table 7. Results for the mediating effect of a firm's political connection (PC) and state ownership (STATESHARE) for the impact of province-managing-county on corporate innovation by ordinary least squares. fewer financial constraints because of their political connections with government officials. We argue that before the PMC reform, SOEs had three channels (county, city and province) to obtain resources to conduct innovation.
After the reform, SOEs now have only two channels (county and province). All else being the same, we expect that post-reforms, SOEs should have relatively fewer resources to drive innovation. By contrast, we expect that Note: We use a firm's operation cost (LNCOST) and a county's per capita road area (PCRA) to proxy county government service. For brevity, we do not present the coefficients of control variables. Table 1 shows the definitions of variables. The standard errors of estimates are clustered at the firm level.
prior to the reforms, non-SOEs diversified their funding sources for innovation because it was challenging to obtain funding. After the reform, non-SOEs should find it easier to obtain funding from the simplified government structure because they now need to deal with fewer layers of government. Hence, we hypothesize that, compared with non-SOEs, the PMC reform has a weaker positive impact on innovation in SOEs. Provincial leaders are responsible for managing their provinces. Their political careers are on the line. When the PMC reform is under the spotlight, a new provincial leader is eager to show his or her performance in a short period in any possible way. Thus, the leader is genuinely motivated to carry out reforms to facilitate their success. In contrast, a long-tenured leader has many ways to show off his or her performance. The PMC reform may be just one of them. Consequently, the motivation to push for PMC reform is weaker among long-tenure leaders when compared with new leaders.
To test this conjecture, we define PROPERTY as a (1, 0) indicator variable with a value of 1 if a firm is stateowned and 0 otherwise. Similarly, PARTY is a (1, 0) variable with a value of 1 if the provincial government is new at t -1 and t (t is the PMC year), and 0 otherwise. We then augment equation (1) with PROPERTY and DIRECT*-PROPERTY (PARTY and DIRECT*PARTY, respectively).
The terms DIRECT*PROPERTY and DIRECT*PARTY capture the moderating effect of state ownership and new political leaders. Table 6 presents the results.
In panel A of Table 6, the coefficients of DIRECT*-PROPERTY are negatively significant at the 5% and 10% levels for the total and inventive patent columns. The results suggest that conditional on the PMC reform, SOEs have less innovation success. In panel B, the coefficients of DIRECT*PARTY are positively significant at the 5% and 10% levels in the total and utility patent columns. Interestingly, new political leadership does not affect the inventive patents held. We interpret this to mean that new political leaders are more interested in patents that can help advance their political careers. Hence, they do not have the patience to help the advancement of inventive patents.
4.5. Transmission mechanism 4.5.1. A firm's political connection and government involvement We define PC as a (1, 0) indicator variable with a value of 1 if a firm's chairperson or CEO has had previous experience as a government official or been a representative at any level of the People's Congress or People's Political Consultative Conference, and 0 otherwise. In addition, we take STATESHARE as the percentage of a firm's state ownership. Panels A-C in Table 7 present the findings.
Panel A of Table 7 presents the effect of the PMC reform on a firm's PC and STATESHARE. As expected, the coefficients of DIRECT are negatively significant at the 1% level. After the PMC reform, firms have a reduced desire to utilize political venues, leading to a lower likelihood of using political connections and state ownership. In panels B and C, we augment equation (1), using PC or STATESHARE. If PC and STATESHARE are mediating variables, their coefficients should be significant in explaining the variation of innovation metrics. In panel B, the coefficients of PC are significantly negative for the inventive patent applications and patents held equations only. Interestingly, in panel C, the coefficients of STATE-SHARE are negatively significant for all six innovation metrics. Thus, a reduction in political connections and state ownership (lower PC and STATESHARE values) mediates the impact of the PMC reform on corporate innovation.

Improvement of government services
We gauge government services using two metrics: the natural logarithm of a firm's operational cost (LNCOST) and a county's per capita road area (PCRA). The logic is that the PMC reform lowers the operating cost of a firm and/or improves a county's ability to enhance its road service. Panels A-C in Table 8 present the findings. Similar to the results in Table 8, the PMC reform significantly lowers the firm's operating costs and enhances the county's road services (in panel A). In panels B and C, the coefficients of LNCOST (PCRA) are negatively (positively) significant in explaining variations in the innovation metrics. Hence, the findings suggest that operating cost reduction and county government improved services are mediating factors for the effect of the PMC reform on corporate innovation.

CONCLUSIONS
We have examined the impact of a flattened government hierarchy on corporate innovation. Drawing from the public administration framework of reducing red tape, enhancing communication between provincial and county governments, and increasing the motivation of county governments, we have hypothesized that after the flattening of the government hierarchy, firms in reform counties have better corporate innovation than those not in reform counties. Leveraging the staggering events of changing province-city-county administration to PMC hierarchy in some provinces in China, we use a DID research design to examine our testable hypothesis. The findings support our testable hypothesis.
Additional analyses suggest that government service improvement mediates the impact of the PMC reform on corporate innovation. The results of the moderating analysis show that the impact of the PMC reform is more salient for non-SOEs or when a province has new leadership. In both cases, firms or government officials are more motivated to engage in innovation activities, further corroborating the testable hypothesis.
Overall, our study has two policy implications. First, it is generally a good practice to simplify the government hierarchy. This practice makes local governments more efficient and contributes to better corporate activities, ultimately enhancing local economic development. Our findings echo several recent studies on how the decentralization of regional and local governments (Ruhrmann et al., 2022), or the empowerment of regional and local change agents (Kristensen et al., 2022), can enhance regional and local innovation. Second, in corporate innovation, it is essential to motivate firms by simplifying government bureaucracy to facilitate innovation. While government taxes, subsidies, or other pro-innovation policies are helpful, it is also beneficial for a flattened government to make these policies effective.
Our study has three limitations. First, the results were obtained from a single country. Although we do not have a priori reason to suggest that the findings do not apply to other emerging markets, it would be meaningful to examine other emerging markets to validate our findings further. Second, corporate innovation is a long-term endeavour. However, our research design focuses primarily on corporate innovation at t + 1. Thus, the impact of PMC reform on corporate innovation depicted in our study is likely to underestimate its real effect. In the future, it will be more appropriate to design a research setting that incorporates the multi-year nature of corporate innovation. Third, corporate innovation success requires an appropriate corporate culture. While we account for corporate culture by controlling for the firm fixed effect, this may not be enough. Future extensions may be made to quantify each firm's corporate culture and examine how these cultural factors interact with the flattening of government hierarchies in impacting corporate innovation.

NOTES
1. For instance, a government can build a science park (a form of infrastructure) to subsidize the occupancies of innovative firms in the park (a form of subsidy). At the end of the year, the government allows these firms to deduct R&D spending (a form of tax relief) to lower their taxable income. 2. A unique feature of the government structure in China is its Hukou system. Essentially, the system imposes hard restrictions on people's movement. For instance, an individual cannot work, live or study in elementary and high schools in a location (e.g., a city) without the Hukou of the location. However, the Hukou system has been a soft system since the economic reform in the early 1980s. Due to a general labour shortage in the developed regions (e.g., large cities), individuals can work or live in a location without the Hukou. These individuals without Hukou do not get the full convenience as those with the Hukou, such as the right to buy a house. Nonetheless, individuals without the Hukou can fully move to work and live in other locations. 3. Our argument does not negate the possibility that reducing red tape improves the productivity and efficiency, and that external resources improve economic performance (such as more revenues and higher market shares). 4. We also examine a subsample of manufacturing firms in the sample by including R&D investment as an additional control variable for the robustness check. We justify this robustness check for two reasons. First, firms in the manufacturing industry make up approximately 71% of the sample. Second, manufacturing firms typically have the discretion to engage in R&D. If they do not report R&D investments, it is very unlikely that they do not have R&D activities. Specifically, we use three metrics for robustness: (1) per capita R&D investment (RDP), which is the R&D investment per employee; (2) per unit sale R&D investment (RDS), which is the ratio of R&D investment to sales; and (3) per unit asset R&D investment (RDA), which is the ratio of R&D investment to total assets. We present the findings in Appendix B in the supplemental data online. Except for column (2) of panels A and B, the coefficients of DIRECT in panels A-C of Appendix B are similar to those of Table 4. Hence, after including R&D investment as a control variable, the baseline results remain intact. 5. An alternative approach to conducting the PSM is by matching counties. We consider the firm-level PSM superior to county-level PSM for two reasons. First, we examine corporate innovation, not county-level aggregate innovation. Hence, it is better to match by firms because matching a treatment firm (located in a PMC county) with a control firm (located in a non-PMC county) always implies we are comparing a PMC county with a non-PMC county. Second, if we match counties, there is no guarantee that other firm characteristics are matched. There may be firm characteristics that affect a firm's choice to locate in a PMC county and also affects innovation. Thus, if we apply a PSM county match, we cannot fully mitigate the endogeneity due to missing firm level variables. 6. We also use patent citations as an alternative metric for robustness. In addition, we use the patent approval rate (patents approved/patents applied) and inventive patent approval rate (patents approved/inventive patents applied) as alternative metrics (Rickard et al., 2016). The unreported results are consistent with the baseline results in Table 4.