The Heterogeneous Effects of Central and Local Subsidies on Firms’ Innovation

Abstract

Four decades of rapid economic growth have enabled the Chinese government to dedicate more resources to research and development. China is the world’s second largest spender on food and agricultural research in terms of absolute expenditures and the largest investor on a purchasing power parity basis. Using a unique panel dataset collected in 2019 in China’s seed company and fixed effect models, this study analyzes the heterogeneous effects of central subsidies and local subsidies. Specifically, this study first tests whether government subsidies have a positive impact on firms’ innovation. Then, this study answers whether the impact of local subsidies differs from that of central subsidies. The estimation results show that the central subsidies positively contribute to firm’s innovation, while the impact of local subsidies on firms’ innovation has not been confirmed. Further analysis shows that local subsidies positively affect firms’ economic performance. That is, rather than focusing on research capacity, local governments are more concerned about firms’ current economic performance due to the performance-based promotion scheme in China. Based on this study, local governments should implement similar methods to those of the central government in research project funding and criteria for selecting research projects to promote firms’ innovation.

1. Introduction

As the world’s second largest economy, China spends hundreds of billions of dollars in research and development (R&D). The total expenditure of China on R&D was USD 526 billion (or 22% of the global total), which makes China the second largest spender in 2019 [1]. In addition, the annual increase in China’s R&D is 10.6%, which is approximately twice that of the United States from 2010 to 2019 [1]. Chai et al. (2019) showed that China outspends the United States on food and agricultural research on a purchasing power parity basis [2].

Although it is believed that governments should subsidize firms’ R&D investment due to its positive externality [3,4], empirical studies on the impact of subsidies are inconsistent. On the one hand, some studies have shown that government subsidies have a positive impact on firms’ innovations [5,6,7,8,9,10,11,12]. On the other hand, some studies have shown that there is no clear relationship between subsidies and firms’ innovation [13,14,15].

In addition, to our knowledge, few studies distinguish the heterogeneous effects of local subsidies and central subsidies except for Gao et al. (2021). Using data collected in manufacturing firms in one of China’s provinces (i.e., Jiangsu province), Gao et al. (2021) showed that local subsidies exhibited a more salient effect than central subsidies [16]. However, the findings of this study seem to contradict others that showed that the central government is more concerned about firms’ innovation capacity, while local governments are more concerned about firms’ economic performance [17,18,19].

The objective of this study is to investigate the impact of government subsidies on firms’ innovation. Specifically, this study has two objectives. First, this study will test whether government subsidies have a positive impact on firms’ innovation in seed companies in China. Second, this study will test whether the impact of local subsidies differs from that of central subsidies.

This study is based on data collected in China’s seed companies for two reasons. First, with the largest population and a per capita cultivated land far below the world average, China’s grain production and food security is not only a domestic issue, but also has important international implications [20]. Surprisingly, however, previous studies that estimate the impact of subsidies on firms’ innovation are all based on data collected in high-tech industries [13,15,21], manufacturing industries in China [10,11,14,22], and/or listed companies in China [23,24]. To our knowledge, no study has analyzed data collected from national representative agricultural companies. Second, China plays an important role in agricultural R&D worldwide. As a developing country, China’s investment and innovation in the agricultural field have been ignored for decades. However, after four decades of rapid economic development, China’s investment in R&D, including agricultural R&D, has increased substantially. According to Chai et al. (2019), China now outspends the United States on food and agricultural research on a purchasing power parity basis [2].

The rest of the paper is organized as follows. In the next section, we briefly introduce R&D investment and firm innovation in China’s agriculture. In the third section, we will first discuss the data used in this study. Then, we will descriptively analyze the relationship between subsidies and seed firms’ patent applications and patents granted. Econometric models and estimation results will be discussed in the fourth section. The final section concludes the paper with policy implications.

2. China’s Agricultural R&D Investment and Innovation

China had very little investment in agricultural R&D before the agricultural reform started at the end of the 1970s. Because China is a developing country and because of its policy of prioritizing the growth of heavy industry, the total investments in food and agricultural R&D were only USD 0.3 million in 1950 [2]. In addition, the Great Leap Forward policies, the Anti-Rightist Campaign, and the Cultural Revolution were highly disruptive and hindered the development of agricultural research in the following years [25,26]. Consequently, China’s total investment in agricultural R&D in 1978 was just USD 233 million (in 2011 prices), which is 0.11% of China’s gross domestic product (GDP) [2].

However, investment in R&D has accelerated since the 1980s, especially during the past two decades. According to national statistical data, both the expenditure of the central government and that of local governments increased significantly. As shown in Figure 1, the total expenditure of the central government on R&D was CNY 9.8 billion in 1990, while it was CNY 113.6 billion in 2019 (in 1990 prices). That is, the expenditure of the central government increased by 10.6 fold during 1990–2019. During the same time period, the expenditure of the local government on R&D increased from CNY 4.2 billion to CNY 192.4 billion, an increase of 44.8 fold.

Figure 1 also shows that the increase in local government expenditure on R&D is faster than that of the central government. The local government’s expenditure was 42.6% of the expenditure of the central government (9.8 billion) in 1990. However, in 2012, the expenditure of the local government surpassed that of the central government for the first time. In 2019, the expenditure of the local government was 69.3% higher than that of the central government. That is, the local government began to play a more important role in subsidizing science and technology.

To encourage the development of agricultural R&D, the Chinese agricultural research system has undergone a series of reforms focusing on shifting public agricultural R&D funding toward competitive grants and enabling the commercialization of research products emanating from public R&D agencies since the mid-1980s [27]. Since 2003, agricultural development has been identified as an area of national priority in China’s annual Central Government No. 1 Central Policy Documents [28]. The Chinese government implemented a number of policies that included the provision of low-interest loans, subsidies, and various tax incentives designed to encourage private companies to invest in agricultural R&D activities and strengthen intellectual property rights (IPR) protection [29,30,31,32,33]. These policy incentives have spurred a number of initiatives intended to promote agricultural innovation and increase investment in agricultural R&D [28].

Consequently, in 2010, China began outspending the United States on total food and agricultural R&D [2]. National statistical data showed that China spent USD 18.4 billion, which is more than 40% higher than that of the United States (13.1 billion), on food and agricultural R&D on a purchasing power parity basis in 2013 [2]. During 1995–2020, China’s R&D investment increased 40-fold after adjusting for the consumer price index (NSBC, various years). In addition, data from the National Statistics Bureau of China (NSBC) showed that total R&D investment was 2.4% of China’s total GDP in 2020 (NSBC, 2021).

Similar to the dynamics of agricultural R&D investment, both the number of patent applications and the number of patents issued have surged in China since 2000. According to national statistics (NSBC, 2021), the number of agricultural patent applications increased from 1845 in 1995 to 108,711 in 2018, while the number of patents issued increased from 1045 to 53,895 during the same time period, as shown in Figure 2. In other words, both the number of agricultural patent applications and the number of patents granted increased by more than 50 times during 1995–2018. Similar dynamics of agricultural patent applications and grants were found in other studies that were based on other patent classification taxonomies and data sources [34]. China has led the world in the number of patent filings since 2011 and filed 1.4 million patents, or 43.4 percent of the world’s total patent applications in 2019, according to the World Intellectual Property Indicators (WIPO) [35].

Previous studies have shown that the impact of central subsidies and local subsidies on patents might be different because local governments’ targets and evaluation criteria are often different from those of the central government [18,19]. To avoid subjective bias, the central government often adopts objective quantitative measurement criteria to evaluate the effect of subsidies [36]. However, rather than focusing on research capacity, most local governments actively engage in and are heavily involved in fostering the economic growth of firms [18]. The reason behind this difference is that local governments directly experience the benefits and costs of firms’ operations [17], which have an important impact on local officials. Due to the performance-based promotion scheme in China, career-minded local officials face pressure to increase their economic growth rate [37,38,39]. On the other hand, the average time period that a local official stays in one position is only approximately 2.8 years [40]. Consequently, local officials are more concerned about firms’ gross output than their technology stocks [41].

Consequently, we have the following hypothesis: central subsidies have a positive impact on firms’ innovation (in this study, innovation is measured by the number of patent applications and the number of patents granted), while local subsidies contribute to firms’ economic performance. We will use empirical data and econometric models to test this hypothesis.

3. Data and Models

3.1. Data

The firm-level data for various types of seed companies were collected in 2019 by the Center for Chinese Agricultural Policy, Peking University. The samples for this study were selected as follows. First, 1038 seed companies were selected from the total 5200 seed companies according to two criteria: (1) registered capital is more than CNY 10 million and (2) seed sales of corn, cotton, rice and soybean are more than 50% of the total sales of the company. Second, 200 companies were randomly selected from the 1038 companies. The selected 200 companies were then sorted according to their total sales in descending order. Then, all 200 companies were divided into 10 groups with 20 companies in each group. In each group, 10 companies were randomly selected, while the remaining 10 companies were labeled as alternative samples. During the field survey, seven companies finally accepted the interview, even though they refused the first time. Finally, eight listed companies that are engaging in the industrialization of transgenic technologies were also included in the survey. Our final data include 115 firms in 19 provinces in China. The sample distribution of these 115 firms is shown in Figure 3.

To collect detailed information about the sample companies’ investment in R&D, the number of patent applications, and the number of patents granted, a detailed questionnaire was designed. The questionnaire was re-edited several times during the pretest to reduce potential problems. The questionnaire includes several sections. The first section records the basic information about the sampled seed companies, such as when the seed companies were established, property rights of the company, total sales and seed sales, and total number of employees.

Second, the questionnaire has a long section that records the company’s investment in R&D. For each company with a separate R&D department, detailed information about the R&D department, such as the number of researchers, their education status, and their wage rates, was requested. Specifically, the questionnaire also asked whether this company had obtained subsidies from governments. If the answer was yes, then the company was asked for detailed information about the subsidy, such as how much the subsidy was, when the company received the subsidy, and whether the subsidy was from central government or local government.

At the same time, the survey team also collected a great deal of information about the company’s R&D patent applications and patents granted. Specifically, the survey team asked whether the firm had applied for patents. If the answer was yes, the survey team then asked for the number of patents for which the company had applied. Similarly, the number of patents granted by the company was recorded.

For all the input variables (e.g., subsidy, investment, number of researchers) and output (i.e., the number of patent applications and the number of patents granted) information, the survey team not only asked the numbers in 2018 (the latest year when the survey was conducted), but also for three more years: 2016, 2017, and 2010 (for those companies which were established after 2010, we asked for the information for the year when the company was established). By collecting four years of information for each seed company, panel data are formed. In total, this study includes 460 observations.

3.2. Subsidy and Firms’ Innovation

Basic characteristics of the sample are summarized in

Table 1. Basic characteristics of major variables used in this study.

Variable	Mean	Std.Dev
The number of patent applications	0.598	4.269
The number of patents granted	0.324	1.709
Central R&D subsidy (Million CNY)	0.718	4.311
Local R&D subsidy (Million CNY)	0.178	1.148
Total subsidy (Million CNY)	0.897	5.196
Firm’s R&D investment (Million CNY)	7.281	26.797
Entrusted investment (Million CNY)	0.593	3.419
Total business income (Million CNY)	155.670	628.300
Main business income (Million CNY)	89.775	268.464
Growth rate of total business income	0.043	0.516
Growth rate of main business income	0.052	0.527
Share of operation income	0.009	0.034
Herfindahl–Hirschman index (HHI)	0.138	0.036
Number of researchers in R&D department	20.765	58.417
Age of the firm (Year)	10.163	7.048
Large firm dummy (Yes = 1)	0.096	0.294
State-owned dummy (Yes = 1)	0.130	0.337
Private property right dummy (Yes = 1)	0.826	0.379
R&D cooperation dummy (Yes = 1)	0.435	0.496

Note: Except for HHI and share of operation income, all the other continuous variables are in their natural logarithmic form.

. As shown in Table 1, the average total business income is CNY 156 million, while the major business income is CNY 90 million (or 58% of the total income). The large standard deviations indicate that our sample includes both large and small seed companies. Table 1 also shows that, on average, there are 21 researchers in the R&D department, while the age of the company is more than 10 years. Finally, the vast majority of the sampled companies are private (83%), while the percentage of state-owned companies is less than 10%.

Consistent with Figure 1, the data used in this study show that both the number of patent applications and the number of patents granted increased rapidly. Even though the average patent application is only 0.598 and the average number of patents granted is 0.324 (rows 1 and 2 of Table 1), both have high increasing rates. Our data show that the average number of patent applications is only 0.03 in 2010, while it is 0.60 in 2018. That is, the number of patent applications increased by more than 20 times. Similarly, the average number of patents granted increased from 0 to 0.50 during 2010–2018.

The increases in patent applications and grants are accompanied by an increase in government subsidies. Rows 3 and 4 of Table 1 show that the average subsidies from the central government and local government are CNY 0.72 and 0.18 million, respectively (USD 1 = CNY 6.9 in 2020). Further study shows that the average subsidy from the central government increased from CNY 0.20 to 0.99 million during 2010–2019. That is, the central subsidy increased 3.86 times, with an annual growth rate of 42.85%. Similarly, the local subsidy increased from CNY 0.06 to 0.22 million, with an annual increasing rate of 31.31%.

Figure 4 attempts to link the subsidies with the number of patents. As shown in Panel A, there is a clear positive relationship between central subsidies and the number of patent applications. On the other hand, there seems to be no relationship between local subsidies and the number of patent applications. Replacing the number of patent applications with the number of patents granted, we obtained similar results. That is, Panel B of Figure 4 also shows that central subsidies positively affect the number of patents granted, but local subsidies do not.

4. Empirical Models

The descriptive analysis of the impact of central subsidies and local subsidies on firms’ innovation (i.e., patents applied and granted) might be misleading, as other factors confound the result. To isolate the impact of government subsidies and quantitatively measure their impact, we then set up and estimate an econometric model as follows:

$$\begin{array}{l}Paten{t}_{i,t}={\alpha }_0+{\alpha }_1*Subsid{y}_{i,t}+{\alpha }_2*Firm\_characteristic{s}_{i,t}+{\displaystyle \sum _{\mathrm{i}=1}^N{\alpha }_{3,i}*I{D}_i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\alpha }_4*Yea{r}_t+e_{i,t}\end{array}$$

(1)

In Equation (1), i is the i^th household; t is the t^th year; and e is the error term. The dependent variable, Patent, is used to measure firms’ innovation. As discussed above, we use two variables: (1) number of patent applications and (2) number of patents granted to measure firms’ innovation.

The first variable on the right-hand side of Equation (1), Subsidy, is used to measure the subsidy that the firm received. This vector includes three variables: central subsidy, local subsidy, and total subsidy. As discussed above, central subsidies are expected to have a positive impact on firms’ innovation. That is, the estimated coefficient of the variable should be positive. Different from the central government, as discussed in the second section, local governments are more concerned about firms’ economic performance (i.e., sales) rather than innovation. Hence, we expected the impact of local subsidies to be smaller than that of the central government. Finally, considering the correlation between central subsidies and local subsidies, we also include total subsidies, which is the sum of local subsidies and central subsidies.

The second vector, Firm_Characteristics, is a vector to measure the characteristics of the seed company. In this study, the Firm_Characteristics vector includes the basic characteristics of the firm, such as firm age, property rights of the firm (i.e., state-owned or private), and a large firm dummy (yes = 1 and no = 0). In addition, a firm’s investment in R&D, sales, total number of researchers in the R&D department, and the investment entrusted to other units are also included in this vector. Finally, an index to measure the power of the firm in the seed market, which is calculated by the ratio of the firm’s sale in the total seed market, is included in the vector. In this study, we use the Herfindahl–Hirschman Index (HHI) to measure the competitiveness of the seed industry, i.e., $HHI={\displaystyle \sum {\left(x_i/X\right)}^2}$, where $x_i$ is the total business income of the i^th firm and $X$ is the total industry income.

It should be noted that even though Firm_Characteristics is included in Equation (1), some characteristics of a firm might still be missed. For example, a firm’s ability is unobserved and hence is missed in Equation (1). With omitted variables, the estimation results of Equation (1) might be biased. To solve this problem, fixed effect models are estimated in this study (i.e., by adding ID dummies). It is worth noting that in the firm fixed effect models, some time-constant variables, such as the property rights of the firm and the large firm dummy, are dropped. Finally, year dummies (i.e., Year) are added in Equation (1) to capture the impact of variables varying over years.

The estimation strategy should be discussed before we move to the estimation results. Our data show that there are some firms without patent applications. That is, their number of patent applications is zero. Similarly, some firms have no patents granted. To solve this problem, the Tobit model is estimated.

5. Estimation Results

5.1. Impact of Subsidies on Firms’ Innovation

The estimation results of Equation (1) are shown in

Table 2. Determinants of patent applications (fixed effect Tobit model).

	Number of Patent Applications
	(1)	(2)	(3)	(4)	(5)
Central subsidy	0.093 *** (0.017)	0.075 *** (0.028)	−0.011 (0.042)
Local subsidy			−0.111 *** (0.040)	0.003 (0.020)	−0.104 *** (0.027)
Total subsidy		0.001 (0.024)	0.114 ** (0.047)		0.103 *** (0.022)
Private investment		−0.025 (0.017)	−0.027 (0.018)		−0.027 (0.017)
Entrusted investment		0.004 (0.020)	0.006 (0.021)		0.005 (0.021)
Age of firm		0.020 (0.068)	0.006 (0.067)		0.007 (0.067)
Total business income		0.034 * (0.019)	0.030 (0.020)		0.031 (0.020)
Share of operation income		3.615 ** (1.596)	2.856 * (1.606)		2.910 * (1.593)
Number of researchers		0.076 (0.057)	0.081 (0.056)		0.079 (0.055)
HHI		0.058 (0.188)	0.095 (0.187)		0.092 (0.187)
Sigma squared	0.097 *** (0.006)	0.094 *** (0.006)	0.093 *** (0.006)	0.104 *** (0.007)	0.093 *** (0.006)
Constant	−0.072 (0.158)	−0.636 ** (0.272)	−0.621 ** (0.270)	−0.102 (0.163)	−0.622 ** (0.269)
Firm dummies	Yes	Yes	Yes	Yes	Yes
Year dummies	Yes	Yes	Yes	Yes	Yes
Observations	460	460	460	460	460

Note: (1) Standard errors in parentheses; (2) * p < 0.1, ** p < 0.05, *** p < 0.01; (3) Except for HHI and share of operation income, all the other continuous variables are in their natural logarithmic form.

. Since the estimated coefficients are not marginal effects, we then calculate the marginal effects of their determinants, which are shown in

Table 3. Marginal effects of subsidies on firms’ innovations.

	Marginal Effect
	Number of Patent Applications			Number of Patents Granted
	(1)	(2)	(3)	(4)	(5)	(6)
Central subsidy	0.032 *** (0.012)	−0.005 (0.018)		0.026 ** (0.012)	−0.007 (0.017)
Local subsidy		−0.048 *** (0.017)	−0.045 *** (0.012)		−0.043 *** (0.016)	−0.038 *** (0.011)
Total subsidy	0.001 (0.010)	0.049 *** (0.020)	0.044 *** (0.010)	−0.013 (0.010)	0.030 (0.019)	0.024 *** (0.009)

Note: (1) Standard errors in parentheses; (2) ** p < 0.05, *** p < 0.01; (3) Except for HHI and share of operation income, all the other continuous variables are in their natural logarithmic form.

. As shown in Table 2, the estimated coefficient and standard error of sigma squared are statistically significant (row 11). That is, using the Tobit model is appropriate.

As expected, Table 2 confirms the positive impact of central subsidies on firms’ innovations. As shown in the first row, the estimated coefficients of central subsidies are positive (i.e., 0.093) and statistically significant (S.E. = 0.017) when other control variables are not included, indicating that a high central subsidy is associated with high firm innovation (first column). In addition, the positive impact holds if other control variables are included, as shown in the second column. Finally, when local subsidies are also included in the estimation, the estimated coefficient of central subsidies becomes insignificant (column 3). However, the estimated coefficient of the total subsidy becomes significantly positive (row 3). Further observation shows that the sum of the estimated coefficient of central subsidy and that of the total subsidy is close to that in column 1. That is, the positive impact of central subsidies on the number of patent applications holds. According to Table 3, if the central subsidy doubles, the number of patent applications increases by 3–4%, while the number of patents granted increases by 2–3%.

However, the impact of local subsidies has not been confirmed. As shown in the fourth column, the estimated coefficient of local subsidies is insignificant when control variables are excluded. After adding other control variables, the estimated coefficient of local subsidies is −0.104 and statistically significant (column 5). On the other hand, the estimated coefficient of total subsidy is statistically positive (i.e., 0.103), which is very close to the absolute value of the estimated coefficient of local subsidy. Further research shows that the sum of these two estimated coefficients is zero (p-value is 0.000). Similarly, when the central subsidy is also included in the model, the sum of the estimated coefficient of the local central subsidy (−0.111) and that of the total subsidy (0.114) is also zero, as shown in the third column (p-value is 0.016). That is, the estimation results show that local subsidies have no impact on firms’ innovation.

Replacing the dependent variable with the number of patents granted, we obtained similar results (

Table 4. Determinants of patent granted (fixed effect Tobit model).

	Number of Patents Granted
	(1)	(2)	(3)	(4)	(5)
Central subsidy	0.065 *** (0.016)	0.061 ** (0.027)	−0.015 (0.039)
Local subsidy			−0.099 *** (0.036)	−0.014 (0.019)	−0.088 *** (0.025)
Total subsidy		−0.031 (0.022)	0.069 (0.043)		0.055 *** (0.020)
Private investment		−0.029 (0.018)	−0.031 (0.019)		−0.030 (0.019)
Entrusted investment		0.041 ** (0.019)	0.042 ** (0.019)		0.041 ** (0.018)
Age of firm		0.107 * (0.062)	0.095 (0.062)		0.097 (0.061)
Total business income		0.056 *** (0.019)	0.053 *** (0.018)		0.054 *** (0.018)
Share of operation income		0.964 (1.467)	0.289 (1.477)		0.363 (1.464)
Number of researchers		0.107 ** (0.052)	0.112 ** (0.051)		0.109 ** (0.051)
HHI		−0.122 (0.173)	−0.089 (0.172)		−0.093 (0.172)
Sigma squared	0.085 *** (0.006)	0.080 *** (0.005)	0.078 *** (0.005)	0.088 *** (0.006)	0.078 *** (0.005)
Constant	−0.077 (0.148)	−0.833 *** (0.250)	−0.820 *** (0.248)	−0.102 (0.150)	−0.822 *** (0.248)
Firm dummies	Yes	Yes	Yes	Yes	Yes
Year dummies	Yes	Yes	Yes	Yes	Yes
Observations	460	460	460	460	460

Note: (1) Standard errors in parentheses; (2) * p < 0.1, ** p < 0.05, *** p < 0.01; (3) Except for HHI and share of operation income, all the other continuous variables are in their natural logarithmic form.

). As shown in columns 1 and 2, the estimated coefficients of central subsidies are positive and statistically significant, indicating that central subsidies contribute to the number of patents granted (the F value of the test that both the estimated coefficients of central subsidy and total subsidy are 2.77, with a p-value of 0.064). On the other hand, the estimated coefficient of local subsidies is insignificant when other variables are not controlled (column 4), and the sums of the estimated coefficients of local subsidies and total subsidies are not more than zero (columns 3 and 5). That is, local subsidies do not contribute to the number of patents granted.

To check the robustness of the estimation result of the Tobit model, we rerun the model using firm-level fixed effect OLS models. The estimation results are available upon request. The estimation results of the fixed effect OLS model are consistent with those of the fixed effect Tobit models (i.e., as shown in Table 2 and Table 3). That is, central subsidies have a positive impact on firms’ innovation (i.e., the number of patent applications and the number of patents granted), while the impact of local subsidies has not been evidenced.

5.2. Why Has the Impact of Local Subsidies Not Been Evidenced?

As discussed above, unlike the central government, local governments are more concerned about the current economic performance of seed companies. That is, the major objective of local government is firms’ gross outputs, rather than the patents for which they applied and were granted. To check this hypothesis, we then set up and run the following econometric equation:

$$\begin{array}{l}Sal{e}_{i,t}={\beta }_0+{\beta }_1*Subsid{y}_{i,t}+{\beta }_2*Firm\_characteristic{s}_{i,t}+{\displaystyle \sum _{\mathrm{i}=1}^N{\beta }_{3,i}*I{D}_i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\beta }_4*Yea{r}_t+{\epsilon }_{i,t}\end{array}$$

(2)

In Equation (2), the dependent variable, Sale, is used to measure a firm’s economic performance, which is measured by the total income and its growth rate. To test whether other businesses affect the estimation results, we also use income from the main business (i.e., seed sales) and its growth rate as the dependent variable. Finally, Ɛ is the error term. All the other variables have been discussed.

The estimation results are shown in

Table 5. Impact of subsidies on the economic performance of seed companies.

	Total Business Income			Main Business Income
	(1)	(2)	(3)	(4)	(5)	(6)
Central subsidy	0.058 (0.054)		0.068 (0.054)	0.089 (0.074)		0.105 (0.074)
Local subsidy	0.079 (0.056)	0.087 * (0.052)		0.133 * (0.077)	0.144 * (0.076)
Age of company	1.156 *** (0.173)	1.162 *** (0.173)	1.168 *** (0.174)	0.952 *** (0.236)	0.961 *** (0.237)	0.972 *** (0.237)
No. of researchers	0.748 *** (0.149)	0.804 *** (0.139)	0.782 *** (0.147)	−0.176 (0.203)	−0.089 (0.190)	−0.118 (0.201)
Share of operation income	1.461 *** (0.522)	1.443 *** (0.522)	1.377 *** (0.519)	1.044 (0.713)	1.016 (0.712)	0.903 (0.710)
Private investment	0.200 *** (0.047)	0.192 *** (0.047)	0.200 *** (0.048)	0.559 *** (0.064)	0.548 *** (0.063)	0.559 *** (0.065)
Entrusted investment	−0.068 (0.057)	−0.048 (0.054)	−0.061 (0.057)	−0.057 (0.078)	−0.026 (0.073)	−0.046 (0.078)
HHI	−6.258 *** (1.281)	−6.320 *** (1.280)	−6.290 *** (1.283)	−5.754 *** (1.749)	−5.850 *** (1.748)	−5.809 *** (1.754)
Constant	9.060 *** (0.884)	9.043 *** (0.884)	9.030 *** (0.885)	8.787 *** (1.206)	8.761 *** (1.207)	8.736 *** (1.210)
Firm dummies	Yes	Yes	Yes	Yes	Yes	Yes
Year dummies	Yes	Yes	Yes	Yes	Yes	Yes
Observations	460	460	460	460	460	460

Note: (1) Standard errors in parentheses; (2) * p < 0.1, *** p < 0.01; (3) Except for HHI and share of operation income, all the other continuous variables are in their natural logarithmic form.

. Surprisingly, the estimated coefficient of central subsidies and that of local subsidies are positive but insignificant (column 1, rows 1 and 2). Considering the correlation of these two variables, these estimated coefficients might be misleading. To solve this problem, we drop one of these variables and rerun the model. The estimation results are shown in the second and third columns.

As expected, column 2 shows that the central subsidy has no impact on firms’ total income, but column 3 shows that the local subsidy has a positive relationship with firms’ total income. As shown in the second column, after dropping the local subsidy, the estimated coefficient of the central subsidy is still insignificant, confirming that the central subsidy has no impact on a firm’s total income. On the other hand, the estimated coefficient of local subsidies is positive and statistically significant after dropping the central subsidy (row 2). That is, the estimation results confirm that the local subsidy has a positive impact on a firm’s total income.

Replacing the total income by the income from the main business (i.e., seed sales), we obtained similar results. As shown in the fourth column, the estimated coefficient of central subsidies is insignificant, but the estimated coefficient of local subsidies is statistically positive. Removing one of these two variables and rerunning the model, we obtained the same results (columns 4 and 5).

Similar results are obtained if we consider the growth rate of total income and that of income from the main business. As shown in

Table 6. Impact of subsidies on the growth rate of seed companies.

	Growth Rate of Total Business Income			Growth Rate of Main Business Income
	(1)	(2)	(3)	(4)	(5)	(6)
Central subsidy	0.034 (0.056)		0.032 (0.057)	0.032 (0.058)		0.030 (0.058)
Local subsidy	0.136 ** (0.061)	0.135 ** (0.062)		0.138 ** (0.063)	0.137 ** (0.062)
Age of company	−0.658 * (0.353)	−0.665 * (0.352)	−0.668 * (0.356)	−0.442 (0.360)	−0.449 (0.359)	−0.453 (0.363)
Total business income	0.268 ** (0.122)	0.273 ** (0.121)	0.268 ** (0.123)	0.288 ** (0.124)	0.293 ** (0.123)	0.289 ** (0.125)
Share of operation income	−0.276 (0.201)	−0.283 (0.200)	−0.269 (0.203)	−0.338 (0.205)	−0.339 (0.207)	−0.330 (0.207)
No. of researchers	−0.245 (0.220)	−0.231 (0.218)	−0.236 (0.222)	−0.365 (0.225)	−0.351 (0.224)	−0.356 (0.227)
Private investment	−0.012 (0.060)	−0.009 (0.060)	−0.020 (0.061)	−0.001 (0.061)	0.003 (0.062)	−0.008 (0.062)
Entrusted investment	0.011 (0.062)	0.021 (0.060)	0.012 (0.063)	0.012 (0.063)	0.021 (0.062)	0.013 (0.064)
HHI	−0.100 (0.287)	−0.097 (0.286)	−0.049 (0.289)	−0.298 (0.293)	−0.295 (0.292)	−0.246 (0.295)
Constant	0.831 (0.724)	0.793 (0.721)	0.877 (0.730)	0.720 (0.740)	0.686 (0.736)	0.768 (0.746)
Firm dummies	Yes	Yes	Yes	Yes	Yes	Yes
Year dummies	Yes	Yes	Yes	Yes	Yes	Yes
Observations	345	345	345	345	345	345

Note: (1) Standard errors in parentheses; (2) * p < 0.1, ** p < 0.05; (3) Except for HHI and share of operation income, all the other continuous variables are in their natural logarithmic form.

, the estimated coefficient of central subsidies is insignificant (row 1), while the estimated coefficient of local subsidies is positive and significant (row 2). That is, Table 6 confirms that the central subsidy has no impact on firms’ economic performance, while the local subsidy has positively affected firms’ growth rate.

Even though this study focuses on the heterogeneous effects of central and local subsidies on firms’ innovation, and the impact of different innovations on sustainability varies, it is worth noting that there is a debate on the opportunities and limits of subsidies [42]. Some studies showed that the relationship between technological innovation and sustainability is positive in the energy industry [43,44,45,46] and agricultural sector [47]. Other studies showed that innovation either fails to meet the sustainable goals [48] or the impact varies over time [49,50] and across counties [51,52]. In addition, Heyl et al. (2022) showed that agricultural subsidies need to be substantially downscaled and implemented as complementary instruments in transitioning the agricultural sector to be in line with global environmental goals.

6. Conclusions

From a poor developing country to the second largest economic power in the world, China is playing an increasingly important role in global R&D [2]. To encourage firms to invest in R&D, both the central government and local governments provide substantial subsidies. However, the empirical analysis of this study shows that the impact of central subsidies is different from that of local subsidies. Specifically, central subsidies have a positive impact on firms’ innovation, while local subsidies have a positive relationship with firms’ economic performance. That is, the formulated hypothesis is tested and confirmed.

The results of this study have important policy implications. Subsidies are pervasive in the agricultural sector, even though there is a debate on the opportunities and limits of subsidies [42]. To strengthen firms’ research capacity, the Chinese central government not only directly provides subsidies to seed companies, but also encourages/requires local governments to provide subsidies [53]. However, local governments are more concerned about firms’ current economic performance (i.e., gross output) than their innovations [18] because of the performance-based promotion scheme in China for local officials [37,38,39]. The results of this study confirm that local subsidies did not contribute to the development of firms’ capacity in R&D, but rather to current economic performance. To address this issue, local governments should implement similar methods to those of the central government in research project funding and criteria for selecting research projects.

Although this study offers insight into heterogeneous effects of central subsidies and local subsidies, it is not without limitations. First, due to data availability, this study is based on four-year data of 115 seed companies. Although the existing results are sufficiently robust, the small number of samples prevents us from discussing the heterogeneous mechanism of central subsidies and local subsidies. Second, all the samples are selected from seed companies. That is, whether the heterogeneous effects of central subsidies and local subsidies in agricultural R&D hold in other industries should be investigated. This is also a very important direction for our future study.