Internal Relationship and Impact Path Between Innovation and Entrepreneurship: Based on China’s High-tech Industry

Innovation is the source of entrepreneurship, entrepreneurship is the value embodiment of innovation, and the two are inseparable. At a time when dividends such as population, reform and opening up, and resources and environment are gradually disappearing, China urgently needs to accelerate scientific and technological innovation to support economic development, incubate scientific and technological enterprises, and ease labor market pressure with technological progress and efficiency improvement. This paper focuses on China’s high-tech industry, which is dominated by scientific and technological innovation. Starting from the overall, local, and regional perspectives, it organically integrates the traditional DEA, similar SFA, Malmquist index decomposition, chain multiple intermediary effect, and other multilevel research through cross-level analysis. Based on the research foundation of innovation efficiency after eliminating environmental and random factors, it deeply discusses the action path and impact mechanism of “double innovation” and provides targeted policy recommendations for the government and relevant local departments. The research confirms that the total effect of innovation on entrepreneurship is always positive, i.e., promoting “people-to-people innovation” is conducive to promoting “mass entrepreneurship” whether it is analyzed from the whole or from the part.


Introduction
Innovation is the source of starting a business. In the process of starting a business, one needs to have a continuous and vigorous sense of innovation before one can produce creative ideas or schemes and explore new modes and outlets. Entrepreneurship is the embodiment of the value of innovation. The value of innovation mainly lies in transforming technology into productivity, and the fundamental way to realize this transformation is entrepreneurship. It can be seen that innovation and entrepreneurship are inseparable. Studying the internal relationship and mechanism between the two can not only provide an important theoretical basis and guidance for China to promote "double innovation" and prevent "deviation and deviation" in the middle of the process but also lead enterprises to carryout efficient innovation input and output, accelerate enterprises to adapt to the new normal, and provide power and guarantee for promoting innovationdriven strategy in depth. 1 1 The corresponding path of chain multiple mediation effect is shown in Figure 1. As shown in Figure 1, the multiple mediation model includes two mediation variables M 1 and M 2 , at this time, the multiple mediation effect analysis can be carried out from three angles. The first is to analyze from the total mediation effect, namely: a 1 b 1 +a 2 b 2 +a 1 a 3 b 2 . Secondly, on the premise of controlling other intermediate variables (such as control M 1 ), we can study the specific mediation effect of each intermediate variable, such as a 1 b 1 , a 2 b 2 and a 1 a 3 b 2 . Third, a comparative mediating effect can be obtained so as to be able to judge which of the effects of multiple mediating variables (e.g. a 1 b 1 , a 2 b 2 ) is more effective, e.g a 1 a 3 b 2 -a 2 b 2 , and a 1 b 1 -a 2 b 2 , a 1 a 3 b 2 -a 1 b 1 (Preacher and Hayes [16] ).

Input-output indicators
Data on capital input, personnel input and innovation output are all derived from China's High-tech Industry Statistical Yearbook. As capital investment data are often affected by price fluctuations, the relevant data should be reduced by using fixed asset investment price index (last year = 100) of corresponding years in all provinces and cities, and the provincial fixed asset investment price index data of each year comes from China Statistical Yearbook. In addition, Tibet and Xinjiang were excluded from the sample due to the serious lack of data. In order to eliminate the time lag between input and output and the difference between innovation activity cycles, this study chooses the time lag as one year, i.e. the input data is 2004-2013 and the output data is 2005-2014. In addition, because the internal expenditure of R&D funds, expenditure of new product development and expenditure of technological transformation not only have an impact on the input-output efficiency of high-tech industries in the current period, but also will have an important impact in some future periods, it is first necessary to convert the flow data into stock and then measure the impact of these three capital inputs on the innovation-output efficiency. Taking the calculation of R&D fund stock as an example, the method of permanent storage is used to deal with it.
In the Eq(2), x i,t and x i,t-1 are the R&D expenditures of the t and t-1 years of the province i respectively; e i,t-1 is the R&D expenditure of the t-1 year after eliminating the price factor; δ is the depreciation rate, according to Griliches [17] can set it to 15% δ = Assuming that the average annual growth rate of R&D expenditures during the study period is g, the base R&D expenditure is  (2) and (3), the R&D fund stock of each year can be obtained by using the perpetual inventory method. The above method can be used to process the funds for new product development and technological transformation, thus converting the flow data into stock. Input-output related variables are described in Table 1.  Ten thousand yuan The stock of R&D funds is calculated by the perpetual inventory method. I n n o v a t i o n ( c a p i t a l ) investment 3 x3 Ten thousand yuan The stock of new product development funds shall be accounted for by the perpetual inventory method. I n n o v a t i o n ( c a p i t a l ) investment 4 x4 Ten thousand yuan The stock of funds for technological transformation shall be accounted for by the perpetual inventory method.

Environmental variables
Environmental variables are mainly used to reflect the geographic location of high-tech industries, macroenvironment, and government innovation support policies. Based on the relevant research foundation at home and abroad, and considering the characteristics of R&D and the availability of data, this paper mainly selects the following indicators:

Regional real GDP
Real GDP can accurately reflect regional differences and the real level of regional economic development, so this paper uses real GDP excluding price factors instead of nominal GDP. Generally speaking, the higher the regional economic development level, the more capable it is to invest in R&D innovation in high-tech industries, and the higher the R&D innovation level.

Geographical location
Whether a region is located in a superior geographical position is crucial to the development of high-tech industries in the region. This paper quantifies this variable by setting virtual variables, in which "1" is used to represent the eastern region, including Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong and Hainan. "2" means the central region, including Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei and Hunan; "0" means the western region, including Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Guangxi and Ningxia.Due to the serious lack of data in Xinjiang and Tibet, this article will not consider it for the time being.

Quality of local workers
In view of the availability of data, this paper chooses the number of college students in the studied area as the characterization variable of laborer's quality. The higher the quality of local laborers, the more likely it is to provide more talents and intellectual support for the development of high-tech industries, and the more conducive to the improvement of R&D efficiency.

Local government innovation support
Local government support has an important impact on the R&D and innovation process of high-tech industries. This paper selects the government funds from the funds raised for scientific and technological activities to represent the innovation support of local governments. Generally speaking, local government support can provide guarantee for the infrastructure construction of high-tech industries in the region, as well as financial support. The greater the support, the more conducive to R&D and innovation in high-tech industries.

Regional science and technology development level
Due to the spatial aggregation effect and spillover effect of high-tech industries, the degree of technological development in a region often plays an important role in the development of high-tech industries. This paper chooses the number of high-tech industrial enterprises and the number of scientific and technological institutions to represent the degree of scientific and technological development in a region. In general, the more high-tech enterprises and scientific and technological institutions there are, the more conducive they are to R&D and innovation in high-tech industries.
Among the above environmental variables, except that the actual GDP of each region comes from China Statistical Yearbook, other data come from China High-tech Industry Statistical Yearbook. Descriptive statistics of relevant variables are shown in Table 2.

Mediating effect variables
Taking the total factor productivity (X) obtained by Malmquist index decomposition in the third stage as the independent variable. The dependent variable is the logarithm (Y) of the number of newly-built entrepreneurs in individual and private enterprises in high-tech industries.  Tables 3, 4 and 5. According to the calculation results before adjustment, the average value of China's R&D comprehensive technical efficiency is 0.48-0.67 from 2005 to 2014, without considering the interference of external environment and random influence factors, which generally shows a rising trend. The average value of pure technical efficiency is 0.63-0.73, with room for improvement to varying degrees. The average value of scale efficiency is 0.753-0.926, and the overall efficiency level is relatively high. Generally speaking, the R&D efficiency of most provinces is below the efficiency frontier. As the result does not eliminate the influence of circumstantial factors and random interference factors, it cannot accurately measure the true level of the comprehensive technical efficiency of each province, so further adjustment and measurement are needed. Note: "BA" means Before Adjustment, "AA" means After Adjustment  Note: "irs" indicates increasing scale efficiency; "crs" indicates that the scale efficiency is unchanged; "drs" indicates a decline in scale efficiency. "BA" means Before Adjustment, "AA" means After Adjustment

Regression analysis of SFA in the second stage
In order to separate the efficiency values influenced by external environment and random error factors, this paper observes the influence of circumstantial factors and random error respectively by constructing similar SFA models in the second stage. SFA regression analysis is carried out by Frontier software. The relaxation values of each input variable obtained in the first stage are taken as dependent variables, and the selected 6 environmental variables are taken as independent variables to establish regression models to estimate the influence of circumstantial factors on the relaxation values of each input variable. If the regression coefficient is negative, it indicates that the positive change of environmental variables is conducive to reducing the input redundancy of high-tech industries, thus avoiding the loss and waste of input and improving the efficiency of research and development innovation. On the contrary, it will reduce the efficiency level of R&D innovation. The results of SFA regression analysis are shown in Table 6.  As can be seen from Table 6, the estimation results 2 σ and γ are both significant at the level of 1%, indicating that it is necessary to eliminate the influence of environmental variables in the model. Specifically, it can be seen from the following aspects: (1) Regional actual GDP has a significant negative impact on R&D personnel's FTE slack, new product development spending slack, and technological transformation spending slack. This shows that the higher the level of regional economic development, the more conducive it is to reduce the redundancy of input variables, and the more effective it is to promote the improvement of R&D innovation efficiency in high-tech industries. (2) Geographical location has a positive effect on all input slack variables at a highly significant level, and compared with other environmental variables, geographical location has a stronger effect on all kinds of innovation input slack.The results show that the eastern region has higher investment redundancy compared with the central and western regions. This is mainly due to the greater demand for innovation in the eastern region, which often spends huge sums of money on research and development innovation, thus easily causing excessive investment in innovation. (3) The influence of regional workers' quality on four types of input slack variables also shows significant positive effects. The higher the number of college students in the region, the higher the quality of workers in the region. However, on the other hand, with the continuous improvement of China's education level, the number of talents engaged in high-tech industries is increasing and even reaching saturation, which to a certain extent increases the redundancy of input variables and reduces the innovation efficiency level of China's high-tech industries. (4) The number of enterprises in high-tech industries has a significant negative impact on the relaxation of R&D funds, new product development funds and technological transformation funds, which shows that the increase of competition in the industry can effectively reduce the redundancy of capital investment and is beneficial to the improvement of R&D innovation efficiency. (5) Local government innovation support has significant positive effects on four types of input slack variables, which indicates that the stronger the government's support for high-tech industries, the easier it is to increase the redundancy of each input, and the more unfavorable it is to improve innovation efficiency. This may be because some inappropriate subsidy policies adopted by the government distort the market, undermine the selfregulating function of the market and hinder the improvement of innovation efficiency. (6) The number of scientific and technological institutions has a significant positive effect on R&D personnel FTE slack variables, while it has a significant negative effect on R&D funds, new product development funds and technological transformation funds. This shows that although the increase in the number of scientific and technological institutions will increase the redundancy of human capital investment and hinder the improvement of innovation efficiency, it can effectively reduce the waste of R&D funds, new product development funds and technological transformation funds, thus promoting the improvement of innovation efficiency.
To sum up, environmental variables have significant effects on input redundancy and their effects are different. Therefore, it is necessary to eliminate the effects of circumstantial factors and random factors when studying the true level of innovation efficiency.

DEA empirical analysis and malmquist index
Using DEAP software again, DEA analysis is conducted based on the original output data and the adjusted input variable data. Based on the results shown in Tables 3, 4 and 5 (after adjustment), the estimated outcomes are shown in Table 6.
Overall, the efficiency values before and after the adjustment have changed. The adjusted average technical efficiency is 0.25-0.53, the average pure technical efficiency is 0.98-0.96, and the average scale efficiency is 0.254-0.556.Compared with before adjustment, the average value of technical efficiency in different years has decreased, but the average value of pure technical efficiency has increased significantly, while the average value of pure technical efficiency is relatively stable, approaching 1.The average value of scale efficiency has not only dropped significantly, but also there are only two states after the adjustment: increasing scale returns and constant scale returns. This shows that compared with the situation before the adjustment, China's high-tech industry does not actually have the problems of excessive scale and overcapacity. From a local point of view, the regions where the average value of technical efficiency has increased include Beijing, Shanghai, Jiangsu, Zhejiang, Shandong, Guangdong and Shaanxi, indicating that the low efficiency of these provinces before adjustment is due to circumstantial factors. The average value of comprehensive technical efficiency in other regions has decreased, which indicates that the high level of innovation efficiency shown in these regions is mainly influenced by favorable environment and random factors.
In order to further investigate the changes in total factor productivity of China's high-tech industries, this paper constructs Malmquist index model based on the DEA model adjusted in the third stage. The results are shown in Tables 7,  8 and 9. The columns in the table are the index changes for every two consecutive years from 2005 to 2014 and the index changes for the whole period from 2005 to 2014. The last two rows are the arithmetic average and geometric average of the index respectively. Generally, geometric mean is used to calculate average ratio and average speed, so this paper uses this value to represent the average growth rate of total factor productivity, technical efficiency index and technical progress index. Table 10 summarizes Malmquist index and its decomposition index of technological innovation in order to meet the needs of total factor productivity of R&D innovation in various regions in subsequent research.  This shows that the increase in technical innovation efficiency in China's high-tech industries in recent years is due to the improvement in management level and technical efficiency. In addition, from a local point of view, except for Tianjin, Shanghai and Fujian, where the total factor productivity has not increased significantly, the total factor productivity in other places has increased by different ranges, with Hainan having the largest growth range, reaching a growth rate of 64.7%. Except Hebei, Liaoning, Heilongjiang, Zhejiang, Anhui and Hunan, the index of technological progress in other regions showed slow growth, with Beijing having the largest growth rate, with a growth rate of 5.4%.

Chain multiple mediation effect analysis
In order to clarify the internal mechanism of China's high-tech industry innovation and "mass entrepreneurship" and find the key to accelerate the process of "double innovation", this paper adopts a multi-chain intermediary effect model to analyze the direct effect and various indirect effects between innovation and entrepreneurship from the overall, regional (east, middle and west) and local (provinces and cities) perspectives.

Overall and regional
The overall analysis results of the multiple chain intermediary effect model are shown in Figure 2, and the effect results are shown in Table 11.    On the whole, the total effect of innovation on entrepreneurship is 1.769, that is, the increase of total factor productivity is conducive to the increase of the number of new entrepreneurs. However, from a further perspective, the direct effect of innovation on entrepreneurship is shown as inhibition (c' < 0), while the mediating effect through different paths shows a significantly higher promotion effect (a 1 b 1 , a 2 b 2 , a 1 a 3 b 2 are significantly positive).This shows that the promotion effect of innovation on entrepreneurship is not a simple direct effect. The government cannot rely solely on improving innovation efficiency to promote entrepreneurship so as to relieve the pressure on China's labor market. Instead, it should promote entrepreneurship by increasing new ventures and promoting economic growth on the basis of promoting innovation drive.

Effect path of western region
From the point of view of the east, middle and west regions, the total effect of innovation on entrepreneurship is promotion, and the promotion in the east region is obviously stronger than that in the middle and west regions, which may be caused by the superior geographical position and relatively developed economic level of the east region. The direct effects in the east and west are significantly negative, indicating that innovation will play a restraining role in the direct path of entrepreneurship. However, the indirect effects are significantly positive, especially the path in the eastern region, whose influence is much stronger than other paths, which shows that the eastern region mainly promotes entrepreneurship through the path of "innovation → enterprise increase → economic growth → entrepreneurship". For the central region, whether it is a direct or indirect path, the impact of innovation on entrepreneurship always presents a significant positive effect.

Provinces, cities and regions
The analysis results of direct effect, indirect effect and total effect of innovation on entrepreneurship in various provinces and cities are shown in Table 13. The direct effects of Beijing, Shanxi and Jilin are significantly positive, while the direct effects of Yunnan, Hainan, Guangdong and Fujian are significantly negative, while the direct effects of most other provinces are not significant. On the contrary, in terms of indirect effects, most provinces have significant positive effects, while only Beijing, Shanxi and Shanghai have negative effects.

Conclusions and policy recommendations
This paper decomposes total factor productivity on the basis of DEA model excluding environment and random factors and then analyzes the influencing process and mechanism among variables by using chain multiple intermediary effect model and further analyzes the relationship between innovation and entrepreneurship in high-tech industries. The research confirms: (1) Before circumstantial factors and random interference are eliminated, the comprehensive technical efficiency and scale efficiency of R&D in China's high-tech industry are obviously overestimated, while the pure technical efficiency is underestimated. Therefore, it is necessary to eliminate environmental and random factors and increase the credibility of the research. (2) No matter from the overall or partial analysis, the total effect of innovation on entrepreneurship has always been positive, that is, promoting "people-to-people innovation" is conducive to promoting "mass entrepreneurship." (3) In most cases, innovation does not directly promote entrepreneurship but indirectly promotes entrepreneurship through a micro-macro combination path.

Relevant policy recommendations
Generally speaking, the government can encourage China's high-tech industry to explore new market areas and gain competitive advantages by formulating relevant industrial policies and promoting innovation, thus providing motive force for innovation drive. Through continuous entrepreneurship, innovative achievements will be continuously transformed into real productive forces to promote the sustainable development of high-tech industries and social economy. We should strengthen the combination of "production, study, and research" and focus on improving the "double-creation" mechanism and the "double-creation" environment.
First of all, as the direct effect of innovation on entrepreneurship is positive in Beijing, Shanxi, and Jilin, the government and relevant local departments can appropriately increase the amount of financial allocation for research and development of high-tech products and conquering high-end technologies, strengthen the government's financial support, and give full play to the policy's guidance and support for innovation by strengthening the government's innovation subsidies and tax incentives, thus realizing a "double harvest" of innovation and entrepreneurship on the basis of effectively improving the innovation efficiency of high-tech industries in the region. Secondly, in Fujian, Guangdong, Hainan, and Yunnan, the government should make great efforts to promote the development of other industries while developing high-tech industries, so that the local advantageous industries and characteristic industries can develop continuously and steadily and all kinds of industries will be put together. Thirdly, in Tianjin, Heilongjiang, Zhejiang, Fujian, and Qinghai, the government should actively promote economic development. While making great efforts to improve GDP, it should also pay attention to the per capita level of GDP, scientific and technological content, and green degree and strive to provide a good macroenvironment for promoting "double innovation" by correctly grasping the "three improvements." Finally, in Beijing, Shanghai, and Shanxi, the number of new high-tech enterprises should be controlled to ensure that a reasonable market structure can not only bring into play economies of scale but also promote effective competition, thus effectively stimulating the improvement of innovation efficiency and entrepreneurship scale.