Inﬂuence Mechanism of Multidimensional Structure of External Knowledge Search on Firm Performance: Based on Chinese Manu-facturing

. In the context of open innovation, the diversiﬁcation of external innovation resources makes more and more enterprises choose to obtain new knowledge from the outside and search for external knowledge, so as to achieve the purpose of enterprise innovation and development, improve quality and increase e ﬃ ciency. This paper discusses the di ﬀ erent inﬂuences of the multidimensional spatial structure of external knowledge search on corporate ﬁnancial performance and innovation performance. Through the empirical analysis of the knowledge search data of 2053 Chinese manufacturing listed companies, it is found that di ﬀ erent external knowledge search structures have di ﬀ erent e ﬀ ects on corporate performance. Compared with the three-dimensional development search structure, exploratory search in any dimension is not conducive to the improvement of corporate ﬁnancial performance and innovation performance


Introduction
In recent years, with the deepening implementation of the national strategy of innovationdriven development and "Made in China 2025", it is extremely important to promote innovative development and improve the quality and efficiency of the manufacturing industry. Therefore, the Ministry of Industry and Information Technology and other eight departments jointly issued the "14th Five-Year Plan" for the development of intelligent manufacturing, proposing to implement the four key tasks of innovation, application, supply and support during the "14th Five-Year Plan", and vigorously promote the high-quality development of China's manufacturing enterprises.
In the context of open innovation, the diversification of external innovation resources makes more and more enterprises choose to acquire new knowledge from the outside, to make up for the lack of internal technical knowledge, reduce the risk of innovation, quickly seize the market, track advanced technology, so as to improve enterprise performance. At present, scholars have conducted a large number of studies on the relationship between external knowledge search and firm performance. In the process of studying the influence of external knowledge search on firm performance, it is found that external knowledge search has a positive promoting effect on firm performance [1,2]. However, there is usually an inverted U-shaped relationship between the two, that is, no matter an enterprise conducts external knowledge search along one or two dimensions, when it reaches a certain critical point, the positive effect of external knowledge will be weakened or even have a negative effect [3,4]. Drawing on the idea of organizational binary balance, scholars began to emphasize the multidimensional structure of external knowledge search [5,6]. In other words, in the process of external knowledge search, it is better not to go to the two poles along any single dimension. However, the existing researches fail to incorporate all dimensions into an overall research framework, which makes the external knowledge search space lack of integrity, and the relevant theoretical research needs to be further carried out.
Based on the above background, this paper uses 9,477 knowledge search data collected from 2,053 listed manufacturing companies in China from 2010 to 2019 to study the influence of the multidimensional spatial structure of external knowledge search on firm performance as a whole. This paper provides three aspects of research contributions: First, this paper considers the multidimensional spatial structure characteristics of knowledge search as a whole, and makes an in-depth study of the balanced search structure of the three dimensions of time, geography and technology, which enriches the research literature on multidimensional knowledge search balance. Secondly, the impact of existing research on the multidimensional spatial structure is mainly focused on the innovation aspects of enterprises, such as breakthrough technology [6], quality of innovation [7], and continuous innovation by enterprises. This paper further expands the research field of enterprise performance and its influencing factors from the perspective of differentiation. Finally, through the empirical study of Chinese manufacturing enterprises, this paper further puts forward the different influences of the multidimensional structure of external knowledge search on the performance of manufacturing enterprises, which provides a new research idea for the innovation and development, quality and efficiency improvement of Chinese manufacturing industry.

Theoretical Basis
In the context of open innovation, knowledge search for enterprise technological innovation has attracted the attention of foreign scholars since the 1980s, and has also attracted the attention of domestic scholars in recent years. From the perspective of Enterprise Boundary, enterprises can effectively expand the knowledge base of the organization and promote the recombination of different knowledge elements by searching for knowledge beyond the enterprise boundary, thus promoting the improvement of enterprise innovation performance [8]. From the perspective of professional knowledge search, under the technology-oriented search strategy, enterprises acquire technical knowledge through the "Inside-out" search process, which is the main resource for enterprises to develop new products [9,10]. According to the theory of open innovation, under the background of highly uncertain competitive environment and the increasing complexity of innovation, it will be difficult for enterprises to gain long-term competitiveness only by relying on their own existing knowledge base. More valuable knowledge resources exist outside enterprises. Therefore, enterprises should adopt an open knowledge search method, break the traditional way of thinking, establish more extensive and in-depth external contacts, so as to obtain diversified innovative knowledge and realize technological innovation of enterprises.
However, due to the stickiness of resources themselves and the path dependence of core competencies [11][12][13]. In order to reduce transaction costs and potential risks, enterprises tend to acquire knowledge from familiar and fixed external entities. In addition, influenced by many factors such as heterogeneous resources and external environment, enterprises have different choices in different dimensions of knowledge search [7]. As a result, different enterprises have different external knowledge search structures, which may be completely exploitative search or exploratory search, but more knowledge search is based on the balance between exploration and development. Relevant studies show that the cross-domain balanced search behavior is an important factor for organizations to get rid of path dependence and realize technology leapfrog [14]. On this basis, Zhou et al through the empirical study, it is found that enterprises can effectively promote the generation of breakthrough technologies by conducting cross-domain balanced exploratory and exploitative knowledge search among various dimensions [6]. Therefore, this paper draws on Li et al's three search dimensions of time, geography and technology [15], and divides these three dimensions into exploratory search and exploitative search respectively according to the organizational duality theory. Among them, the geographical dimension is divided according to the geographical distance of knowledge search, the distance of knowledge search is divided into exploratory search, and the other way is development search. The division of time dimension is based on the old and new degree of knowledge. The old knowledge with a long time to search knowledge is divided into development search, and the other way is exploratory search. The division of technical dimension is based on the enterprise's familiarity with the knowledge to be searched, the knowledge in the familiar technical field is divided into development search, otherwise it is exploratory search.
In addition, the key difference between this study and previous studies is that this paper argues that firms' different knowledge search structures have different impacts on firms' economic income and innovation output. Therefore, this paper divides firm performance into two aspects [16]. On the one hand, innovation performance refers to the increase of enterprise value brought by new products or services created by enterprises in knowledge and technology research and development(R&D) innovation, focusing on the innovation output of enterprises. On the other hand, financial performance refers to the economic benefits brought by the implementation and execution of the enterprise strategy, focusing on the economic income of the enterprise.

Research Assumptions
In different dimensions, enterprises will choose exploratory search and exploratory search according to their own resource characteristics, and different search methods are bound to have different impacts on the financial performance and innovation performance of enterprises. For example, in the dimension of time, some scholars believe that due to the complex and changing environment, old knowledge is difficult to meet the needs, while new knowledge can bring breakthrough inventions to enterprises, get rid of path dependence and realize technological leapfrog, so as to improve the innovation performance of enterprises and obtain greater economic benefits [17]. However, some scholars believe that old knowledge focuses on specific knowledge fields and has higher innovation value for enterprises, while new knowledge has lower reliability and higher utilization risk [18]. In terms of geographical dimension, the geographical proximity between enterprises can reduce the transportation time of goods and save the transaction costs of enterprises, but excessive development search will lead enterprises into the "risk of redundancy", thus hindering enterprise innovation [18]. Because of the spatial heterogeneity, the knowledge with a long geographical distance has regional characteristics and novelty. It can create a new combination with the existing knowledge base of the enterprise, which is conducive to the innovation of the enterprise. However, due to the geographical distance, a large number of transportation costs and transaction costs will be generated, thus increasing the financial burden of enterprises. Moreover, due to the differences in regional culture and system, knowledge is different, and enterprises are faced with greater uncertainty and higher degree of information asymmetry, which brings huge challenges to enterprises [19,20].On the technical dimension, similar products in the horizontal technology market can benefit from each other's innovation and R&D activities [21,22]. Search for similar technical knowledge that can be digested and absorbed at lower cost and higher efficiency; But excessive exploitation of similar technologies can lead to short-termism [23]. Technical knowledge with large differences is novel for enterprises, and the combination with existing technical knowledge can produce breakthrough technologies, but it requires enterprises to pay a large cost to digest and absorb.
Through the above analysis, it can be seen that exploratory search in different dimensions has different impacts on corporate financial performance. On the one hand, due to the low reliability of knowledge acquired by exploratory search, high utilization risk, high transaction cost and difficulty in digestion and absorption, exploratory search in all dimensions will cause adverse effects on corporate financial performance [19,20,23]. On the other hand, exploratory search can bring new products and technologies to enterprises, improve product quality and production efficiency of enterprises, and thus obtain huge economic benefits [21]. Based on this, this paper proposes the following research hypothesis: Hypothesis 1. Compared with full development search, the external knowledge search structure of exploratory search in different dimensions has different influences on corporate financial performance.
Similarly, exploratory search in different dimensions has different influences on enterprise innovation performance. On the one hand, the new knowledge obtained by exploratory search is of great novelty and differs greatly from the existing knowledge of enterprises, which is difficult to be fully digested and absorbed [6,7]; Therefore, exploratory search in all dimensions will have a negative impact on enterprise innovation performance. On the other hand, exploratory search can bring new knowledge with high novelty and heterogeneity to the enterprise, effectively expand the knowledge base of the organization, promote the recombination of different knowledge elements, produce breakthrough technology, realize technology leapfrog, and thus facilitate the technological innovation of the enterprise [8,11]. Based on this, this paper proposes the following research hypothesis: Hypothesis 2. Compared with the fully developed search, the external knowledge search structure of exploratory search in different dimensions has different influences on the innovation performance of enterprises.

Data Source and Sample Selection
This paper takes listed companies with inward patent licensing in China's manufacturing industry as the research object, and adopts the inward patent licensing data of listed manufacturing companies in China as the sample. The data in this paper include three levels: the patent level, the enterprise level and the regional level; The patent data are mainly from the State Intellectual Property Office and the China research data service platform(CNRDS); The data at the enterprise level are mainly from the database of China stock market & accounting research database(CSMAR) and the annual report of listed companies, including the basic information of listed companies. The data at the regional level are mainly from the official website of the National Bureau of Statistics.
In the process of sample selection, firstly, 452,750 citation data of authorized patents of 2,977 listed companies during 2001-2021 were obtained from the data of authorized patents of listed companies obtained from CNDRS after missing values were excluded. Secondly, take the average value according to the enterprise level and match the data of patent acquisition and financial index of listed companies. Finally, according to the manufacturing code "C" in "Industry Classification and Code of National Economy", the final data sample of this paper is obtained after industry screening, namely, 9,477 knowledge search data of 2,053 listed manufacturing companies in China from 2010 to 2019.

Measurement of Variables
The improvement of enterprise performance through external knowledge search is usually reflected in two aspects. One is the financial performance of the enterprise. In this paper, the return on assets (ROA) of the patent licensee in the next year after accepting the technology is used to measure the financial performance of the enterprise. The other is the innovation performance of the enterprise. This paper measures the innovation performance of the enterprise by adding 1 logarithm to the number of invention patents authorized by the patent licensee in the next year after receiving the technology [24].
For the division of external knowledge search structure of enterprises, this paper firstly uses the citation information of authorized patents to calculate three knowledge search dimensions: time distance, geographical distance and technical distance. Referring to the empirical research conducted by domestic and foreign scholars on external knowledge search, this paper adopts the time dimension distance of the difference measurement from the authorization year of the cited patent to the application year of the cited patent [24], and subtract the original value from the maximum value. The smaller the difference value is, the lower the novelty of the searched knowledge is and the easier it is to be digested and absorbed by the enterprise. The larger the difference value is, the higher the novelty is. The specific calculation method is shown as Formula 1. (1) In Formula 1, year li is the grant year of the reference patent i, and year a j is the filing year of the referenced patent j. Calculation of geographical dimension The spherical straight-line distance between the cities where the knowledge flow occurs between the two companies, the smaller the value is, the closer the geographical distance between the two firms and the simpler the knowledge search. The larger the value is, the farther the distance is. The specific calculation method is shown as Formula 2.
(2) In Formula 2, where R is the average radius of the earth, long A and the lat A are respectively the latitude and longitude of place A, long B and lat B respectively the latitude and longitude of place B.
To calculate the cognitive dimension, refer to Jaffe [22]. According to the classification number of the referenced patent and the referenced patent, the first three letters and digits (i.e., part and category) are selected as the technical index of the patent, and set as dummy variable. 0 indicates that the two patents belong to the same part and category, and 1 indicates that the two patents do not belong to the same part and category. In addition, considering that the same company has multiple patent citations, the average value of the three dimensions is taken at the company level. Secondly, the medians of the three knowledge search dimensions are calculated, and the three dimensions are divided into exploitation and exploration according to the medians respectively. The ones below or equal to the median are exploitation search, and those above the median are exploratory search. Then it can be divided into four categories and eight search structures according to the number of exploration dimensions. This paper includes control variables at both the enterprise level and the regional level. At the enterprise level, it includes six indicators, namely enterprise age, enterprise size, ownership type, redundant resources, R&D investment and enterprise profit. At the regional level, there are three indicators, which are regional economic growth, regional technology market scale and regional knowledge level. In order to reduce the influence of extreme values on the regression results, the continuous variables were reduced by 1% at both ends. In addition, due to the differences in enterprise performance in different regions and at different times, dummy variables of year and province are also added as controls.

Model Setting
Through variable design and theoretical analysis, this paper adopts Bayesian multiple regression analysis method to establish the model. The specific model is shown in Formula 3.
In Formula 3, Y it is enterprise performance; β 0 is the constant term, β j ( j = 1, 2...β j , 8) represents the regression coefficient; X it is the structure of external knowledge search; Control it is the control variables, including firm age, firm size, ownership type, redundant resources, R&D input, firm profit, regional economic growth, regional technology market size, regional knowledge level; ε is the error term.

Descriptive Statistics and Correlation Analysis
The descriptive statistics and correlations of key variables of the matched samples are shown in Table 1. The correlation coefficients do not show any major multicollinearity problems, although there are individual control variables with correlation coefficients greater than 0.5, such as 0.73 between firm R&D investment and firm size, but a review of the correlation shows that the mean variance inflation factor (VIF) is 1.58 and the maximum VIF value is 2.73. Below the critical threshold of 10, there is no multicollinearity problem [24].

Regression Results
In order to verify the influence of external knowledge search structure on enterprise performance, this paper takes the first knowledge search structure (namely, three dimensions of development search) as the benchmark and constructs eight models. Models 1-7 are the influence of seven knowledge search structures on enterprise performance respectively, Model 8 is the full model, and the explained variables are enterprise financial performance and innovation performance respectively. The regression results are shown in table 2 and table  3. Table 2 lists the empirical results of the influences of different external knowledge search structures on corporate financial performance. It can be seen that different knowledge search structures have different influences on corporate financial performance. Specifically, compared with the three dimensions of the exploratory search structure, in one dimension of   Note: *** means p < 0.01, ** means p < 0.05, * means p < 0.1; Other dummy variables are not included in this table. Note: *** means p < 0.01, ** means p < 0.05, and * means p < 0.1 exploratory search and the other two dimensions of exploratory search structure, the geographical dimension of exploratory search structure (β = 0.735, p < 0.05) and the technical dimension of exploratory search structure (β = 0.784, p < 0.05) have a significant positive impact on the improvement of corporate financial performance, the time dimension of the exploratory search structure on corporate financial performance although the coefficient is positive, but the impact is not significant (β = 0.317, p > 0.1);In the two dimensions of exploratory search and the third dimension of exploratory search, the geographical and technical dimension of exploratory search structure (β = 1.138, p < 0.01) and the time and geographical dimension of exploratory search structure (β = 1.46, p < 0.01) have a significant positive impact on corporate financial performance. The influence coefficient of time and technology exploratory search structure on corporate financial performance is positive but not significant (β = 0.179, p > 0.1).Exploratory search in the three dimensions of time, geography and technology also has a significant positive impact on corporate financial performance (β = 1.065, p < 0.01).In addition, the coefficient of redundant resources, R&D investment and profits is positive and significant, indicating that redundant resources, R&D investment and profits are conducive to the improvement of financial performance of enterprises, the coefficient of firm size is negative and significant, and the factors of firm age, ownership type and regional level have no significant influence on financial performance. In general, exploratory search is not beneficial to the improvement of corporate financial performance in any one or more dimensions, indicating that different external knowledge search structures have different influences on corporate financial performance. Hypothesis 1 is supported. Table 3 shows the empirical results of the influence of external knowledge search structures on firms' innovation performance. The results show that different knowledge search structures have different influences on firms' innovation performance. Specifically, compared with the three dimensions of exploratory search, only the exploratory search structure of geographical dimension (β = 0.274, p < 0.01) and the exploratory search structure of time and geographical dimension (β = 0.16, p < 0.01) have positive and significant influences on firm innovation performance. While the other five knowledge search structures have significantly negative impacts on firms' innovation performance. In addition, firm size, ownership type, R&D investment and profit have a significant positive impact on firm innovation performance, while the three indicators of firm age, redundant resources and regional level have no significant impact on firm innovation performance. In general, exploratory search on one or more dimensions is not conducive to the improvement of firms' innovation performance, indicating that firms' different external knowledge search structures have different influences on firms' innovation performance. Hypothesis 2 is supported.

Robustness Test
In order to prove the reliability and non-randomness of the research results in this paper, the following three methods are further selected to test the robustness of the above regression results (the table is omitted due to space limitation). First, this paper only considers the impact of one year lag after knowledge search on short-term firm performance. Therefore, in the robustness test, the two dependent variables lag three years to observe the impact of external knowledge search on long-term firm performance. The results show that, in the long run, the influence of external knowledge search structure on firm performance is different from that in the short run, but the difference does not change much.
Second, in existing studies, some scholars also use average to divide the external knowledge search structure of enterprises [7]. Therefore, this paper adopts the knowledge search structure divided by average to conduct robustness test. The results show that, in terms of corporate financial performance, the exploratory search structure of time dimension and the Note: *** means p < 0.01, ** means p < 0.05, and * means p < 0.1 exploratory search structure of time and technology dimension change from insignificant to significant, and the significance of other structures remains unchanged, indicating that exploratory search in any one or more dimensions is conducive to the improvement of corporate financial performance. In terms of enterprise innovation performance, the coefficients of exploratory search structure and full exploratory structure in time dimension are no longer significant, the coefficients of exploratory search structure in geography and technology dimension become positive and significant, and the significance of other structures remains unchanged. Third, the return on equity (ROE) is also used as the proxy index of corporate financial performance, and the number of patents authorized by the patent licensees in the next year after accepting the technology is used as the proxy index of corporate innovation performance. The significance of knowledge search structure is not changed much.

Conclusions and Discussion
Based on 2,053 listed companies with internal patent licenses in China's manufacturing industry from 2010 to 2019, this paper uses Bayesian multiple regression for empirical analysis to study the influence of different structures of external knowledge search on the improvement of enterprise performance. It is found that different external knowledge search structures have different impacts on corporate performance. Compared with the three-dimensional development search structure, exploratory search in any dimension is not conducive to the improvement of corporate financial performance and innovation performance.
According to the analysis results, in terms of financial performance, exploratory search with a single time dimension has no significant impact on the improvement of corporate financial performance. The reason may be that the knowledge obtained from the search is too novel, which belongs to the forefront of the research field and has a large distance from the existing knowledge of the enterprise. Therefore, the enterprise cannot make full use of the new knowledge, resulting in the waste of time and resources. The influence on the improvement of enterprise financial performance is not significant; In addition, due to the limited absorption capacity of enterprises, enterprises need to spend a lot of financial resources, material resources and manpower to absorb the knowledge obtained from knowledge search, which is relatively novel and greatly different from the original technology of enterprises, and there are certain risks, so the exploratory search structure of time and technology dimension has no significant influence on the improvement of corporate financial performance. In terms of innovation performance, only the exploratory search structure of geographical dimension and the exploratory search structure of time and geographical dimension have positive and significant effects, while the other five knowledge search structures are not conducive to the improvement of innovation performance. There may be two reasons for its emergence: first, the knowledge obtained from external knowledge search can be converted into capital income in a short period of time, while it is difficult to generate invention patents in a short period of time. In addition, no matter in the knowledge search structure of single dimension or multidimension, the great difference of the searched knowledge technology is detrimental to the improvement of the innovation performance of enterprises. The knowledge from different industries or fields is incompatible with the existing knowledge base of enterprises, so it is very challenging for enterprises to absorb cross-domain knowledge.
This article provides the following three contributions. Firstly, this paper studies the balanced search structure from the three dimensions of time, geography and technology, and enriches the research literature on multidimensional knowledge search balance. The existing research paradigm of external knowledge search usually abstractions the external technology of enterprises, or takes the view of "point", "line" and "surface". This paper considers the multidimensional space characteristics of knowledge search as a whole, focuses on the balance of multidimensional space structure, and enriches the research literature on multidimensional knowledge search balance. Secondly, the influence of existing researches on the multidimensional spatial structure is mainly focused on the innovation aspects of enterprises, such as breakthrough technology, innovation quality and enterprise continuous innovation [5][6][7]. This paper focuses on the differentiated effects of different multidimensional spatial structures on financial performance and innovation performance, and further expands the research field of enterprise performance and its influencing factors from the perspective of differentiation. Finally, existing research fields on enterprise external knowledge search are mainly concentrated in high-tech industries with high patent output, pharmaceutical industry, artificial intelligence, etc [5,6,8], As China is a manufacturing country, the research on the manufacturing industry can provide new ideas for the innovation and development of China's manufacturing industry and the improvement of its quality and efficiency.
This paper provides management enlightenment for the innovation development, quality and efficiency improvement of Chinese manufacturing enterprises. The results show that the exploratory search structure of single geographical dimension, time and geographical dimension is beneficial to the improvement of enterprise financial performance and innovation performance. In practice, Chinese manufacturing enterprises can search for knowledge in the field of remote technology, which can effectively balance the exploratory search and development search of external knowledge, help enterprises realize technology leapfrog, obtain greater economic benefits, and realize the purpose of innovation, development, quality and efficiency improvement.
There are still some shortcomings in this paper. First, based on the existing literature, this paper only divides the spatial heterogeneity of external knowledge search into three dimensions: time distance, geographical distance and technical distance, and studies its impact on enterprise performance, without considering other dimensions of external knowledge search (such as organizational dimension, depth and breadth of knowledge search, etc.). Therefore, future research can further explore external knowledge search and expand new fields of external knowledge search research from more spatial dimensions. Secondly, although the number of patents used in this paper to measure the innovation performance of enterprises can objectively reflect the technological innovation ability of enterprises, it is not comprehensive and accurate enough, because not all innovative achievements will be applied for patents. Future research should adopt more comprehensive and accurate indicators for measurement. Finally, due to the limitations of data, the sample selection in this paper is limited to listed companies in the manufacturing industry, which has a single industry. In the future, in-depth research can be conducted from other industries or multiple industries to improve the universality of the research conclusions.