【作者】 李娜；

【导师】 李北伟；

【作者基本信息】 吉林大学 ， 工商管理， 2024， 博士

【摘要】 目前,全球正经历第四次工业变革的浪潮,中央经济工作会议明确指出:“要以科技创新推动产业创新,特别是以颠覆性技术和前沿技术催生新产业、新模式、新动能,发展新质生产力”^[1]。作为我国产业结构转型的重要着力点,高技术产业已经成为我国经济和社会发展的主要推动力^[2]。2011年,我国规模以上高技术制造业总产值已达到9.2万亿元,产业规模居世界第二位^[3],但区域间却存在显著差异。怎样准确地评估我国高技术产业的技术创新效率、找出不同地域间的差异特征,阐明其背后的影响机制并给出合理化建议?针对这些问题,学术界已经开展了大量研究并取得了一定成果,但仍存在一些限制和不足之处:以往文献将技术创新过程视作未知的“黑箱”,忽视了创新系统内部的运作机制和流程,缺乏对技术创新过程不同阶段的划分测评;在创建技术创新效率评价体系时,没有考虑到环境污染排放等非期望产出问题;在分析技术创新效率的差异时,多数研究只聚焦时间维度上的差异,较少考虑空间维度,或对空间维度的分析不足,只考虑空间上的差异性,忽略了空间的关联性和动态演变趋势;在研究技术创新效率差异产生的原因时,主要关注于少数因素对高技术产业的影响分析,缺乏对高技术产业技术创新效率影响因素全面的归纳和分析;在进行影响因素的机制分析时,得到的作用路径往往较为单一,对多重影响因素下复杂的作用路径研究不够深入,且主要靠专家打分等主观方法进行权重赋值,缺少客观的研究方法等。鉴于此,本文基于理论分析和文献梳理,在概念界定的基础上,构建了高技术产业技术创新效率测度体系,基于2000—2020年我国30个省份面板数据,测度了高技术产业技术创新整体及各子阶段的效率,然后利用多种实证分析方法,对高技术产业技术创新效率的时空分异特征和收敛性进行了分析,利用BP-DEMATEL模型识别了高技术产业技术创新效率的关键影响因素,并在此基础上利用解释结构模型,对作用机制进行了理论分析与实证检验,最后基于全文的研究结论,制定了有针对性的改进建议。本研究的创新点可以概括地归纳为以下几个方面:（1）确立了高技术产业技术创新效率测算的新范式,将非期望产出纳入高技术产业技术创新效率测度评价体系中。通过文献回顾发现,国内外高技术产业技术创新效率的研究中往往缺乏对环境约束的考虑,也忽视了投入要素的“松弛”和“拥挤”问题。基于这种情况,本文参考绿色创新理论,把能耗与生态影响融入常规的效率计量模型,同时,鉴于研发不成功、成果转化不足等非有效创新要素,构建了包括非期望产出（能源使用、环境影响及无效创新）的Super-SBM模型,不仅对测算方法的选取依据进行了较为详细的理论论证,而且基于创新价值链理论,将技术创新细分为不同阶段（技术研发、产品开发、市场转化）来分析和评估效率,并分阶段讨论了高技术产业技术创新效率的差异化特征,更能贴近高技术产业技术创新的真实情况。（2）跨学科借鉴了地理信息学中的探索性空间数据分析技术,通过引入Arc GIS和GEODA等空间处理工具,不仅突破了只依赖时间序列分析的局限,还在揭示空间差异的同时,将空间异质性图像化,并进一步引入了空间关联视角,通过运用Moran散点图、LISA聚集图、空间热点格局图,对技术创新效率的空间差异特征、空间关联特征和演化特征进行了多维印证、逐层递进的解读和分析,更直观地展现高技术产业技术创新效率的时空差异。这种跨学科的研究方法契合了现代及今后的科研范式革新,是一次横跨不同领域和学科的创新性探索。（3）对传统DEMATEL方法进行改进,引入BP神经网络构建了BP-DEMATEL模型。通过将BP神经网络融入到传统的DEMATEL方法中,不仅解决了以往研究只能对少量影响因素进行分析的片面性问题,实现了对大量影响因子的全面评估,还通过融合神经网络的客观数据处理特性,大幅降低了影响因素归纳和权重赋值环节的主观偏差。现有研究通常聚焦于影响技术创新效率的单一或几个变量,缺乏对影响因素的全面探索,并且常常依赖于问卷和专家评价等主观数据源,这可能会导致其在处理复杂影响因素时可信度降低。本文通过改进的BP-DEMATEL模型,能够在保证客观的前提下,全面系统地识别出高技术产业技术创新效率的影响因素,并根据原因度和中心度,分别找出高技术产业技术创新整体及各子阶段的关键影响因素,增强了结果的全面性与客观性。（4）利用系统工程学ISM模型,揭示了具有多元影响因素、多重传导路径的复合层次作用机制。在探索高技术产业技术创新效率的作用机制时,不再依赖主观分析、局限于单一作用路径,而是通过引入系统工程学ISM模型,揭示了一个具有多元影响因素、多重传导路径的复合层次作用机制,成功识别出10条作用途径。目前对于作用机制的研究通常只聚焦几种影响因素的单一作用路径,主要依赖于主观构建作用机制,然后用回归分析等传统方法进行验证。但在许多情况下,社会经济现象是由多元影响因素通过复杂的机制共同作用产生的,传统模型往往无法充分捕捉这些复杂的互动关系。本文通过引入系统工程学中的ISM模型,厘清了复杂影响因素之间的层级结构,明确了每个因素在不同作用路径中的作用和影响度,成功揭示了10条作用路径,为制定更有效的策略和政策提供了新的视角和深度。更多还原

【Abstract】 At present,the world is experiencing the fourth wave of industrial change,and the Central Economic Work Conference clearly pointed out: "We must promote industrial innovation with scientific and technological innovation,especially with disruptive technologies and cutting-edge technologies to generate new industries,new models,new momentum,and develop new quality productivity".As an important focus of China’s industrial structure transformation,high-tech industry has become the main driving force of China’s economic and social development.In 2011,the total output value of China’s high-tech manufacturing industry above designated size has reached 9.2 trillion yuan,ranking second in the world,but there are significant differences between regions.How to accurately evaluate the technological innovation efficiency of China’s high-tech industry,find out the differences between different regions,explain the influencing mechanism behind it and give reasonable suggestions?To solve these problems,the academic community has carried out a lot of research and made some achievements,but there are still some limitations and shortcomings:previous literature regards the technological innovation process as an unknown "black box",ignores the internal operating mechanism and process of the innovation system,and lacks the classification and evaluation of different stages of the technological innovation process;When establishing the evaluation system of technological innovation efficiency,the problem of undesirable output such as environmental pollution emission was not taken into account.When analyzing the regional differences of technological innovation efficiency,most studies mainly focus on the differences in time dimension,and rarely consider the difference structure in space dimension,or the comprehensive analysis of spatial differences is insufficient,and only consider the spatial differences,ignoring the correlation characteristics and dynamic evolution trend of space.When studying the causes of the difference of technological innovation efficiency,the author mainly focuses on the influence analysis of a few factors on the high-tech industry,and lacks the comprehensive induction and analysis of the factors affecting the technological innovation efficiency of the high-tech industry.In the mechanism analysis of influencing factors,the obtained action path is often relatively simple,and the research on complex action paths under multiple influencing factors is not deep enough,and the weight assignment mainly relies on subjective methods such as expert scoring,and there is a lack of objective research methods.In view of this,based on theoretical analysis and literature review,this paper constructs a measurement system of technological innovation efficiency of high-tech industries on the basis of concept definition.Based on the panel data of 30 provinces in China from 2000 to 2020,the Super-SBM model containing non-expected output is used to measure the overall efficiency of technological innovation of high-tech industries and the efficiency of each sub-stage.Then,the paper analyzes the spatio-temporal differentiation and convergence of technological innovation efficiency in high-tech industries.Finally,BP-DEMATEL model is used to identify the key influencing factors of technological innovation efficiency in high-tech industries.On this basis,ISM structural interpretation model is used to analyze the mechanism theoretically and empirically.The innovations of this study can be summarized as follows:A new paradigm for the measurement of technological innovation efficiency in high-tech industries has been established.The non-expected output is included in the evaluation system of technological innovation efficiency in high-tech industries.Through literature review,it is found that the study of technological innovation efficiency in high-tech industries often lacks the consideration of environmental constraints,that is,the problem of non-expected output.In addition,the problems of "slack" and "crowding" of input factors are usually ignored in the research.Different from previous studies,this paper refers to green innovation theory and integrates energy consumption and ecological impact into the conventional efficiency measurement model.At the same time,in view of the non-effective innovation factors such as unsuccessful research and development and insufficient transformation of results,a Super-SBM model including non-expected outputs(energy use,environmental impact and innovation inefficiency)is constructed.The selection basis of the measurement method is not only theoretically demonstrated in detail,but also based on the innovation value chain theory.The technological innovation is divided into different stages(technology research and development,product development,market transformation)to analyze and evaluate the efficiency,and the differentiated characteristics of technological innovation efficiency in high-tech industry are discussed in stages,which can be closer to the real situation of technological innovation in high-tech industry.Interdisciplinary reference is made to exploratory spatial data analysis technology in geoinformatics.By introducing spatial processing tools such as Arc GIS and GEODA,it not only breaks through the limitation of relying on time series analysis,but also reveals spatial differences,maps spatial heterogeneity,and introduces spatial correlation perspectives.By using Moran scatter plot,LISA aggregation plot and spatial hotspot pattern map,this paper verifies the spatial difference characteristics,spatial correlation characteristics and evolutionary characteristics of technological innovation efficiency in multiple dimensions,and interprets and analyzes it layer by layer,so as to more intuitively show the regional differences in technological innovation efficiency of high-tech industries.This interdisciplinary research method caters to the modern and future innovation of scientific research paradigm,and is an innovative exploration across different fields and disciplines.The traditional DEMATEL method is improved.By introducing BP neural network,the improved BP-De MATel model is proposed,which not only solves the one-sidedness problem that previous studies only analyze a few influencing factors,but also realizes the comprehensive evaluation of a large number of influencing factors.Moreover,by integrating the objective data processing characteristics of neural network,The subjective bias of factor induction and weight assignment is overcome.Existing studies usually focus on single or multiple specific variables affecting the innovation efficiency of high-tech industries,lack a comprehensive and systematic exploration of all key factors,and often rely on subjective data sources such as questionnaires and expert evaluations,which may lead to lower credibility in dealing with complex influencing factors.This paper builds an improved model of BP-DEMATEL,which can comprehensively and systematically identify the influencing factors of technological innovation efficiency in high-tech industries under the premise of ensuring objectivity,and identify the key influencing factors of technological innovation in high-tech industries as a whole and each sub-stage according to the cause degree and the centrality degree,thus enhancing the objectivity and accuracy of the results.When exploring the action mechanism of technological innovation efficiency in high-tech industries,it no longer relies on subjective analysis and is limited to a single action path,but introduces the ISM model of system engineering to reveal a composite hierarchical action mechanism with multiple influencing factors and multiple conduction paths,and successfully identifies 10 action paths.The current research on the mechanism of action usually only studies the single mechanism path of several influencing factors,and mainly relies on subjective construction of the mechanism of action,and then verifies it with traditional linear analysis methods such as single or multiple regression analysis.But in many cases,socioeconomic phenomena are generated by multiple factors interacting in complex ways,and linear models often fail to adequately capture these complex interactions.By introducing the ISM interpretation structure model in system engineering,this paper clarifies the hierarchical structure among complex influencing factors,clarifies the role and influence degree of each factor in different action paths,and successfully reveals multiple action paths.Through the hierarchical analysis framework built by ISM model,it not only highlights the key driving factors and their impact at the level,but also reveals the complex causal relationship and interaction among factors,so as to more accurately analyze how different factors work together on the efficiency of technological innovation in high-tech industries,and provide a new perspective and depth for formulating more effective strategies and policies.更多还原