【作者】 任娟；

【导师】 陈圻；

【作者基本信息】 南京航空航天大学 ， 管理科学与工程， 2012， 博士

【摘要】 竞争战略识别是战略管理领域的重要问题之一。竞争战略实证研究存在两个基本问题，一是验证竞争战略的存在性，二是探讨竞争战略的绩效机制和影响机制。显然，这两个基本问题都必须以竞争战略识别为前提和核心。但是，竞争战略概念的模糊性和缺乏可操作性，战略维度、变量选择的主观性和数据可得性限制，识别方法选择传统聚类方法所具有的缺陷，使竞争战略识别成为一个实证难题，严重制约着竞争战略的理论发展与实践应用。本文针对竞争战略识别方法研究中存在的问题，在认真收集、挖掘、整理相关研究文献基础上，提出了两种竞争战略识别路径，构建了竞争战略结构维度研究框架，提出了过程视角和动态视角的竞争战略识别问题，通过引入数据包络分析和多元统计分析的最新理论与方法，分别提出了基于DEA和基于面板数据聚类分析的竞争战略识别方法，实证研究了我国制造企业的竞争战略识别。本文将外部特征视角研究拓展到内部过程视角研究，将静态视角研究拓展到动态视角研究，不仅为竞争战略识别提供了新的实现手段，而且为竞争战略理论在企业管理中的操作化与实践化提供了一种有效的解决途径。本文的主要内容包括以下几个方面：第一，本文构建了竞争战略结构维度研究框架。现有的竞争战略识别研究均缺乏研究框架的构建。而本文通过对国内外竞争战略识别文献的深入挖掘，总结梳理了竞争战略识别的理论基础，提出了类型学和分类学两种识别路径；在妥善处理单维度与多维度、量化实证与概念分解、静态描述与动态分析之间矛盾的基础上，构建了竞争战略结构维度研究框架，给出了整合研究路径及其实现途径；分析了不同研究视角的竞争战略识别的逻辑结构以及识别方法的内部逻辑结构，构建了竞争战略识别框架。最终，为竞争战略识别的方法创新和实证研究提供了结构化和概念操作化的解决方案。第二，本文构建了交叉效率DEA模型并应用于过程视角竞争战略识别。现有的竞争战略识别均存在从外部战略特征识别的局限性和主观性缺陷。而本文首次运用DEA理论将外部特征视角研究拓展到内部过程视角研究，提出了过程视角的竞争战略识别问题；充分提取了DMU之间相似性和相异性信息，通过引入区间数，构建了多指标区间交叉效率DEA模型，提出了一种基于投入、产出权重的聚类方法并应用于竞争战略识别；通过引入多人非合作博弈，构建了博弈效率DEA模型，设计了以虚拟最差DMU效率为初始值的逼近搜索算法，证明了算法具有收敛性，求解出了Nash均衡解；构建了两阶段博弈交叉效率DEA模型，实现了自互评体系下的各DMU间的博弈，改进和优化了博弈交叉效率；在融合内外部评价结果的基础上，提出了一种基于博弈效率权重比和博弈交叉效率矩阵的聚类方法并应用于竞争战略识别；实证研究了我国机械制造业上市公司的竞争战略，识别出了低成本、差异化、低成本差异化竞争战略类型。实证结果表明，前一种方法能够区分有效决策单元，综合评价具有统一性和合理性；后一种方法能够实现自互评体系下的Nash均衡求解，效率评价的区分度更高；这两种识别方法与传统战略识别方法相比，分类效果更好。最终，解决了权重多重解和各DMU间竞争的问题，充分考虑了资源配置和运营范围两大类战略投入、产出维度以及战略指标的相对重要性和因果关系的特点，使其更具解释力和客观性。第三，本文提出了多指标面板数据的聚类方法并应用于动态视角竞争战略识别。现有的基于截面数据聚类方法的静态视角竞争战略识别均缺乏企业时间序列维度信息的提取。而本文首次运用面板数据多元统计分析理论将静态视角研究拓展到动态视角研究，提出了动态视角的竞争战略识别问题；描述了面板数据的时间、空间、指标三维数据格式和数字特征，提出了面板数据的因子分析，改进了多指标面板数据的系统聚类方法并应用于竞争战略识别；结合了时间序列的局部变化特性与整体距离关系，综合提取了面板数据的水平指标、增量指标和增量变化率指标的时间序列特征，提出了一种基于时序特征的多指标面板数据的聚类方法并应用于竞争战略识别，解决了非同步问题；提出了自适应滑动窗口分段方法，实现了时间序列局部变化的形状特征提取，提出了一种基于形状特征的多指标面板数据的聚类方法并应用于竞争战略识别；实证研究了我国家电企业的动态视角竞争战略，识别出了低成本、差异化、低成本差异化、“夹在中间”四种竞争战略类型。实证结果表明，第一种方法能够满足系统分析的统一性要求，保证指标之间的不相关；能够克服时间维度上均值化处理造成的偏误，信息损失较少。后二种方法能够综合提取空间、时序信息，能够有效降低噪声影响，分类效果较好。最终，解决了指标相关性和数据突变引起的识别效果失真问题，提高了竞争战略的识别效果。第四，本文提出了多指标面板数据的融合聚类方法并应用于竞争战略转折点识别。现有的基于战略稳定期的竞争战略识别均缺乏企业竞争战略演化信息的提取。而本文首次将面板数据的聚类方法引入竞争战略的演化分析中，提出了竞争战略的演化分析问题，利用Frobenius范数构造了离差平方和函数，提出了多指标面板数据的有序聚类方法，通过引入融合思想，提出了多指标面板数据的融合聚类方法并应用于战略转折点识别，实证研究了我国家电业上市公司的竞争战略演化，识别出了竞争战略类型和竞争战略演化的时间点。实证结果表明，新方法能够保证指标之间的不相关；能够克服时间维度上均值化处理造成的偏误，减少了信息损失；能够解决面板数据有序聚类的问题；弥补了单一分析的片面性和局限性。最终，解决了基于战略稳定期的静态识别对多期数据均值化处理导致的演化信息缺失问题，提高了战略转折点的识别效果。更多还原

【Abstract】 Competitive strategy identification is an important issue in the competitive strategy research.However, because of the ambiguity of the concept and its lack of operability, subjectivity of strategicdimension and strategic variable selection, availability of the data, as well as the flaw of statisticalclustering methods, there exist difficult empirical problems which seriously limit the development ofthe competitive strategy theory.In accordance with the existing problems in competitive strategy identification research, we putforward two identification paths, construction of the competitive strategy’s structural dimensionframework and competitive strategy recognition from process and dynamic perspective by collectingand refining the relevant research literature. By introducing the latest theories and methods in themultivariate statistical analysis and data envelopment analysis, we put forward new competitivestrategy identification methods based on panel data clustering and DEA theory respectively. Theempirical research is made on competitive strategy identification of China’s manufacturing enterprises.To expand the research perspective from external characteristics to internal process and static view todynamic view provides not only new means for competitive strategy identification but also enterprisemanagement operation and practice for competitive strategy theory. The main contents of the thesisinclude the following aspects:Firstly, the structure dimension research framework of competition strategy is constructed. Theexisting competitive strategy identification literature is lack of research framework. While this thesisput forward two identification paths of typology and taxonomy through deep excavation of thedomestic and foreign related literature. Based on proper treatment of the contradiction between singledimension and multi dimension, quantitative analysis and concept decomposition, as well as staticdescription and dynamic analysis in study of competitive strategy identification, we construct thecompetitive strategy structure framework and give the integration path and the way of its realization.We construct identification framework of competitive strategy based on analysis of the logicalstructure from different perspectives and internal logic structure among various identification methods.It provides a structured and concept operation solution for method innovation and empirical researchfinally.Secondly, a cross efficiency DEA model is constructed and applied in competitive strategyidentification from process perspective. The existing study is limited to external strategic characteristics identification and subjective defect. While this thesis first use DEA theory to expandthe perspective from external characteristics to internal process and put forward identificationproblem from process perspective. Based on the DMUS’ similarity and dissimilarity information ofinput and output weights, we put forward an input-output weights clustering method of competitivestrategy identification by introducing interval number and constructing the multivariable intervalcross efficiency DEA model. By introducing several DMUS’ non cooperative game, we construct thegame efficiency DEA model and design the virtual worst DMU efficiency as the initial values for anapproximate search algorithm. The algorithm is presented to be with convergence and Nashequilibrium solution is obtained. We construct two-stage DEA game cross-efficiency model. Thegame between each DMU is realized in the self-evaluation and peer evaluation system and game crossefficiency is improved and optimized. By fusing inner and outer evaluation results, a new clusteringmethod based on game efficiency weight ratio and game cross-efficiency matrix is proposed to beapplied to the empirical study for competitive strategy identification of China’s listed machinerymanufacturing companies. The empirical results show that the former method can distinguish efficientDMUS with Unity and rationality in comprehensive evaluation, and the latter method can realize theNash equilibrium solution in the self-evaluation and peer evaluation mode with higher distinguishdegree. The two methods are of better explanatory power and objectivity compared with traditionalclustering methods. The problem of the multiplicity of DMUS’ weights and competition amongDMUS is resolved. We fully consider the resource allocation and operation range as the strategic inputand output dimensions and present the relative importance of index and their causal relations.Thirdly, a multivariate panel data clustering method is put forward and applied in competitivestrategy identification from dynamic perspective. The perspective of existing study based on crosssection data clustering is static that is lack of information extraction in time series dimension. Whilethis thesis first use panel data multivariate statistics theory to expand the perspective from static todynamic and put forward identification problem from dynamic perspective. We describe threedimensions and format of panel data in time, space and index. A factor analysis of panel data is putforward and the ward clustering method for multivariable panel data is improved and applied incompetitive strategy identification. By combining the panel data’s local changes characteristics in timeseries dimension with global distance, we comprehensively extract the panel data’s time seriescharacteristics in “positional index”,“incremental index” and “incremental rate index”, then putforward a multivariate panel data series clustering method to identify competitive strategy and avoidthe problem of non synchronization. We propose an adaptive sliding window segmentation methodand realize the shape feature extraction of local changes of time series. Then a multivariate panel data clustering method based on shape characteristics is put forward and applied in competitive strategyidentification. The empirical study for dynamic competitive strategy identification of China’shousehold electrical appliance enterprises show that there exist four kinds of competitive strategiesincluding cost-leadership strategy, differentiation strategy, hybrid strategies and stuck in the middle.The empirical results show that the first method can meet the requirement of systematic uniformityand ensure the indicators are not related. It is able to overcome the errors caused by mean of timeseries dimension and less information loss. The second and third method can comprehensively extracttime and space information which effectively reduce the influence of noise and get betterclassification results. The distortion of identification results caused by correlation of indices and datamutation is solved and the result accuracy is improved.Fourth, a multivariate panel data fusion clustering method is put forward and applied incompetitive strategy turning point identification. The existing relative study based on strategic stabletime period (SSTP) is lack of information extraction of strategic evolution. The thesis first introducedpanel data clustering method into studying competitive strategy evolution and put forwardevolutionary problems. We put forward a multivariate panel data ordinal clustering method by using Fnorm to construct square deviation function. A multivariate panel data fusion clustering method basedon fusion theory is proposed to identify strategic turning point. Empirical study of China’s householdelectrical appliance enterprises is made to identify the strategic types and turning point of strategicevolution. The empirical results show that the new method can ensure no correlation of indices,overcome the errors caused by mean of time series dimension and less information loss. It can solvethe panel ordered data clustering problem and make up the one-sidedness and limitation of singleanalysis. The problem of evolutionary information loss caused by mean of SSTP is solved and theresult accuracy of strategic turning point is improved.更多还原