CAJ | 학술논문

数据包络分析(DEA)%三阶段组合效率测度模型%幅度调整测度(RAM)%随机前沿分析(SFA)%技术研发效率%中国大中型工业企业數據包絡分析(DEA)%三階段組閤效率測度模型%幅度調整測度(RAM)%隨機前沿分析(SFA)%技術研髮效率%中國大中型工業企業
수거포락분석(DEA)%삼계단조합효솔측도모형%폭도조정측도(RAM)%수궤전연분석(SFA)%기술연발효솔%중국대중형공업기업
data envelopment analysis(DEA)%hybrid three-stage model for efficiency measure%range-ad-justed measure(RAM)%stochastic frontier analysis(SFA)%technological research and develop-ment efficiency%China ’ s large and medium-sized industrial enterprises

在系统梳理现有可调整环境因素或／与统计噪音影响的三阶段组合效率测度模型的基础上，利用幅度调整测度（ RAM）与随机前沿分析（ SFA ）结合，构建了无需强制性调整环境因素与统计噪音影响的三阶段组合效率测度改进模型———RAM-SFA-RAM．相对现有模型，新的组合测度模型不但在估计影响与调整投入和产出时可实现较为客观地设定环境因素与松弛变量之间存在非线性生产函数关系，以降低效率估计偏误，而且充分利用了RAM模型全面测度无效的优点，为从投入和产出两个方面同时过滤环境因素与统计噪音的差异性影响提供了新的分析途径；更重要的是，选用的RAM具有平移不变性，可使组合测度模型克服现有模型只能通过最值强制性对投入和产出进行正向调整以适应选择的效率估计模型无法处理非正值的不足，避免了由此产生的效率估计偏误．实证研究部分用这一组合效率测度模型分析了中国省域层次上大中型工业企业的技术研发效率，为效率的客观比较提供了可行的模型方法．
재계통소리현유가조정배경인소혹／여통계조음영향적삼계단조합효솔측도모형적기출상，이용폭도조정측도（ RAM）여수궤전연분석（ SFA ）결합，구건료무수강제성조정배경인소여통계조음영향적삼계단조합효솔측도개진모형———RAM-SFA-RAM．상대현유모형，신적조합측도모형불단재고계영향여조정투입화산출시가실현교위객관지설정배경인소여송이변량지간존재비선성생산함수관계，이강저효솔고계편오，이차충분이용료RAM모형전면측도무효적우점，위종투입화산출량개방면동시과려배경인소여통계조음적차이성영향제공료신적분석도경；경중요적시，선용적RAM구유평이불변성，가사조합측도모형극복현유모형지능통과최치강제성대투입화산출진행정향조정이괄응선택적효솔고계모형무법처리비정치적불족，피면료유차산생적효솔고계편오．실증연구부분용저일조합효솔측도모형분석료중국성역층차상대중형공업기업적기술연발효솔，위효솔적객관비교제공료가행적모형방법．
This paper proposes a flexible and enhanced hybrid three-stage model for efficiency measure , RAM-SFA-RAM, without the mandatory adjustment of the effects of environment factors and statistical noise , which is based on the comprehensive discussions and review of extant hybrid three-stage models for efficiency measure with adjustments for the effects of contextual environment and statistical noise .In contrast to the exist-ing models, our hybrid model not only obeys the non-linear productive functional relationship between the en-vironmental factors and slacks in estimating slacks and in adjusting the inputs and outputs to reduce the bias of efficiency estimation , but also takes advantage of the RAM which can present a new approach of adjusting the different effects of environmental factors and statistic noise based on both inputs and outputs .More importantly , our model takes full advantage of the translation invariance of the RAM to deal with the non -posi-tive values, which can only work for existing models by arbitrary positive adjustments in virtue of the maximum or minimum of estimated slack datasets , and avoids the bias of efficiency estimation .In the empirical study , this article applies this new model to measure the research and development efficiency of large -and medium-sized industrial enterprises at China ’ s provincial level , to show that our method presents a feasible modelling approach of objectively comparing efficiency scores .