An analysis of impact factor of R&D and profitability efficiency for global semiconductor assembly and testing industry

近年半導體封裝產業發生多起併購案件，令人不禁好奇該產業發生之變化及對未來之影響。本研究針對全球半導體封測產業市場佔有率前十名之廠商，以包絡分析法進行CCR模式及BCC模式之研發效率及獲利效率。第一階段研究2012~2014年目標廠商之研發資源運用效率，投入項為研發費用及知識存量(加權累積專利數)，產出項為專利權數量；第二階段研究2014~2016年目標廠商之獲利效率，投入項為專利資本及銷售與管理費用，產出項為營業毛利。分析完個別之研發效率及獲利效率後，再將兩者以Pearson相關性分析檢測其相關係數，證實研發效率與獲利效率僅有低度相關。此外為探討日月光及矽品兩企業併購案，本研究將該兩家廠商之投入產出項合併進行另一次DEA包絡分析。研究結果發現整體產業呈現有效率狀態，以平均整體效率而言，研發效率及獲利效率皆達0.7。另外BCC模式下之技術效率表現更佳，惟規模效率部分表現較差，研究發現中度規模廠商(市佔率排名6至10之廠商)多處於規模報酬遞增的狀態，而規模程度高之廠商(如日月光、Amkor)則為規模報酬遞減之狀態。因此研究推論大者恆大之現象確實存在，然規模太大之廠商可能反而出現資源分配不均、整合程度不足等因素所造成的規模無效率。再以Tobit迴歸分析尋找影響效率因子，證實規模對於無論研發或是獲利效率均無顯著影響，研發效率的顯著因子有申請專利變動幅度，說明企業應有效規畫管理研發產出；獲利效率顯著因子包含直接生產力與銷管能力，證實成本控制為該產業獲利效率的重點因素。最後，本研究希望根據實證結果分析給予產業界及學界之未來研究建議。Recently, many merger and acquisition cases have occurred between the enterprises of the semiconductor assembly and testing industry. To imply about how it happened and how will it influence the whole industry,this thesis present a study which utilizes the input oriented CCR model and BCC model of two-stage Data Envelopment Analysis (DEA), which separates R&D efficiency and profitability efficiency to evaluate the performance for the global semiconductor assembly and testing industry. The research targets are enterprises which included in top-10 market share in the industry. Research period of the first stage and second stage respectively is during 2012 to 2014 and 2014 to 2016. In order to deal with the result of DEA, this study takes the Pearson correlation analysis to find relationships between the R&D efficiency and the profitability efficiency. In the R&D efficiency, it take R&D expenses and accumulated patents number as input; while take the number of new patents as output. In the profitability efficiency, it take patent capital and operating expenses as input; while take the operating margin as output. Through the BCC mode DEA, this study finds that most enterprises are efficiently in the technical efficiency part. However, the scale efficiency are worse. Those medium scale enterprises are in the” increasing returns to scale” mode, while the high scale enterprises are in the” decreasing returns to scale” mode. It confirmed that the trend of “The powerful are always powerful”, while those high scale enterprises might suffered in the problem of lake of integration. Finally, we hope the results can provide actual suggestions for both the industry and future research