台灣養豬戶經營效率及影響效率因素之研究：網絡資料包絡分析法的應用

Analysis of Technical Efficiency in Farrow-to-finish Swine Production: The Application of Multi-activity Network Data Envelopment Analysis

本文主要建立一個可以同時衡量毛豬生產之育種與肥育兩個階段生產效率之多部門網絡模型，相較於傳統之一階段DEA模型，此模型除可同時處理不同生產活動間之要素共用（shared inputs）與中間產出／投入（intermediates）問題，還具有可以考慮所有潛在差額（slack）的優點。藉由此模型的使用，我們將可以更務實與完整的角度來衡量一貫化豬場之生產效率。 實證分析係根據2005~2007年之計帳養豬戶樣本來進行。效率衡量結果主要包括以下幾點：首先，多部門網絡模型之整體無效率值雖未顯著異於傳統一階段DEA模型的結果，根據其結果卻指出豬場之無效率來源並不相同，大部分豬場之無效率是來自肥育生產階段，而較少部分來自育種生產階段。這些訊息是傳統模型無法提供但對豬場效率改善很有價值的訊息。其次，考慮非零差額與否對於樣本養豬戶之效率結果影響顯著，說 明DEA模型考慮非零差額的重要性。最後，迴歸結果發現不同生產階段之效率影響因素並不完全相同，顯示各別豬場應各自針對其無效率之生產階段來決定應改善的因素。此亦說明，具有多部門或多階段生產活動的廠商若要辨認影響效率大小之因子時，不同生產階段應該分開各自進行，而不應該混合在一起處理。

This study aims to propose a multi-activity network data development analysis (MNDEA) model to measure the technical efficiency of farrow-to-finish swine production in Taiwan. Production phases are explicitly divided into two activities, namely, breed-to-farrow phase and wean-to-finish phase. By using this model, the problem of shared inputs and dynamic intermediates among activities which characterize pig production are taken into account in an integrated framework simultaneously with the consideration of all the potential slacks, and allow us to examine aspects of production in a more comprehensive and factual manner.For the empirical results based on sample data from 2005~2007, it is shown that the overall technical inefficiencies obtained from MNDEA do not reveal obvious difference from those from a traditional one-stage model. However, the MNDEA results show us explicitly that the inefficiency sources for each farm are different. Furthermore, regression results also tell us that the determinants of efficiency for each production phase are not completely the same, and indicate the need to identify the influential factors for each production phase separately.