本文旨在衡量2011年APEC及EU計31個成員國稅務機關在技術效率與稽徵技術水準之差異,應用三階段DEA作為評估方法,摒除環境變數之影響,合理評估該31國稅務機關之技術效率差異,期滿足稅務機關多樣產出、政策方針差異及經營環境不同之特性;另考量稅務機關追求稅收極大化及稽徵成本最小化之目標,須要同時考量縮減投入與擴張產出,以及各國稅務機關具不同稽徵技術水準之差異,乃採用方向性距離函數結合O’Donnell et al.(2008)以共同邊界為概念的資料包絡分析,進行不同國家之稽徵績效與技術差距之比較,以揭露各國在資源運用及稽徵技術水準之差異,進而提供稅務機關相互學習之標竿。實證結果顯示,經三階段DEA調整後,整體樣本國家共同技術效率及技術缺口率均有明顯增加。而共同邊界模型亦解釋稅務機關受資源限制及政府政策等外在因素,存在稽徵技術差距之現象。
The subject of this article is to examine the difference of the technical efficiency and tax collection technology among the tax agencies of 31 member countries of APEC and EU in 2011. The three-stage DEA is applied as the evaluation method. The technical difference among the tax agencies of 31 member countries would be evaluated reasonably without considering the influence of environmental variation. It is expected to comply with the multiple output of tax agency, the difference of policy guideline and the different characteristic of operating environment. In addition, it is required to consider to reduce the investing and expanding output and the of difference production technology of each tax agency as the tax agency is seeking the maximum taxation and minimum tax collection cost. Therefore, the directional distance function is adopted to be combined with O’Donnell et al.(2008)and further take the metafrontier as the concept for data envelopment analysis to execute the comparison of the tax collection efficiency and technical gap among different countries so as to expose the difference of the resource utilization and technical standard of tax collection of each country. This will further provide the learning benchmark of each other for the tax agencies. According to the experimental result, the meta technical efficiency and technology gap ratio of the entire sample country are increased significantly after the adjustment of three-stage DEA. Meanwhile, the metafrontier model can even be explained that the technical gap of tax collection does exist when the tax agency is restricted by the resource constraint, government policy and other external factors.