Sensitivity Analysis of Parametric Production Frontiers to Multivariate Outliers : Application of Multivariate Outlier Detection using Stata to Japanese Water Utility Analysis

IR Web Site

Bibliographic Information

Other Title
  • 多変量外れ値に対する パラメトリックな生産フロンティアの感度分析　ーSTATA を用いた多変量外れ値検出の水道事業分析への適用ー
  • 水インフラ整備の課題と展望 多変量外れ値に対するパラメトリックな生産フロンティアの感度分析 : STATAを用いた多変量外れ値検出の水道事業分析への適用
  • ミズ インフラ セイビ ノ カダイ ト テンボウ タヘンリョウ ハズレ チ ニ タイスル パラメトリック ナ セイサン フロンティア ノ カンド ブンセキ : STATA オ モチイタ タヘンリョウ ハズレ チ ケンシュツ ノ スイドウ ジギョウ ブンセキ エ ノ テキヨウ

Search this article

Description

The purpose of this paper is to conduct a sensitivity analysis of production frontiers and technical efficiencies estimated by a one-step model in stochastic frontier analysis (SFA) against the removal of multivariate outliers. Specifically, this study applies three commands available in STATA, hadimvo・bacon・mcd, on the 12 variables used by Yane and Yane (2018), which applied Wang (2002)’s model to Japanese water utilities. The main findings are as follows : 1) although by default the number of outliers detected is the highest with mcd followed by hadimvo and bacon, it is consistent in that bacon’s outliers are those of hadimvo and hadimvo’s outliers are those of mcd; 2) removing the only two observations with extreme values among the 1,243 water utilities, customer density variable cusden, which is an environmental variable and was not statistically significant to technical inefficiency, becomes statistically significant at the 0.1％ level; 3) the results of sensitivity analysis against outliers using bacon and hadimvo on production frontier and technical efficiency including the forementioned environmental variable, are mostly robust; and 4) mcd, however, detects almost 30％ of the sample as outliers, and extremely contracts the variance of the one-sided error term (which indicates technical inefficiency). Hence, it cannot be used for this frontier analysis as a default

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top