以Double-LASSO評估政府採購案中招標方式及決標原則對標比的影響

Abstract

本研究分析臺北市政府轄下某機關3,276筆公開決標資訊，利用多元迴歸分析法估計處置變數「招標方式」及「決標原則」對被解釋變數「標比」的影響程度。為了避免遺漏變數，透過計量方法Double-LASSO選擇合適的控制變數。\n研究結果顯示，在無控制變數，單純以「招標方式」及「決標原則」作為變數分析時，最有利標決標原則及限制性招標對「標比」的影響係顯著的。當改以Double-LASSO挑選控制變數時，從74個變數中挑出40個遺漏變數，此時最有利標決標原則仍然呈現顯著，但招標方式全部呈現不顯著。簡言之「招標方式」及「決標原則」對「標比」的影響會隨著控制變數之增加而越來越不顯著，本研究中唯一持續呈現顯著且負相關的變數是「投標家數」。

This study analyzes 3,276 public bidding information of an institution under the Taipei City Government, using multiple regression analysis to estimate the degree of impact of "tendering method" and " bid decision principle " on "bid-ratio". In order to avoid omission of variables, the study selects appropriate control variables by the measurement method Double-LASSO.\nThe results of the study show that when there are no control variables and only "tendering method" and "bid decision principle" are used to explain the variable analysis, the impact of best value principle and the revised bidding on "bid-ratio" are significant respectively. When using Double-LASSO to select control variables, it picks out 40 missing variables from 74 variables. At this time, the impact of best value principle is still significant, but all tendering methods are not significant. In short, the impact of "tendering method" and "bid decision principle " on "bid-ratio" will become less and less significant with the increase of control variables. The only variable in this study that continues to show significant and negative correlation is "number of bidders".