公共運輸使用率之關鍵影響因素分析

Abstract

旅運者為克服環境阻抗，會選擇不同運輸方式來達成旅次目的，近年來，臺灣私人運輸的成長，造成了交通壅塞和環境污染等問題，為達永續運輸之目的，促使各地方政府不斷地發展公共運輸，以減少私人運輸的使用，或提升當地經濟發展。然而，策劃有效之提升公共運輸使用率策略，仍需仰賴對於影響公共運輸使用之重要因素的了解。 基於上述原因，以往為找出影響公共運輸使用率之因素，因受到區域特徵屬性限制，研究常局限於較小範圍，故本研究將以臺灣348個鄉鎮市區為研究樣本，分別建構全域型（Global）與區域型（Local）兩模式，其中全域型模式因估計出參數值，於各地區屬一均質狀態，故稱之，採用模式為迴歸分析與總體羅吉特分析，而區域型模式可估計出因地而異之參數值，採用模式為地理加權迴歸（GWR）。將臺灣348個鄉鎮市區之現有能取得資料，分別以社會經濟、公共運輸供給和私人運輸相關變數，來進行模式推估，其中區域型地理加權迴歸為與全域型迴歸分析參數做比較，採用與迴歸分析所推估出相同之顯著變數。最後，透過修正後MAPE來比較各模式之績效，找出最佳模式後，提出公共運輸使用率之關鍵影響因素，並給予相關改善策略。 模式推估結果顯示，因地理加權迴歸可估出因地而異之參數值，其修正後 MAPE相較於全域型模式，為準確度最高之模式。而該模式結果表示公共運輸使用率與未成年人口比例於花蓮地區呈正相關，於臺中以北地區呈負相關；與低收入戶於桃園縣、新竹縣和宜蘭縣等地呈正相關，於南投縣、花蓮縣和苗栗縣等地呈負相關；與公路客運平均班次數於新竹以北地區呈正相關；與市區公車路線數於南投以北地區呈正相關；與市區公車平均車齡於桃園以北地區呈負相關；與機車持有率於新竹縣、屏東縣等地呈正相關，於東北部呈負相關；與道路長度於新竹以北、中央山脈中南部分地區呈負相關。基於以上各地區之關鍵影響因素，提出相關對策之建議。

With the rapid growth in private vehicles in Taiwan in recent years, traffic congestion and air pollution become a major challenge for urban development. Towards transportation sustainability and continuing economic development, to effectively reduce travelers’ dependency in using private vehicles and increase their preference to public transportation is obviously the only solution. According to recent official national travel survey in Taiwan, usage rate of public transportation remarkably differs across Taiwan. Without knowing the key factors affecting traveler’s mode decision, it is impossible to propose effective public transportation improvement strategies. Based on this, this study employs global and local regression models to recognize the key determinants for the usage rates of a total of 348 areas (township or districts) in Taiwan. The global regression model (i.e. Ordinary Least Square, OLS; Aggregate Logit Model, ALM) is to estimate one set of parameters associated with explanatory variables to explain the differences in the usage rates of all study districts, while the local regression model (i.e. Geographically Weighted Regression, GWR) allows the estimated parameters differ depending upon the correlation among neighboring areas. By referring related studies, a total of 29 potential explanatory variables are selected, which can be divided into three categories: social-economic variables (11 variables), private vehicle variables (3 variables) and public transportation variables (15 variables). Additionally, the performances of these three models are compared in terms of corrected MAPE. The estimation results show that the GWR model performs best with the lowest corrected MAPE value. A total of 7 variables are significantly tested and most of them have rather different parameters across Taiwan, where the percentage of population aged 0-17 has a positive effect to public transportation usage rate of the areas in the eastern Taiwan, especially in Hualian county, but negative effect in most of areas in the northern Taiwan. Since most of these people are captive riders, they depend on public transportation (in rural areas) or their parents private vehicles to school. Number of low-income households has a positive effect in Taoyuan, Hsinchu, and I-lan counties but a negative effect in Nantou, Hualian, and Maoli counties. As anticipation, frequency of inter-city bus has a positive effect to the areas in the northern Taiwan and average bus age of city bus has a negative effect to the areas in the northern Taiwan. It is interesting to note that motorcycle ownership has a positive effect to the areas in Hsinchu and Pingtung counties and a negative effect to the areas in the north-eastern Taiwan, suggesting that people in rural areas use motorcycles to reach public transportation stations but people in urban areas prefer to use motorcycles as the major mode in the whole trip instead. Length of roads has a negative effect to the areas in northern and middle-southern Taiwan. Based on these findings, corresponding strategies are then proposed.