地方政府支出效率之研究

Efficiency of Local Government Expenditures in Taiwan

吳 仂(Le Wu)

指導教授 ： 李顯峰

國立臺灣大學/社會科學院/經濟學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.02507

摘要

本研究主要目的在於探討地方政府的支出是否具有效率，以及各縣市支出的可改善空間，最後討論影響縣市支出效率的因素。本研究以台灣地區18個縣市做為觀察對象，樣本期間包含民國91至100年間共計10年的年度資料，投入項為單一投入，即歲出總決算，產出項從各支出政事別選取一項指標代表，分別有國中小學生數、縣鄉道長度、老年加身心障礙人口、行政機關公務人員數、刑案破獲數及垃圾清運量，並運用資料包絡分析法(Data Envelopment Analysis, DEA)其中之CCR(Charnes , Cooper and Rhodes)與BCC模式(Banker, Charnes and Cooper)來估算出各地方的支出效率。
經計算後發現，在CCR模式下，效率值前三高的縣市為彰化縣、雲林縣和桃園縣，最低的前三名為新竹市、新竹縣和苗栗縣。於BCC模式下，新北市、台南市、彰化縣、雲林縣、澎湖縣和嘉義市為最具效率的縣市，表現較差的縣市為基隆市和苗栗縣。
接著再運用差額變數分析(slack variable analysis)，探討各縣市十年來合計之可節省支出空間，在CCR模式下，可節省空間最多的是台南市的1,011.62億元，所有縣市總計可節省空間為4,994.92億元。在BCC投入導向模式下，可節省支出空間最多的縣市為苗栗縣，共計417.01億元，所有縣市總計可節省空間為2,379.13億元。
最後利用Tobit迴歸模型(Tobit regression model)分析下，人口密度越低、平原面積比例越高、服務業就業比例越高、公教人員人力素質越高、補助依存度越低、融資依存度越低的縣市對支出效率帶來正面的影響。15歲以上高教比例和支出效率呈負相關，但只有在CCR模式下顯著；每人可支配所得和支出效率呈負相關，但在三個模式下皆不顯著。

關鍵字

地方財政支出 ； 績效評估 ； 資料包絡分析法 ； 差額變數分析 ； Tobit迴歸模型

並列摘要

The study mainly focuses on discussing the efficiency of local government expenditure, and the improvement in Expenditure for every county and city. At last, it will discuss about the factors with influence on the efficiency of local government’s expenditure. 18 counties in Taiwan in the period of 2002 to 2011 were investigated. The empirical model includes one input variable that is total expenditures, and output variable selects one criteria from the number of students in primary and junior high school, the length of the, the population of elders and disabled people, the population working in the government, the number of the criminal cases solved, and the number of waste conveying. In this paper, Data Envelopment Analysis (DEA) is employed, and two models are applied to assess their efficiency, namely the CCR model initiated by Charnes, Cooper and Rhodes in 1978, and the BCC model proposed by Banker, Charnes and Cooper.
In the CCR model¸ the most efficient three Counties are Chunghwa, Yuanlin, and Taoyuan. The lowest efficient three Counties are Hsinchu City, Hsinchu, and Miaoli. In the BCC model, Tainan City, Chunghwa County, Yuanlin County, Penghu County and Jiayi City are the most efficient counties. Keelung City and Miaoli City are bad in efficiency.
Then, by the slack variable analysis finds out the expenditure can be saved in the last ten years. In the CCR model, 101.162 billion N.T. dollars can be saved. All counties and cities can save 499.492 billion N.T. dollars. In the BCC model, Miaoli County can save the most expenditure 41.701 billion dollars, and all counties and cities can save 237.913 billion dollars.
Finally, in the Tobit regression model, factors can have positive influence on expenditure are the “lower the density of the population”, “higher proportion of the plain area”, “higher rate of services industry”, “the higher quality of the public servants”, and “ lower dependence on the loan and subsidy”. The Higher Education proportion above 15 has negative relevance to the efficiency of expenditure, but it is only significant in the CCR model. Every person’s disposable Income has negative relevance to the efficiency of expenditure; however, it is not significant in the three models.

並列關鍵字

Local Government Expenditure ； Performance Evaluation ； Data Envelopment Analysis ( DEA ) ； Slack Variable Analysis ； Tobit Regression Model

參考文獻

王肇蘭、許義忠、徐偉初(2008)，「台灣地區地方政府效率暨生產力之評估」，《行政暨政策學報》，84，71-120。

汪明生、曾玉祥(2007)，「台灣縣市政府自有財源運用效率之衡量－資料包絡分析(DEA)之應用」，《台灣銀行季刊》，58(1)，99-112。

吳濟華、劉春初、馮永猷(2010)，「台灣地方政府公共支出效率衡量之實證研究」，《行政暨政策學報》，50，33-88。

林健次、蔡吉源(2003)，「地方財政自我負責機制與財政收支劃分」，國立政治大學《公共行政學報》，9，1-33。

章定煊、劉小蘭、尚瑞國(2002)，「我國各縣市財政支出與經營績效之研究」，《台灣土地研究》，5，45-66。