臺灣地方政府歲入估測之研究

Abstract

地方政府財政壓力遽增，地方財政問題日益惡化凸顯歲入資源估測的重要性，本研究透過文獻分析與量化統計法檢視民國40年至99年臺灣25個地方政府的歲入估測實務，以平均絕對誤差百分比（Mean Absolute Percentage Error, MAPE）及平均誤差百分比（Mean Percentage Error, MPE），評估歷年地方政府總歲入及各來源別收入的估測準確度及偏頗度表現，並建立迴歸分析模型瞭解影響歲入估測誤差的影響因素。 本研究結果發現，首先，在估測準確度部分稅課收入準確度最高，財產收入和其他收入的準確度最低，MAPE起伏變化最大；歷年數據顯示雲林縣除了財產收入估測最不準確，在其他收入項目表現優良，為估測準確度最高的前三名，花蓮縣、高雄市和連江縣表現亦佳；數據中未發現估測最不準確的單一縣市。其次，估測偏頗度部分，早期多以低估總歲入為主流，民國80年代之後樂觀高估頻率增加；稅課收入、罰款及賠償收入和規費收入主要以低估為主，財產收入和其他收入歷來以高估居多，且偏頗程度高，近年來補助及協助收入的高估明顯，特定地方政府出現虛列此項收入之傾向。 影響歲入估測誤差之決定因素中，代表遞延效果的前期估測誤差最顯著，表示歲入估測準確度和偏頗度直接受到前期誤差的影響，若前期誤差高，當期估測誤差亦隨之升高。在影響準確度因素模型中，總歲入、罰款及賠償收入和規費收入受到時間變項影響；稅課收入、罰款及賠償收入和財產收入在代表經濟環境變動的經濟成長率估測誤差的變項中具顯著性；直轄市的自變項在所有項目中皆不顯著。影響偏頗度之因素模型中，稅課收入、罰款及賠償收入、財產收入和補助及協助收入受到經濟成長率估測誤差影響；代表財政壓力的縣市人均赤字變項在總歲入、罰款及賠償收入、規費收入和補助及協助收入具顯著性；直轄市變項顯著影響補助及協助收入的偏頗度，直轄市傾向低估補助及協助收入；選舉變項僅稅課收入達顯著水準，選後容易出現稅課收入高估情況。 不確定風險下的估測行為必然存在估測誤差，然而估測誤差所生的估測成本並非對稱。在財政壓力下，估測者必須衡量風險承擔能力以決定估測結果，估測主責單位應正視估測上的限制，強化資訊公開並核實估測以實踐預算管理之效。

Under dramatic fiscal pressure, the municipal governments in Taiwan have been facing severely fiscal crisis. It indicates the importance of revenue forecast in municipal governments. This study examines the results of revenue forecasting of 25 municipal governments in Taiwan from the year 1951 to 2010 through literature analysis and quantitative analysis. The research firstly uses Mean Absolute Percentage Error (MAPE) and Mean Percentage Error (MPE) to present the accuracy and bias of revenue forecast errors, then applys panel data model to analyze the impact factors behind. This study finds out that, first, during the period from 1951 to 2010, estimates of tax recipients were the most accurate item among all different revenue. Estimates of property revenue and the item other revenue were the least accurate. According to historical data, Yunlin County had the most inaccurate estimation of property revenue, but in other revenue items it performed the most accurate estimation among all the municipal government. Hualien County, Kaohsiung City and Lianjiang County displayed good performance as well. Second, in the section of assessing the degree of bias, municipal governments tended to underestimate total revenue in the early years. After 1990, municipal governments appear to overestimate total revenue gradually. Tax recipients, fines and compensation revenue and fee revenue have the tendency to be underestimated. On the other hand, property revenue tends to be overestimated and displays the highest degree of bias. In recent years, grants and assistance revenue has been overestimated significantly, and it seems to be an alert of abnormal condition. There are plenty of factors influencing revenue forecast error. The deferred effect of the forecast error in previous year is the most significant. If previous year’s error arises, the inaccuracy and bias of forecast error will then increase. In the accuracy model, total revenue, fines and compensation revenue, and fee revenue were influenced by time. Tax recipients, fines and compensation revenue, and property revenue were significant in the variable of economic growth rate estimation error, reflected that they were also influenced by changes in the economic environment. The dummy variable of municipality was insignificant in all the dependent variable. In the bias model, tax recipients, fines and compensation revenue, property revenue, and grants revenue were influenced by the variable of economic growth rate estimation error. Besides, this study uses counties deficit per capita as the representative of financial pressure. The result shows that deficit per capita was significant in total revenue, fines and compensation revenue, fee revenue, and grants revenue. The dummy variable of municipality was significant in grants revenue and municipalities. For example, Taipei City and Kaohsiung City tend to underestimate it. Elections are prone to affect the forecasting of tax recipients. In the circumstance of uncertainty and risks, every forecast activity would be accompanied with forecast error which has asymmetric cost. Under financial pressure, forecasters must measure the ability of risk-taking to make decisions. The responsible authority should face the restrictions on forecasting behavior, encourage municipal governments to disclosure information, and perform a relatively reasonable estimation.