空氣污染防制費自民國84年7月1日開徵，希望能藉由空污費的徵收來減少污染的排放量，同時能透過空氣污染防制費的有效運用來改善空氣品質，而本文僅針對空氣污染防制迄今，對家計部門所產生的清潔效益進行估算。

首先，利用家計部門生產函數建立理論模型，接著以錢玉蘭、蕭代基(1996)，「台灣大都會地區改善空氣品質之經濟效益評估與酸雨風險認知調查」的問卷調查資料為樣本，對超對數及近似理想需求函數的實證模型進行配適比較。結果顯示：近似理想需求函數配適之判定係數較高，符合加總性、齊次性及對稱性限制條件，而價格變數與屬性變數之估計結果亦相當合理，因此，推論近似理想需求函數為解釋台灣清潔需求型態的較適模型。

由近似理想需求函數所推估之85及86會計年度空污費執行所帶來台灣地區家戶清潔效益分別為每戶2101元及651元，加總每戶效益後，台灣地區總清潔效益分別為128億元及39億元，與空污費實際執行的歲出相較，推論空污費執行迄今之效益大於成本。並由推算之各都會地區補償變量性質，建議政府以效率的觀點使用空氣污染防制費時，應將防制資源優先使用於改善都市化程度愈高，人口數愈多，污染程度愈嚴重地區的空氣品質。

The Environmental Protection Agency of Taiwan launched the Air Pollution Emission Fees Program to levy emission fees on petroleum products, July 1, 1995. The purpose of such a program is to provide economic incentives to change peoples' behavior towards the consumption of fuels and to collect fees to be used for the improvement of air quahty.

The objective of this study is to evaluate the effect of the emission fees program on household cleanness. In order to achieve this objective, we first employ a household production function in accordance with the theory of consumption behavior. Two models are utilized in this study, i.e., the translog firnction model and the almost ideal demand function model. We utilize cross-section data to fit into these models and find that the almost ideal demand model has more significant explanatory power. Therefore, we infer that this model is a better functional form for expressing the cleanness behavior of Taiwan's households.

The estimated results of the total benefits for household cleanness resulting from the implementation of air pollution emission fees are NT$12.8 and NT$3.9 bilhon for the accounting years of 1996 and 1997, respectively. By comparing expenditure from the air pollution fees, which were designated for the purpose of improving air quality, i.e., NT$4.2 and NTS3.3 billion for the respective years 1996 and 1997, we can see that the pollution control program is conformable with the benefit-cost principle. Finally, based on the empirical results, pohcy recommendations are presented in the conclusion.