傳統上使用隨機效用模型估算自然資源的遊憩使用效益，均假設所得邊際效用為一 常數項而忽略非線性所得效果的存在。本文放棄此一限制條件，同時考量線性與非線性隨機 效用間斷選擇模型，利用民國84年國人國內旅遊狀況調查為實證資料，運用Multinomial Logit 與Nested Logit估算模式，並採用Linear、Generalized Leontief與Translog型式及使用 bootstrapping方式，估算國人前往森林遊樂區的遊憩效益。估計結果顯示在一天一地的樣本 下，每人每次的遊憩價值介於165至241元之間；而以二天一地的樣本，則遊憩效益值則介 於366至714元之間。本研究亦發現，加入非線性所得型式所估計而得的效益值具有較大的 變動區間，但機平圾值與中位數而言，是否加入非線性所得效果的影響並不如模式估計間選 擇的影響。

　　　　　Traditional random utility model imposing a constant marginal utility of income on individual's preference has been widely used to evaluate recreational benefits in natural resources uses In this paper we use both linear and nonlinear random utility models to consider the impacts on benefit estimates from the assumption of constant and non-constant marginal utility of income. Both the multinomial and nested logit models are used to assess the magnitude of forest recreational benefits. By employing the linear, generalized Leontief and Translog functional forms and applying bootstrapping resample method, the results show that benefit estimates range from NT$165 to NT$241 per-trip based on a single-day trip sample The estimates are much larger and range from NT$366 to NT$714 per-trip from estimation for a two-day trip sample Our results also indicate that benefit estimates have more variations from the estimation of a nonlinear random utility model than those from a linear random utility model However, the mean and medium values of benefit estimates do not show much difference between linearity and non-linearity assumption. The impacts on model selection between multinomial and nested logit are larger that those on functional form Our result is consistent with a recent finding from Herriges and Kling (1999).