模型選擇與動態設定－－以台灣漁產品需求為例

Model Selection and Dynamic Specification- An Application of Consumer Demand for Fish in Taiwan

本研究目的在利用1979-1998年漁產品消費需求資料，經由估計及檢定來驗證鹿特丹需求體系（roterdam demand system，RDS）及近似理想需求體系（almost ideal demand System，AIDS）何者較適合台灣漁產品消費需求分析，並尋求適當的需求體系動態結構以掌握漁產品消費資料的特性。研究結果顯示在模型的選擇上，經由Davidson and Mackinnon（1981）非嵌套檢定（non-nested tests）發現，一般化動態近似理想需求體系相對於鹿特丹需求體系有較佳的資料配適。而一般化動態近似理想需求體系可進一步簡化為部分調整及一階自我迴歸兩種近似理想需求動態模型。一階自我迴歸模型不被拒絕，表示漁產品消費需求存在慣性（inertia），消費需求不僅受到當期總支出及價格影響，也會受到前期漁產品總支出與價格影響。部分調整模型不被拒絕，表示習慣持續（habit persistence）在漁產品消費行為中扮演重要角色，即前期消費水準會影響到當期的消費。換言之，台灣消費者在購買漁產品時其消費行為具有慣性及習慣持續。

The aim of this study is to examine systematically the pattern of fish consumer demand in Taiwan during 1979-1998 by estimating and testing how well widely used Rotterdam demand system and Almost Ideal demand system fit the data, and to identify the appropriate dynamic structure to represent the data. Results indicate that the General Dynamic Almost Ideal demand system fits data better than the Rotterdam demand system by non-nested tests. The general dynamic almost ideal demand system as the maintained specification, the partial adjustment and the autoregressive are not rejected at the 1% level of significance. The first autoregressive form is not rejected, this implies that changes in fish demand not only depended on actual total expenditure and prices but also on lagged values of these variables. The first partial adjustment form is not rejected, this implies that habit persistence play an important role in explaining fish consumer behaviour and that past consumption levels has exerted influence on actual consumption. In other words, it can be said that Taiwan consumers show some inertia and habit persistence in their fish consumption behaviour.