摘要: | 此研究擬採用歷年經濟部工廠校正資料中的產出關連表, 估計台灣三欄位製造業的規模經濟, 並利用 實證方法的設計, 一併探討隨機邊界模型中技術無效率應為產出導向或投入導向的問題。我們採用 Basu and Fernald (1997) 的方法, 作為主要估計規模經濟的方法。此方法的優點在於它不需對 生產函數形式、技術無效率存在與否等等做假設, 因此可以避免模型誤設的問題。除了Basu and Fernald 的方法之外, 我們還採用了一般的生產函數迴歸分析法, 用以估計規模經濟。特別的是, 我 們分別估計了具有產出導向技術無效率的生產函數, 以及具有投入導向技術無效率的生產函數。由於 Basu and Fernald 的方法不隨技術無效率的設定而有改變, 因此其結果可作為一個基本參考點, 讓 我們將上述兩個迴歸模型的估計結果與之比較, 藉以判斷何種技術無效率的設定較為正確。由於在相 關文獻中, 技術無效率究竟應設為產出導向或投入導向, 一直有爭議, 且沒有適當的判別標準; 因此, 此研究提出的分析方法, 將為文獻提供一個有價值的參考。 In this research, we use the Input-Output Table of the industry census dataset from the Ministry of Economic Affairs to estimate the returns to scale (RTS) statistics of sub-industries of Taiwanese manufacturing firms. In addition, we design the estimation in such a way that it also allows us to test the input- and output-orientations of the technical inefficiency specification. Our main method of the RTS estimation is from Basu and Fernald (1997). The method has an important advantage that it is invariant to assumptions on the functional form of the production equation, the specifications of technical inefficiency, etc.. Therefore, it avoids many of the model mis-specification problems. In addition to the method of Basu and Fernald (1997), we also estimate the RTS statistics from stochastic production frontier models. In particular, we estimate two versions of the model, one with the input-oriented (IO) technical inefficiency and the other with the output-oriented (OO) technical inefficiency. There is no consensus in the literature of whether the technical inefficiency should be characterized as IO or OO, and there is no good way to test it. In our case, because the result from Basu and Fernald method is invariant to the specification of inefficiency, it serves as the benchmark against which the IO and OO assumptions are tested. This provides a novel way to test the IO and OO assumptions. |