規模報酬與加成定價：台灣資料之實證分析及其應用

Abstract

本文係利用文獻上較可靠的估計方式，並使用較具代表性的長期追蹤資料，估計台灣主要製造業的規模報酬和加成定價等參數。本文證明指出，當影響產出的外生因素爲附著於要素投入上的形式，傳統的規模報酬估計方式將可能產生偏誤，而在此種情況下，本文使用的估計方式仍爲一不偏誤的估計式。本文的實證結果顯示，台灣的製造業大致符合固定規模報酬的生產特性，而定價結構方面則顯示大多數產業具有相當程度的市場力量。此外，本文亦利用此估計式的不偏誤性質，針對隨機邊界模型中，有關技術無效率究竟應以產出導向或投入導向的方式衡量，提出了檢驗的準則。由於技術無效率的設定方式不同，將會影響隨機邊界模型對於規模報酬的估計，但是本文證明發現，上述不偏誤的規模報酬估計方式卻不受影響。因此本文將以隨機邊界模型的估計結果與不偏誤方式的估計結果比較，以判斷何種技術無效率的設定方式較符合台灣的實證資料。

In this paper, we use a more reliable estimation technique in the literature and more representative panel data to estimate returns to scale and markups of main manufacturing industries in Taiwan. We show that when there are factor-embodied exogenous influences on production, conventional estimation method of returns to scale will generate a biased estimate. However, in this situation, we prove that the estimator we use is still unbiased. Our empirical results indicate that constant returns to scale are prevalent, and there is some extent of market power in most industries. In addition, based on the unbiased property of this estimator, we propose a way to determine which representation of technical inefficiency, output-oriented or input-oriented, we should choose. Although the estimator of returns to scale in the stochastic frontier model depends on the choice of the inefficiency orientation, we find and prove that the unbiased estimator is unaffected and still valid. Hence we will compare the estimation results from the stochastic frontier model with those from the unbiased estimator to examine which inefficiency orientation would match empirical data more reasonably.