學習機能考量下之適應性加速調整行為

Learning Mechanism and the Flexible Accelerators

無論是靜態或動態的對偶理論，在進行實證分析時，函數型態的選擇扮演著相當重要的角色。早期關於投資行為的研究多以適應性加速（flexible accelerator）的資本調整方式來探討廠商對資本投入的需求，而應用動態對偶理論的實證文獻除了對於價值函數做了過於缺乏彈性的設定，其中之適應性加速調整更隱含準固定要素以固定速度調整至其長期均衡的水準，這兩個設定均與理論文獻中所得的結果不相符合。本文以范光中與陸怡蕙(2003)之理論模型為基礎，討論非靜態預期與學習機能考量下的適應性加速調整行為，結果發現，在學習機能的考量下，對價值函數的函數型式做較為缺乏彈性的設定是必要的；但準固定要素的調整速度則並非固定不變，廠商對未知資本價格的認知與考慮新增訊息以修正其對參數的估計均將影響準固定要素的調整速度。

This paper investigates the general functional specification of thevalue function and the behavior of the flexible accelerators, undernonstatic expectations and learning mechanism. In most empiricalstudies applying the dual approach, either static or dynamic, choices offunctional forms are critically constrained. For investment behaviorstudies, in particular, the assumption of flexible accelerator has beencriticized for the restricted functional specification and the assumedconstant rates of adjustments. Based on the extended dynamic dualstructure in Fann and Luh(2003), we find that it is inevitable to specify arigid functional form for the value function. However, instead ofadjusting in fixed proportions, quasi-fixed factors adjust in variableproportions that are shaped by firms’ expectations and learning behavior.