本研究的目的在於利用台灣證券交易所掛牌上市的製造業公司在1990至2003年的追蹤資料, 以動態觀點來實證檢驗廠商創新持續性及廠商創新活動的決定因素。利用dynamic random effect probit model, 控制個別廠商無法觀察之異質性及解決初始條件問題後, 實證結果證實台灣製造業廠商存在創新持續的現象。再者, 進一步將樣本區分成高技術機會群與低技術機會群兩群, 分別檢驗不同技術機會環境下的廠商創新行為是否持續。結果顯示, 不論是否為科學性產業的創新行為皆存在創新持續的現象。而此效果尤以科學性產業較為明顯。本文亦強調影響廠商創新績效的重要因素, 此皆為有助於瞭解廠商創新活動的決定因素。

This study examines the persistence of innovation and the determinants of innovations in Taiwan’s manufacturing sector, using a panel dataset of manufacturing firms over the period 1990–2003. This study employs a new estimation method for the dynamic random effect probit model to account for firm heterogeneity and also to handle the problem of initial conditions. The empirical results find a strong effect of state dependence after controlling for firm heterogeneity and the initial condition, supporting the hypothesis of true persistent innovation. This finding is also supported for subgroups of scientific and non-scientific industries, implying that the causal effect from past innovation to current innovation exists across industries.
Less surprisingly, the phenomenon of innovation persistence is stronger in scientific industries relative to non-scientific industries.We also highlight some important factors that affect continuity in the performance of innovations.