台灣積體電路產業赴中國直接投資的動態成長分析

Analysis of Dynamic Growth Model in Foreign Direct Investment of Taiwan IC Industry into China

本文首度利用動態成長模型，探討法令限額內，台灣積體電路（IC）赴中國直接投資資金的動態過程。「在大陸地區從事投資或技術合作許可辦法」限制臺灣廠商需於政府核准限額內赴中國投資，故本研究考量該投資上限額度進行模擬分析，該動態模型係一階微分方程，因此本文利用前瞻的數值方法（numerical method）求解與分析影響廠商赴中國直接投資的主要因素。本文進而將研究期間區分為兩段時期，以估計期間樣本建構未考量投資限額的數值模擬模型與迴歸模型，再比較在測試期間不同模型的預測誤差，以驗證考量投資限額的數值模擬模型是否涵蓋較高的模型精確度。本實證結果發現台灣半導體產業當期赴中國直接投資的金額與累積投資金額呈顯著正相關，隱含當台灣積體電路設計或封測產業赴中國大陸的累積金額增加，個別廠商為維護競爭優勢，會增加當期直接投資金額，產業間同儕競相赴大陸投資的內部影響力是驅動當期繼續赴大陸投資的主要動因。此外，IC封測廠商間內部影響因素均大於IC設計產業，研究結果隱含IC封測廠商若在中國設廠，上游設備商與原料供應商會相繼在中國設據點，吸引更多台灣IC封測廠赴大陸直接投資，IC封測產業赴大陸直接投資的群聚效應較為強烈，反之，IC設計廠對上游設備商與原料商的仰賴不及IC封測廠，故IC設計產業內部影響力較IC封測產業為低。最後，模型的精確度檢定證實考量隨時間變動的投資限額數值模擬模型平均絕對誤差水準最低，模型精確度最高，顯示該模型能有效地解釋台灣廠商赴中國投資的演化過程。

Numerical models which incorporate constant or time-varying foreign direct investment (FDI) limits are simulated in our study. Under ‘Regulations Governing the Approval of Investment or Technical Cooperation in Mainland China', Taiwanese firms have to obtain governmental approvals to implement FDI in China and their FDI amounts are confined to the maximum quotas permitted by the government. It is reliable to consider the regulatory limits to explore the FDI dynamics into China. Our dynamic model incorporates the time-varying FDI limits and is specified as an ordinary differential equation (ODE), where the cumulative flow can be solved. Two models which ignore the FDI limits, the regression model and the numerical model without FDI limits, are also estimated, so forecast accuracy of these two models can be compared with the numerical model incorporating FDI limits. The results indicate the positive internal influence through intra-firm communications on the FDIs into China, which suggests that Taiwan IC firms tend to successively imitate experienced firms to undertake FDIs into China. Once more and more IC corporations proliferate production bases in China, the other IC firms are likely to follow and construct production centers to maintain their competitive advantages and share the available resources. Specifically, the internal impact is stronger for IC packaging and testing firms than for IC design ones, revealing the profound imitating inclination of IC packaging and testing industry. In addition, the results of prediction ability illustrate the lowest prediction errors for the numerical model incorporating time-varying FDI limits. Models incorporating time-varying FDI limits perform superior because of taking account of regulatory restrictions. This supports the important influence of regulatory FDI limits.