Relationship between Industry Agglomeration and Firm Productivity : Evidence from Taiwan Electronic Industry

本文利用2004-2009的經濟部工商校正暨營運調查與鄉鎮經緯度資料，以生產、研發活動的EG Index及NNh方法建構出的產業聚集圈，試圖描述台灣電子產業整體的群聚現象，並探討廠商聚集程度高低是否影響他們的生產力。從生產面EG指數來看，電子產業均值為0.0002，屬低度聚集，但在半導體所屬的電子零組件製造業中則為0.0253，屬中度聚集。足見台灣電子業在不同類型的子產業的聚集程度上有差異，且與傳統上認知的高科技業為聚集產業的現象一致。研發面EG指數的平均則為0.0917，顯示台灣電子業在R&D活動上為高度聚集產業。從聚集圈衡量方式來看，台灣電子業集中在北、中與南部主要城市內及其周邊，與都市經濟的發展有很大的關連性。以二欄位產業觀察，分佈情況與技術密集程度有關。 在產業聚集程度與廠商生產力關係的驗證上，首先我們利用firm-level資料，整理出一個具有7080個個體，4年共28320筆觀察值的Balanced Panel Dataset。接下來以價格調整後勞動附加價值衡量的生產力作為被解釋變數，分別以落後一期的EG指數及聚集圈虛擬變數作為主要解釋變數建構出兩個Panel Regression模型。在利用Hausman test作檢定並決定使用固定效果(Fixed Effect)後，得到我們的實證估計結果。結果發現，台灣電子業廠商的生產聚集程度與生產力為負向且顯著的關係，而在研發聚集程度方面，與生產力成正向且顯著的關係。同時，兩者的平方項係數皆為負，表示兩者與生產力間均有非線性關係。研究結果顯示，(1) 電子業聚集產生的負向外部效果，不管在生產或研發上均已大過正向外部效果，並產生聚集不經濟的情形。(2) 聚集圈方法的估計結果與EG一致。(3) 勞工薪資水準、工廠規模與研發支出均對生產力有顯著影響，且平均薪資越高、研發支出越高、規模越小，生產力越高。In this thesis we use the data of Taiwan’s electronic firms from the period of 2004 to 2009, aiming to understand the relationship between the phenomenon of industrial agglomeration and firm’s productivity. First we adopted EG index and NNh method to describe overall geographical cluster of the industry. From the perspective of production EG index, it has an average of 0.0002 and is defined to be in the range of lowly concentrated industry. However, if we focus on the situation in manufacture of electronic parts and components industry, average came out to be 0.0253. This shows that agglomeration divergence of different sub-sectors within the industry do exists, and ICT industry is still relatively concentrated in accordance with past research. Average R&D EG index is about 0.0917, which means the research activities in the indsutry are highly concentrated. In the other hand, NNh method points out that electronic firms’ factories are located around the northern, middle and southern part of Taiwan, and close to major cities, which implies that the delopment of Taiwan’s electronic firms is closly related to urbanization. Finally we sort out a balanced panel dataset that has 28320 observations and fit two panel regression fixed effect models with price-adjusted labor productivity as dependent variable. There are three major results: (1) Both production and R&D agglomeration has non-linear relationship with productivity, which implies that the effect of negative externalities outruns that of positive externalities and has resulted in agglomeration diseconomies. (2) NNh method has similar results compared to EG Index and could be furthur examnied in the future. (3) Higher Average employee salary, smaller size of factories and larger R&D expenditures are significantly resulted in higher firm productivity