半導體產業資本支出與臺股電子指數之實證研究

Abstract

本研究探討半導體產業變數、資本支出利率、貨幣供給與臺股電子指數之關聯性，利用三種單根檢定(ADF、PP、DFGLS)、Johansen共整合檢定、向量自我迴歸模型、誤差修正模型，並研究變數間之衝擊反應函數、預測誤差變異分解、Granger因果關係檢定。時間序列資料之研究期間自1995年1月至2018年12月，共288期之月資料。實證結果發現五種變數與大部分的經濟變數相同，皆為非定態資料，屬於I(1)。經過共整合檢定，發現存在一組共整合向量，變數具長期均衡關係。透過Granger因果關係檢定發現，除了晶圓代工銷售額外，臺股電子指數可Granger領先其他三種變數，與M1B、北美半導體設備出貨金額為雙向領先落後關係。在解釋誤差變異之部分，北美半導體設備出貨金額是除了臺股電子指數本身外最顯著的解釋變數，其解釋能力隨著期數增加，因此建議投資人在建立操作策略時能夠將此指標納入考量。除此之外，同時發現北美半導體設備出貨與晶圓代工銷值具有雙向的領先落後關係，意即晶圓代工廠增加資本支出後對於整體銷售是有助益的，而當半導體市況轉佳，代工廠會持續增加資本支出以搶市並維持競爭力。

This research studies the relationship between semiconductor industrial variables, interest rate for capital expenditure loans, money supply and Taiwan electronics sector index by using unit root test(ADF, PP, DFGLS), Johansen cointegraiton test, VAR, VECM, impulse response function, forecast error variance decomposition and Granger causality test. Period of monthly data is from January 1995 to December 2018. The empirical results show that the five variables(ELI, R, M1B, SV, EQ) are non stationary and all have the unit roots, just like other common macroeconomic variables. Besides, the five variables are I(1) with the long term relationship, and there exists one cointegration by Johansen test. Through the Granger causalit test, the results show that Taiwan electronics sector index can Granger cause others except the north America semiconductor billings. There also exists the mutually lead-lag relationship with M1B at 0.01 significant level and with the north America semiconductor billings at 0.1 significant level. In terms of impulse response function and forecast error variance decomposition, Taiwan electronics sector index is affected by the semiconductor billings except itself. Therefore, the investors can take the semiconductor billings as reference before making the investment decision.