DRAM季價格預測

Abstract

DRAM (dynamic random access memory) 價格劇烈變化使得DRAM市場中的買賣雙方經常經歷突來的暴利或者嚴重的虧損。在過去的文獻中曾經討論年度的DRAM價格變化，而這種模式較適合用在長期性的決策問題，例如，決定DRAM廠的投資建廠時點；然而，對於中短期的決策問題像是DRAM市場中買賣雙方的存貨決策，則應該以季為基礎的價格分析對實務界才更幫助。 本研究根據近兩年 (1999-2000)的資料,發展一個以64M DRAM為主的季價格預測模式。在本研究中，首先驗證之前學者所提出的年價格預測模式的有效性以及其是否同樣適合用在以季為基礎的分析資料中，而驗證結果並不令人滿意。於是，我們便提出其他影響價格的變數，並且使用簡單的線性回歸技術以找出較好的季價格預測模式。在本研究中經過170條以上的回歸方程式，結果顯示64M DRAM 的約當價格（equivalent price）與DRAM 使用量成長率、DRAM製造商的存貨成長率有非常高度的顯著相關。這個複回歸方程式在預測2001年第一季 64M DRAM 的季價格結果是本研究中所有回歸方程式中最好的。

The price of DRAM (dynamic random access memory) has been so highly fluctuated that DRAM makers and buyers would gain or lost a lot due to unexpected price change. From a literature survey, the price forecast model for DRAM on a yearly basis has been developed. Such a model would be more helpful to the long-term decision making problem, such as deciding the timing of establishing a DRAM plant. However, for the case of mid-term or short-term decisions such as the inventory policies of makers and buyer, the price forecast on a quarterly basis would be more helpful. Based on the data in two years (1999-2000), this research aims to develop a quarterly price forecast model for 64M DRAM. We first justifies the effectiveness of applying the yearly basis model to forecast the DRAM price on a quarterly basis, it turns out that the performance is not satisfactory. We then propose some other variables and use simple linear regression techniques in order to find out a better quarterly price forecast model. Over 170 regression equations have been tested. The results show that the growth rate of 64M DRAM equivalent price is highly correlated to the growth rate of DRAM consumption and the growth rate of DRAM manufacturers’ inventory. The associated regression equation, among the tested ones, showed the best performance in forecasting the 64M DRAM price for the first quarter of 2001.