透過您的圖書館登入
IP:18.217.84.171
  • 學位論文

後金融危機時期匯豐環球礦業指數之預測研究

Research on the Forecasting of HSBC Global Mining Index in Post Financial Crisis Era

指導教授 : 林建甫

摘要


本研究旨在運用自我迴歸移動平均(autoregressive moving average, ARMA)模型,分析與預測後金融危機時期(post financial crisis),匯豐環球礦業指數(HSBC Global Mining Index)週報酬率。匯豐環球礦業指數為許多礦業類共同基金的參考指標(benchmark),選擇本指數為研究標的,並以金融危機以來時期為研究期間,期能更了解礦業基金的特性,進而作為投資決策參考。 ARMA模型普遍運用在股價、匯率或是投資組合報酬率等時間序列資料的預測上,故以ARMA模型,做為分析與預測匯豐環球礦業指數週報酬率的主要模型。本研究將原始指數取自然對數,計算週報酬率後,進行分析與預測,以2009年1月30日至2012年6月29日匯豐環球礦業指數每週收盤資料為樣本內(in-sample)資料,進行ARMA模型估計與週報酬率分析,以滾動預測(rolling forecast),對2012年7月6日至2012年8月31日資料作樣本外預測(out-of-sample forecast)。 週報酬率分析,選擇適配模型部分,刪除係數不顯著變數,剔除1%顯著水準下無法服從白噪音的估計式,以及移動式Chow檢定排除不具長期結構穩定的模型後,獲得AR(8)與AR(3)兩個跳期數的估計模型。其中AR(8)模型的AIC值與SBC值分別為5.8114和5.8298,均小於AR(3)模型,且AR(8)模型解釋度(R2 and adjusted R2)為8.38%,優於AR(3)的解釋度1.98%,因此AR(8)模型為本研究最適配模型。 在週報酬率預測部分,計算9次滾動預測後的AR(8)與AR(3)估計模型的RMSE、MAE與MAPE平均值,其中AR(8)的RMSE平均值為2.2579,MAE平均值為1.7568均較AR(3)模型小,而AR(3)模型的MAPE平均值為92.6857,較AR(8)模型小。兩模型在不同預測指標,各有預測準確性,因此AR(8)與AR(3)模型均為適配的預測模型。 鑒於國內外文獻缺乏礦業股票指數的預測研究,本文以匯豐環球礦業指數為研究標的,提出落後第8期AR(8)與落後第3期AR(3)週報酬率預測模型,期許可做為後人投資研究或避險上參考。

並列摘要


This research presents the Box-Jenkins model as one of the forecasting techniques in the financial time series. The main aim is to predict the weekly return rate for HSBC Global Mining Index. That is achieved by finding the autoregressive moving average (ARMA) models that describe the equation of the forecasting for HSBC Global Mining Index in the post financial crisis. The index and return rate data are accumulated weekly from HSBC Global Research using the historical data in the period from January 30, 2009 to August 31, 2012. The data from January 30, 2009 to June 29, 2012 are in-sample used to build ARMA model and data from July 6, 2012 to August 31, 2012 are used to do rolling and out-of-sample forecasting. I test the number of the equations by using unit root and then use Akaike information criterion (AIC), Schwartz Bayesian information criterion (SBC), root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), Quandt-Andrews breakpoint test (Chow test) and p-statistics value to choose the best ARMA models at 99% confidence interval. The resulted shows that the best model for HSBC Global Mining Index is AR(8), since this model gives the minimum Akaike information criterion (AIC) and Schwartz Bayesian information criterion (SBC). And the best forecast models for HSBC Global Mining Index are AR(8) and AR(3), because they give the minimum mean absolute percentage error (MAPE), root mean square error (RMSE) or mean absolute error (MAE).

參考文獻


葉淑媚、李佳樺、許天維(2007),「ARIMA 模式分析與預測—以鴻海股票市場日收盤價與報酬率為例」,臺中教育大學學報:數理科技類,第二十一卷第二期,頁51-69。
Wu, C. W., Huang, Y.C. and Wang, C. W. (2009), “Pricing TAIEX Options - An ARMA Approach, ” International Research Journal of Finance and Economics, Issue 31, September, 2009,180-193.
Akaike, H. (1973), “Information Theory and an Extension of the Maximum Likelihood Principle,” in Petrov and Csaki, eds., Proceedings of the Second International Symposium on Information Theory, 267-281.
Box, G. E. P., and Jenkins, G. M. (1976), “Time Series Analysis: Forecasting and Control,” San Francisco, CA: Holden-Day.575.
Carvalhal, A. and Mendes, B, V, M. (2008), “Evaluating the Forecast Accuracy of Emerging Market Stock Returns,” Emerging Markets Finance and Trade, January-February 2008, v. 44, iss. 1, 21-40.

延伸閱讀