- 學位論文
現金流量風險值之估計—Linear、STAR與STARX模型之比較分析
The Estimation of Cash Flow at Risk─A Comparison among Linear, STAR, and STARX Models
摘要
近年來隨著金融國際化與自由化的熱絡，全球金融體系正面臨結構性的改變，企業經由海外投資，雖可讓資金運用更有效率，卻也承擔更多的不確定因素與風險。由於產業特性與資本結構的不同，相較於傳統的敏感度分析（Sensitivity Analysis）與VaR而言，CFaR更適合作為非金融產業衡量風險的指標。本研究試圖求得公司最適現金流量預測模型，並透過CFaR模型更精確地評估公司營運的風險衝擊。 本研究以景氣領先指標綜合指數、消費者物價指數、新台幣兌美元匯率與一銀三個月期定期存款利率等四種總體經濟指標來解釋現金流量，並預測現金流量，進而評估公司CFaR。研究對象為台灣50指數中，上市年數長達十年以上的非金融產業公司，研究期間為1996年第4季至2007年第3季共計44筆季資料。此外，為檢驗公司現金流量的變化過程是否為一非線性路徑，以及公司現金流量與總體經濟變數間是否存在著非線性關係，故本文亦採用Teräsvirta and Anderson（1992）與Teräsvirta（1994）所提出的線性檢定程序，分別檢定AR模型與多元迴歸模型是否存在STAR與STARX型式的調整行為。 實證結果顯示，在AR模型的線性檢定中，除了遠紡、台達電子、聯強、寶成等4家公司外，其餘17家公司均存在STAR型式；而在多元迴歸模型的線性檢定中，除了台達電子外，其餘20家公司皆存在STARX型式。在樣本內估計方面，STAR模型與STARX模型相較於線性AR模型與多元迴歸模型而言，皆提供較佳的配適度。至於樣本外預測結果則顯示，除了亞泥、台塑、聯電、仁寶、鴻準、統一超商等6家公司適合採用非線性模型估計CFaR外，其餘15家公司仍以線性模型估計CFaR較佳。這可能是因為台灣50指數中的非金融產業，大多為穩定成長的公司，且本研究受限於樣本數不充足的原因，進而導致非線性模型無法全面優於線性模型。
並列摘要
In recent years, global financial system is facing structural changes, derived from the blooming internationalization and liberalization on the financial environment. Although enterprises can invest overseas more efficiently, they also tolerate the impact of much more uncertain factors and risks. Due to the industrial characteristic and capatial structure, CFaR provides a more proper indicator to measure risk of non-financial industries than traditional sensitivity analysis and VaR. This study attempts to find out the optimal model for predicting cash flow of company, and further to evaluate company’s operating risk by CFaR model. This thesis uses four kinds of macroeconomic indicators, including Composite Leading Index, Consumer Price Index, TWD/USD Exchange Rate, and 3-month Deposit Rate to construct cash flow model and assess CFaR of company. The sample objects are the non- financial companies of the Taiwan Fifty index which have been listed more than ten years. The Sample period spans from the fourth quarter in 1996 to the third quarter in 2007. In order to examine whether the cash flow of company is a nonlinear process and a nonlinear relationship exists between cash flow and macroeconomic variables, we apply linearity tests used by Teräsvirta and Anderson (1992) and Teräsvirta (1994) to examine AR model and multiple regression model. Except Far Eastern Textile, Delta Electronics, Lemel, and Pou Chen companies, other companies exist the type of STAR model in linearity tests of AR model. Except Delta Electronics company, other companies exist the type of STARX model in linearity tests of multiple regression model. The empirical results show that STAR(X) models provide better goodness of fit than linear models. Besides, this study also compares the out-of-sample forecasting performance of linear, STAR, and STARX models. Except Asia Cement, Formosa Plastics, United Microelectronics, Compal, Foxconn, and President Chain Store companies are appropriate to adopt nonlinear model in evaluating CFaR, other companies are appropriate to adopt linear model in evaluating CFaR. The possible reason is that the non-financial companies of the Taiwan Fifty index are the blue chip companies which grow stably, or the sample companies in the Taiwan Fifty index are not enough that nonlinear model is not superior to linear model.