本論文是利用市場模型評價『利差型債券』，債券付息的方式是連結到30年期CMS及10年期CMS，目前在評價利率衍生性金融商品以市場模型最符合實務的需要，並利用市場上的資料來校準波動度結構及遠期利率之間的相關係數矩陣，再利用蒙地卡羅法模擬出各期的遠期利率，再組合成契約需要的各期 CMS，就可模擬出債券的理論價格，而在模擬的過程中發現，當契約中轉換事件訂的越高，則發行商的利潤越低，訂的越低，發行商的利潤越高。

貨幣需求理論中，長期利率與短期利率常被用來作為貨幣持有之機會成本的變數。然 究竟那一種利率較為適用？在學理上，這個問題帶來很多爭論。近幾年來，一些經濟 學家主張貨幣需求函數只考慮長期利率或考慮短期利率均不恰當，貨幣需求函數應該 針對整個利率結構考慮。這種貨幣需求─利率結構理論，在文獻上又有兩種分析方式 ：一種為：INVENTORY APPROACH，另一種為MEAN- VARIANCE APPROACH 。 本文第一章探討貨幣持有與機會成本間的關係。第二、三章分別就上述兩種APPROACH 作一探討；並就台灣資料作實證分析。篛四章討論採用多利率變數時在計量上所遭遇 的問題及其解決方式。第五章為結語。

在過去，不外乎藉由瞬間短期利率的隨機過程或瞬間遠期利率的隨機過程來描述利率期間結構，應用這些方式理論上雖然可行，但是市場上並無法觀察得知這些瞬間利率。1997由Brace、Gatarek及Musiela提出之LIBOR市場模型，直接推導市場上可觀察得到之LIBOR利率的隨機過程，因此不需如傳統評價模型尚須對利率做轉換，可以直接以市場上觀察到之LIBOR報價帶入模型中做評價。由於市場上有愈來愈多的利率衍生性商品，不是由單純的cap或是swaption來組成，因此很難求出封閉解，所以通常使用數值方法來解決評價的問題，常用的數值方法有樹狀圖評價法及蒙地卡羅模擬法，由於使用樹狀圖評價法必須對利率做假設，才能使項樹的節點重合不至於增加太多的運算困難；因此，本文選擇使用蒙地卡羅模擬法，透過機率測度的轉換，推導出符合商品設計的遠期LIBOR利率的動態過程，進而模擬出商品的價格，在LIBOR市場模型下使用蒙地卡羅模擬法的好處在於，只要了解商品的設計方式，針對不同商品尋找合適的遠期LIBOR利率動態過程，便可利用模擬的方式得到商品價格。

2008年初，正值美國籠罩次級房貸風暴影響、全球經濟景氣趨緩、產油國地緣政治因素造成能源價格創新高…等險峻經濟狀況之際，投資銀行設計一包含 :搭配出局條款之CMS Spread雪球型利率結構商品及結合附加WTI上限、USD/JPY匯率上下限之異型選擇權的混合結構性商品提案。本文依據標的資產屬性，參考相關文獻及近期在頂級期刊發表之利率資產評價模型研究中，選用Extended BGM模型(Ting-Pin Wu, and Son-Nan Chen(2007))、遠期曲線模型及匯率評價模型為個案之基礎評價模型；以無套利觀念依取得之市場各資產相關公開報價資料估算各模型所需之參數；由於屬於雪球利率結構型商品及路徑相關特性，在目前相關文獻無封閉解的條件下，使用蒙地卡羅模擬獲得未來各資產之現金流折現值，進而計算預期理論價值。依據上述方法論評價所獲得之預期理論價格顯示，個案並非具公平價值之交易，依此結論強烈建議客戶不應該承做本交易。 個案相當於投資銀行以買入一個5年期附帶出局條件，隱含看空經濟景氣循環之CMS Spread選擇權及買入一個1年期看空WTI價格選擇權建構此混合結構性商品。為強化客戶承做意願，設立一似乎觸及機會很大，但從交易後至今從未觸及的出局條件，又透過每日數位選擇權計息方式將WTI波動度資產化，提供大於10%之相對LIBOR rate 很高，但實際是被低估之半年收息固定費率。由於雪球型利率結構型商品特性，收益不僅取決於是否達成交易付款條件，更重要因素是達成時間點之速度。 在蒙地卡羅模擬資產價格路徑中，觸及頭一次CMS Spread付款條件天數之眾數區間為125至135，貼近實際136天。從評價結果，交易之付款條件內已隱含透過兩個不同標的資產選擇權之高預期獲利相互達到避險、套利及強化收益等效益;投資銀行可以不用額外對受眾多複雜不確定因數影響之WTI價格採取避險策略，而將所有避險成本轉嫁於選擇權賣方的客戶。在資本計提規範下及確保未來預期收益之考量下，投資銀行唯一要做是以低成本尋求中介銀行進行背對背交易以強化因市場風險所衍生之信用風險。 從研究過程，不禁讚嘆個案是投資銀行設計建構在財務工程科學上的卓越藝術及策略，從它一旦出現世界上之瞬間，個人預估其價值將達34,211,458.09美元! / Early 2008 was a steep economic era when U.S. was enveloped by subprimemortgage crisis, world's economy was slowing down, and energy prices were pushed to a historical record high by oil geopolitical factors. Under this situation, an investmentbank designed a hybrid structure product, which includes a CMS Spread Snowball interest rate structured product with USD/JPY FX rate Knock out condition, a WTIoption of an additional upper limit, a USD/JPY exchange rate combined exotic option of upper and lower limits. After considering assets attributes and reviewing the relevant literature and recent research published in top journals related to the interest rate assetpricing model, Extended BGM model (Ting-Pin Wu, and Son-Nan Chen (2007)), forward curve model, and FX Rate model are selected as the basic pricing models. Tocalculate the expected theoretical value of this structured product, the unavailable model parameters of assets are estimated through the public market data based on thearbitrage-free concept, and the discounted values of the assets future cash flows are obtained by Monte Carlo simulation because of snowball interest rate structured product and path dependency characteristic and no close form solution in current relevant literature. The results of the pricing models shows that the net present value(NPV) received by customers is lower than that received by the investment bank, theconclusion is : Strongly recommend customers should not to do this trade ! In this case, the investment bank used a long position of one 5-year period CMSSpread Option with knock out condition, which implies Bearish on the economic cycle, and a short position of a 1-year period WTI option with up and low limits condition to construct this hybrid structure product. To draw customers’ attention to this proposal, the investment bank designed a knock out condition that seemed to be met very easily,but the price never touched by the article finished date. Additionally, a daily accrued digital option is used to transfer WTI volatility to a semi-annual fixed yield over 10% that, compared to LIBOR Rate, is very high but actually is underestimated. For theSnowball structure product, the total profit depends on not only when but also, more importantly, how soon to meet the payment condition. According to the asset pricing path generated by Monte Carlo simulation, the mode range which CMS Spread payment condition first met is 125 to 135 days after the contract’s value date, very close to the actually history data of 136 days. From pricing results, terms of contract implied that two different options combined to hedge risk and gain profit from each other. Hence, the investment bank does not need to make extrahedge strategy to WTI price which is impacted by more complicated risk factors.However, customer must spend hedge cost because of taking much risk as a sell option role. Under the Capital Charge regulation, to lock up the expected profit, what the investment bank needs to do is only to pay a very low cost fee, which like insurancepremium, to look for an intermediary bank to offer a back to back trade to manage thecredit risk caused by market risk! During the research of this paper, I am amazed what an excellent art and strategy that designed by the investment bank based on financial engineering science! As this structure product appeared in this world, I estimated that it would worth 34,211,458.09 USD.

經由愈來愈多的實證研究發現，的確在利率的變動過程中，除了包含連續性行為，即遵循”擴散”模式(diffusion process)，亦包含了不連續性行為，也就是有著跳躍(jump)的情形發生。因此顯示出假設利率隨機過程僅為連續性的擴散模型已是不足夠的，跳躍-擴散模型(Jump-diffusion model)顯然會比純粹擴散模型有著更好的解釋能力。而市場模型(LIBOR market model)的提出，則說明了遠期LIBOR利率模型較能描述市場實際的利率型態，並且可方便使用市場資訊，進行模型參數校準。 所以本研究旨在以LIBOR market model 加上跳躍過程，即遠期LIBOR利率的跳躍-擴散模型，分別針對歐洲美元期貨與利率結構型債券中的滾雪球式累息債券建立評價方法。由於所選用動態模型的複雜度，使得封閉解的求出不易，因此在文中，最後是採用蒙地卡羅模擬法，求兩商品的數值解。在後續研究上，本文還挑出了幾個最直接影響商品價值的因素，如殖利率、波動度、跳躍幅度等，進行各種情境下商品價值的敏感度分析，以提供投資人與發行商在考量風險因子所在時的一個參考。