本研究旨在協助產險公司建立動態財務分析系統. / Despite the promise of DFA, many insurers have grown increasingly frustrated with it. Evolution of DFA models leads to a vicious cycle of implementation, compilation and modification that disrupts the creative evolutionary modeling activity. To overcome this obstacle, MLDFA discussed in this paper provides an improved approach that would meet the desired requirements. MLDFA architecture plays a central role in its successful development. Model Transformation Systems makes a transformation from the conceptual model into programmed model by combining conceptions of Object Oriented Programming (OOP) and Code Generation. With the object-oriented style user interface, MLDFA provides the user with means to conveniently structure the model in a natural way. We believe that the proposed approach is suitable and feasible for the formulation and refinement of DFA models. Although MLDFA is in an early stage of implementation, it has the potential to bring DFA to a wider audience because it could help insurance companies in total cost reduction of the life cycle DFA system.

本文使用政大資管系及風管系所共同開發的動態財務分析模型(dynamic financial analysis model)，並測試該模型是否能準確區別出健全的產險公司及喪失清償能力的產險公司。我們進一步地使用羅吉斯迴歸模型分析該模型在預測產險公司清償能力的準確性。 從迴歸模型實證結果中指出在10%顯著水準下所有變數皆不顯著。此結果顯示我們的實證測試無法提供強烈的佐證以支持”該動態財務分析模型能夠準確地預測保險人清償能力”的說法。 根據實證結果，我們建議往後的研究可以使用不同年度的資料，藉由大量的樣本以增加統計分析的準確度，同時改善該動態財務分析模型以符合保險人擁有多種再保安排的實際狀況。在實證測試中，我們發現該模型仍然存在一些錯誤。假使該動態財務分析模型能有效率地消除這些錯誤，我們期待修正後的動態財務分析模型在預測產險公司的清償能力上有更好表現。 / This paper set out to empirically test whether the dynamic financial analysis model (DFA), developed in a joint project of Department of Management Information Systems and Risk Management and Insurance, National Chengchi University, could accurately classify both solvent and insolvent property-liability insurers. We used a logistic regression model to analyze the solvency prediction accuracy of the DFA model. The empirical results indicated that none of the variables were significant at the 10% level and did not offer strong supporting evidence that the DFA model could accurately predict the solvency of insurers. Based on this, we suggest that further research should perhaps use data over different years to increase the accuracy of the statistical analysis, by using larger samples; this may improve the DFA model by coordinating actual situations with various reinsurance arrangements. In the empirical tests, we found that the DFA model still has some bugs. If these bugs can be efficiently deleted, we expect a revised DFA model to perform well in predicting the solvency of property-liability insurers.

確保保險公司的清償能力一直是保險監理的重心。在所有施行的保險清償監理工具中，風險基礎資本（Risk-Based Capital, RBC）是目前為止最先進的代表。然銀行監理機關已經推薦涉險值（Value at Risk, VaR）系統為資本適足要求的工具，因此涉險值有很大的潛力成為下一代的保險資本適足要求工具，雖然尚未施行。由於保險監理的重要性以及RBC和VaR在其中扮演重要的角色，兩者相對上的精確性是我們所感興趣的。 本篇論文的目的是實際去比較RBC及VaR在破產預測上的相對精確性。我們以美國1995到1998年產險公司的實際清償記錄，用型1及型2錯誤檢視RBC及VaR的破產預測能力。RBC的數據直接從產險公司報給NAIC的年報上就可取得，而VaR的數據來自於我們所建立的現金流量模擬模型。既然RBC的數據是實際的數據，而VaR的估計值也是基於公司實際的財務數據而來，我們能以實例展現VaR相較於RBC的財務預警能力。 我們的結果顯示RBC沒有任何財務預警能力，換句話說，沒有一個破產公司的RBC值小於0.7（監理機關可以根據這個值關掉公司）。另一方面，VaR有較好的財務預警能力，但是它同時也會使許多財務健全的公司必須接受許多沒有必要的檢查。我們VaR模型的整體正確分類能力只比隨意分類稍微好一些。 雖然結果並不如原先預期的好，我們仍然對VaR成為保險監理工具抱持樂觀的態度，因為它是目前為止最嚴密也最先進的風險管理工具。我們認為這些結果可以藉由修正不適當的假設後獲得改善，未來研究可以先朝這個方向努力。 / Assuring insurance company solvency has always been the focal point of insurance regulation. Among the employed solvency regulation methods, RBC represents the currently state-of-the-art capital adequacy requirement. Bank regulators already advocated the use of VaR systems in capital adequacy requirements. Value at risk thus has great potential to be the next-generation capital adequacy regulation, although not implemented yet. Because of the importance of solvency regulation as well as the key role played in that regulation by RBC and VaR, the relative accuracy of RBC and VaR is of great interest. The purpose of this research is to empirically compare the relative effectiveness of RBC and VaR in predicting insolvency. Through the solvency record of property-casualty insurers in the United States from 1995 to 1998, we examine the Type I and Type II error of VaR and RBC in predicting insolvency. The RBC figures are readily available from the annual statement since 1994 and the VaR values come from a simulation model that we build up. Since the RBC figures are the “real” numbers and the VaR estimates also base on the companies’ real financial positions, our research will demonstrate how VaR is compared to RBC in early warning for real cases. Our result shows that RBC doesn’t have any prediction power. In other words, none of the bankrupt insurers has a RBC ratio lesser than 0.7, the threshold according to which the regulator can seize the company. On the other hand, VaR has good early warning ability, but also leads the regulator to take too much unnecessary actions on solvent companies. The overall ability of correct classification of our model is just a little stronger than arbitrary classification. Although our results are not as good as we expect, we are still optimistic about the use of VaR, the currently most comprehensive and advanced approach of risk management, as an insurance solvency regulation tool. We attribute the unsatisfactory outcome to some assumptions that may be inappropriate. Further researches can move toward this aspect.

過去的文獻中，資產配置的方法不外乎效率前緣、動態資產配置等方式，但是，單獨針對銀行探討的文章並不多見，所以本文的貢獻在於單獨針對銀行的資產配置行為進行研究，希望能利用『演化策略演算法』，進行『模擬最佳化』來解決銀行資產配置的問題。基本上這個方法是由兩個動作結合而成，先是模擬，再來尋求最佳解。所以，資產面我們選擇了現金、債券、股票、不動產四項標的，而負債面則模擬了定存、活存與借入款這三項業務，然後透過重複執行模型的方式來求出最適解。並與單期資產配置方法下的結果作一比較，發現運用演化策略演算法有較佳的結果，此外，在不同的亂數下，仍具有良好的穩健性，可作為一般銀行經理人參考之用。 / We focus on the bank’s asset allocation problem in this thesis. We use simulation optimization to solve the problem by evolution strategy, which is relatively new in the financial field. Simulation optimization consists of two steps: simulate numerous situations and search for the optimal asset portfolios. In the simulation, we set up four assets, including cash, bond, stock, and real estate and three business lines, including demand deposits, time deposits, and borrowings. Then we search for the optimal solution by running the ES algorithm. The results show that simulation optimization generates better results than one-period asset allocation. Furthermore, the evolution strategy method generates similar results using different random numbers.

本文考量於Excel介面下設計一「壽險公司現金流量模型」，透過保險財管、精算理論的採用與大量隨機模擬亂數的應用，欲建構一結合理論基礎與實務運用的動態財務分析系統雛形。 模型中，資產面的模擬項目共有七項：1.債券與放款：採用CIR或Vesicek兩利率模型供選擇進行利率期間結構生成，以模擬出各到期期限的債券及放款價格。2.股票：以資本資產訂價模型(CAPM)來模擬各類股股票價格的變動與股票投資報酬。3.不動產：使用幾何布朗運動模擬不動產價值與租金收入。4.國外投資：利用幾何布朗運動模擬匯率的變動。5.現金及銀行存款。6.應收款項，考量壞帳情況下，逐年比率攤回殘餘金額。7.其他資產。 負債面採用定期險、終身生死合險與遞延年金險模擬壽險公司業務經營的現金流量情況。藉由資產與負債的整合，可模擬出公司未來十年內各年度的損益情況，讓使用者了解於承受總體經濟各項不確定風險下，壽險公司資產面、負債面與業主權益的現金流量情況。 文末引用個案範例，進行實務操作的說明，示範如何應用本模型來進行最適資產配置決策與敏感度分析，以證明本系統的合理可行性。最後，並對此系統提出檢討與展望，期待後續研究可加入程式語言的應用而建構出一完備的動態財務分析系統。 / The main purpose of this study is to construct a dynamic cash flow testing for the life insurance company by using Excel. Through the adoption of financial and actuarial theories and the application of stochastic method, we want to provide a rudiment analysis framework of life dynamic financial model that combines theoretical basis and practical application. This analysis framework includes seven categories of assets. The simulation models or related issues for each category will be discussed accordingly. – 1. Bonds and mortgage loans: providing CIR and Vesicek interest rate model for users to generate the interest term structure. 2. Stocks: applying CAPM method to simulate the stock prices and stock returns. 3. Real estate and rental income: using Geometric Brownian Motion to simulate the price of real estate and the rental income. 4. Foreign investment assets: using Geometric Brownian Motion to simulate the movement of exchange rate. 5. Cash and Deposits. 6. Account Receivable: after considering bad loans, we amortize the residual account receivables for a specific period. On the liability side, we use three types of products - term life, whole life endowment, and deferred annuity - to generate the business profile as well as the cash flows patterns of the life insurance company. By integrating the asset and liability sides of the model, we can simulate the revenue of the company for the following ten years and enable the users to predict the future cash flows under uncertain financial conditions. Finally, applications of this model are presented as thoroughly as possible to educate the users about how to make the optimal asset allocation decisions and sensitive scenario analysis. The application results show that the model reasonably fits the desired results. Since the model presented here is not a complete DFA model, future researches may consider adding more refined component into the analysis framework like using programming language.

保險公司清償能力一直是保險監理的重心，在所有現行的制度中風險基礎資本是最重要的，但此項制度仍有其缺點，因此其他動態分析模型被許多學者所提出，如涉險值及情境分析。雖然這些動態分析模型被學者所偏好，但監理機關仍須對這些模型的精確程度加以了解，這也是本篇論文所要研究的目的。 基於此，本篇論文以模擬方式及經濟模型加以分析風險基礎資本、情境分析及涉險值等方法的破產預測的相對精確性。其中風險基礎資本完全採用現有NAIC的年報資料，情境分析及涉險值則採用我們所建立的模型，基於此也可以確認現有監理制度是否有缺失。 我們的結果發現風險基礎資本的預測能力很低，動態模型-情境分析及涉險值皆優於風險基礎資本，且在不同動態模型中涉險值的預測能力較好。因此可知被學者所偏好的動態分析模型應是未來保險監理的方向希望藉由本篇提供監理機關一個參考的依據。 / Solvency prediction of insurers has been the focus of insurance regulation. Among the solvency regulation systems, risked-based capital (RBC) is the most important but RBC still has some drawbacks. Thus, the dynamic financial analyses-scenario analysis and Value at Risk have been developed to be the regulation tool. Although, the scholars prefer the dynamic financial analysis, the regulators still want to make sure the accuracy of dynamic financial analysis. That is the purpose of our paper. Therefore, we use the simulation result and the econometric model to analyze the relative effectiveness of RBC, scenario and Value at Risk (VaR). The RBC is from the annual statement and the scenario and VaR come from our simulation model. Our result shows that the RBC has very low explanatory power, the dynamic financial analysis is better than RBC, and VaR outperform scenario analysis. Thus, we conclude that VaR is the way to go for property-casualty insurance regulators.