Dynamic Structural Change of Financial Intermediation and Restructuring of Business Models in Japanese Banking

在本研究中，我的总体研究目标是分析日本银行业风险-回报优化与冒险决策之间的关系。自马科维茨创立现代投资组合理论以来，定量金融中的风险分析已广泛涉及资本市场的市场风险。在综合风险管理的背景下，美国和欧洲已经推进了对银行业各种风险的延伸，但日本银行业仍然是一个挑战。然而，为了提高风险收益状况，日本银行有必要实施基于风险的政策和实践，从而利用定量方法充分建模和衡量银行风险。我集中讨论了市场风险的建模和度量问题。“风险”一词在金融中有两种不同但相关的用法：投资组合潜在回报的标准差的大小，或者在一段时间内潜在的更多损失。定量金融中的风险模型使用随机方法分析风险。首先，我研究了均值方差分析，其中投资组合风险通过使用协方差矩阵进行统计计算，其中波动性和相关性是风险的两个关键决定因素。从理论上讲，均值方差分析假设股票收益呈正态分布，投资者行为理性。然而，事实上，股票收益的直方图表现出过度的峰值和肥尾分布，而不是正态分布，这被称为程式化事实。在实现优化的过程中需要考虑到这些特性。在实现随机均值方差优化的过程中，我使用蒙特卡罗模拟作为优化器。通过检查蒙特卡罗模拟的预测结果，我扩展到风险的第二种观点：潜在损失的大小，其中风险在分布的尾部进行评估，例如风险价值的概念。目前，在银行业综合风险管理的背景下，将VaR框架作为巴塞尔协议的风险度量方法。少

In this study, my overall research objective was to analyze the relationship between risk-reward optimization and risk-taking decisions in Japanese banking. Since the emergence of modern portfolio theory originated with Markowitz, risk analysis in quantitative finance has extensively addressed market risk in the capital markets. In the context of integrated risk management, the extension to the various risks in banking has advanced in the United States and Europe, but it still remains a challenge in the Japanese banking industry. However, in order to enhance the risk-return profile, it is necessary for Japanese banks to implement risk-based policies and practices, and thus, to adequately model and measure banking risk utilizing a quantitative approach.I concentrated on the issue of modeling and measuring market risk. The term ‘risk' is used in finance in two different but related ways : as the magnitude of the standard deviation of the potential return of investment portfolio, or the p … More otential loss over some period of time. Risk models in quantitative finance analyze risk using stochastic approaches.First, I studied mean-variance analysis, in which portfolio risk is statistically calculated by using a covariance matrix, in which volatilities and correlations are the two key determinants of risk. Theoretically, mean-variance analysis postulates normally distributed stock returns and rational behavior of investors. In fact, histograms of stock returns, however, exhibit excess peakness and fat-tailed distributions rather than normal distributions, which are referred to as stylized facts. These features need to be accounted for in the process of implementing optimization.In the process of implementing stochastic mean-variance optimization, I used Monte Carlo simulation as an optimizer. By examining forecasting results of Monte Carlo simulation, I extended to the second view of risk : the magnitude of the potential loss, in which risk is evaluated in the tail of the distribution, such as the concept of value at risk. Now the VaR framework is used as the methodology of risk measurement of the Basel Accord in the context of integrated risk management in banking. Less