Discussion Papers
Findings from research and studies conducted at the Financial Research Center (FSA Institute) are organized and published as Discussion Papers to stimulate further discussion and comment. Please send comments to: frtc_comments★fsa.go.jp (replace the star sign with @).
The views expressed in the papers are those of the authors and do not necessarily reflect the views of the Financial Services Agency or the FSA Institute.
Discussion Papers (FY 2024)
| No. | Title/Author(s) |
|---|---|
DP2024-5 (January 2025) |
The Significance of “Enterprise Value Mortgage Lien” and Its Standing in the Bankruptcy Act TOMIKAWA Ryo and YASUNAGA Yuji Abstract |  Full text (PDF:1,067KB)(Full text is available only in Japanese) |
DP2024-4 (January 2025) |
A Study on the Current State of Impact Measurement and Management (IMM) and Future Directions from the Perspective of Management Accounting HAYASHI Toshikazu and KOZAKI Aiko Abstract | Full text (PDF:1,749KB) |
DP2024-3 (January 2025) |
Progress in the Evaluation Methodologies of Large-scale Language Models (LLM) in the Financial Sector and Initiatives toward Improving LLM through Retrieval Augmented Generation KIM Kangsoo and MURATA Ken Abstract | Full text (PDF:2,354KB)(Full text is available only in Japanese) |
| DP2024-2 (July 2024) |
Regional Financial Markets during Covid-19 Pandemic: Concentration, Inter-regional Flow of Funds via Loans and Deposits, and Credit Guaranteed Loans UESUGI Iichiro, HIRAGA Kazuki, MANABE Masashi, YOSHINO Naoyuki Abstract | Full text (PDF:1,773KB)(Full text is available only in Japanese) |
| DP2024-1 (June 2024) |
Study on the Business Practices for Lending Collateralized against Enterprise Value at U.S. and U.K. Commercial Banks KAWAHASHI Hitomi Abstract | Full text (PDF:1,908KB)(Full text is available only in Japanese) |
Abstract
DP2024-5
The Significance of “Enterprise Value Mortgage Lien” and Its Standing in the Bankruptcy Act
TOMIKAWA Ryo, Associate Research Fellow, Financial Research Center (FSA Institute)YASUNAGA Yuji, Associate Research Fellow, Financial Research Center (FSA Institute)
The Act on the Promotion of Cash Flow-based Lending was enacted on June 7, 2024. This act has received considerable attention from the public, as it introduced a new collateral scheme, namely, “enterprise value mortgage lien,” which enables a borrower entity to use all of its assets, including both tangible and intangible assets, as collateral for loans, and the act also stipulated special foreclosure procedures to implement this mortgage lien. This study first examines what kind of messages this new act has conveyed to the financial market (particularly financial institutions). Then, it discusses the significance of this new type of collateral for mortgages, considerations for its use, and the treatment of the new collateral scheme in the Bankruptcy Act.
Keywords: The Act on the Promotion of Cash Flow-based Lending, cash flow-based lending, and “enterprise value mortgage lien.”
DP2024-4
A Study on the Current State of Impact Measurement and Management (IMM)
and Future Directions from the Perspective of Management Accounting
HAYASHI Toshikazu, Special Research Fellow, Financial Research Center (FSA Institute)KOZAKI Aiko, Special Research Fellow, Financial Research Center (FSA Institute)
This paper focuses on the increasingly prominent topic of Impact Measurement and Management (IMM), particularly on the setting and utilization of impact-related indicators. Based on interviews with practitioners in Japan and abroad, it describes the current state of IMM practices among companies and investors. The key findings are as follows. First, IMM consists of two distinct layers: “IMM at the firm level by individual companies” and “IMM at the portfolio level by investors.” Second, the objectives and benefits of measuring impact-related indicators within companies fall into two broad categories: engagement and accountability to investors and other stakeholders (external reporting purposes); and business progress management, planning, and improvement (internal use purposes). Third, the indicators effective for external reporting to investors and other stakeholders do not always align with those useful for internal use. Fourth, there is some overlap between the indicators used for IMM and those used for Commercial Performance Measurement and Management (CPMM). And fifth, companies tend to prefer baseline comparisons—comparing outcomes against levels in the absence of intervention—over competitor benchmarks for impact-related indicators. Additionally, by drawing on insights from management accounting and management control research, this paper explores future directions for the practice and study of IMM at the firm level by individual companies. The discussions also highlight the potential for connecting IMM research with the fields of management accounting and management control.
Keywords: Impact Measurement and Management, IMM, Impact Investing, Management Accounting, Management Control.
DP2024-3
Progress in the Evaluation Methodologies of Large-scale Language Models (LLM) in the Financial Sector and Initiatives toward Improving LLM through Retrieval Augmented Generation
KIM Kangsoo, Special Research Fellow, Financial Research Center (FSA Institute)MURATA Ken, Special Research Fellow, Financial Research Center (FSA Institute)
Since the emergence of ChatGPT in 2022, focus on generative artificial intelligence (AI), particularly large-scale language models (LLM), has been increasing. Financial institutions are also considering the use of these technologies to improve operational efficiency and customer service. However, LLMs are probabilistic models, and their output accuracy is not necessarily guaranteed. If the training data itself contains factual inaccuracies or biases, there is a risk that the model may generate seemingly plausible text grounded on incorrect information or prejudice. Moreover, concerns arise from the model’s opaque internal reasoning process and its limited explainability. For financial institutions, where cultivating trust with customers is essential, these issues represent challenges that must be prudently examined when adopting such technologies.
Financial regulators have also expressed great interest in the potential opportunities and risks that generative AI/LLM pose for the financial sector. Considering rapid developments in AI, including generative AI, and the increasing use of AI in the financial sector, the Financial Stability Board (FSB) published a report in November 2024 on the implications of AI on financial stability, which is an update of the November 2017 report titled “AI and machine learning in financial services.” The recent report points out that AI has given rise to new use cases, such as document summarization, information retrieval, and code generation, and identifies potential AI-related vulnerabilities that could increase systemic risks, such as third-party dependencies, cybersecurity, model risk, data quality, and governance.
The purpose of this paper is to identify the use cases of LLM in the financial sector through a literature review and empirical analysis and clarify current issues and challenges, and to explore possible technical solutions to such challenges and the future development of LLM. In particular, it focuses on the latest research on LLM evaluation methodologies in the financial sector and the performance of Retrieval Augmented Generation (RAG), and initiatives to enhance its performance. When introducing LLM in the financial sector, careful consideration is especially needed in assessing model outputs. For language models used generally, evaluation is primarily based on performance measures such as accuracy, consistency of generated data, and frequency of hallucination, but for models used in the financial sector, industry-specific measures that are more granular and multifaceted are required. Interviews with experts in charge of AI at financial institutions and a survey of previous research revealed that decision-making and risk management in the financial sector, in particular, require more complex cognitive skills and that there are many areas that cannot be adequately covered by the existing LLM evaluation standards. With regard to the RAG system, which is drawing attention as an example of the use case of LLM in the financial sector, we conduct a technical overview of the system and identify the points that should be considered when implementing the system in financial institutions’ operation, and examine and verify how the system is assessed and how its performance is improved for practical use based on the FSA’s guidelines and related documents.
Keywords: large language model (LLM), RAG, explainability.
DP2024-2
Regional Financial Markets during Covid-19 Pandemic: Concentration, Inter-regional Flow of Funds via Loans and Deposits, and Credit Guaranteed Loans
UESUGI Iichiro, Associate Research Fellow, Financial Research Center (FSA Institute)HIRAGA Kazuki, Associate Research Fellow, Financial Research Center (FSA Institute)
MANABE Masashi, Associate Research Fellow, Financial Research Center (FSA Institute)
YOSHINO Naoyuki, Chief Advisor, Financial Research Center (FSA Institute)
The Covid-19 pandemic started in early 2020 and has had far-reaching effects on regional economies and financial markets in Japan. This study identifies how regional financial markets have changed during the Covid-19 pandemic by examining the degree of concentration in loan and deposit markets, and indicators of the inter-regional flow of funds through loans and deposits. The production of these indicators is a follow-up to our previous research project at the Financial Services Agency (Uesugi, Hiraga, Manabe, and Yoshino, 2021a, 2021b, 2022). In addition, by combining our indicators and other financial data, we study how financial institutions provide credit guaranteed loans (loans guaranteed by credit corporations) to small and medium enterprises (SMEs), and how the provision of such loans has changed during the pandemic. Our findings are as follows. First, the degree of concentration in loan and deposit markets continued to gradually increase even during the pandemic. For those prefectures where mergers of regional financial institutions took place, there was a substantial increase in concentration. Second, indicators of the inter-regional flow of funds via loans and deposits show that deposits collected in a region were more likely to be used for loans in the same region, and the ratio of deposits used for loans in other regions, including Tokyo, continued to decline. During the Covid-19 pandemic, the government provided various support measures to firms and households, including cash transfers, subsidies, and loans, which led to a surge in the outstanding balance of deposits in the country. However, of that increase in deposits, approximately JPY 40 trillion has been used in the form of loans in prefectures overall, while the remaining JPY 90 trillion has been used for investment other than loans. And third, we observe a few distinctive features for credit guaranteed loans to SMEs depending on where these loans are extended: the regions where banks’ headquarters are located or other regions. These notable features are: (i) financial institutions depend on credit guaranteed loans in regions where their headquarters are located more than they do in other regions, (ii) default rates for credit guaranteed loans in other regions are higher than those for loans extended in their headquarter regions, and (iii) during the pandemic, the higher dependency on credit guaranteed loans in headquarter regions continued to hold, while there is no notable difference in the default rates of guaranteed loans extended in headquarter regions and in regions outside of those headquarters.
Keywords: regional financial markets, Covid-19, lending in regions outside banks’ headquarters, credit guarantees.
DP2024-1
Study on the Business Practices for Lending Collateralized against Enterprise Value at U.S. and U.K. Commercial Banks
KAWAHASHI Hitomi, Special Research Fellow, Financial Research Center (FSA Institute)In this paper, we conducted interviews with senior-level relationship officers and credit underwriting officers at U.S. and U.K. commercial banks about their business practices for lending against enterprise value, referred to as blanket lien, in order to draw implications for the business practices of Japanese financial institutions as the country aims to introduce a new blanket lien type collateral scheme. The survey focuses on the following points: (i) the standing of lending against total enterprise value relative to other corporate loans, (ii) the business practices for enterprise and collateral valuations, (iii) the personnel systems of relationship officers, (iv) the roles of relationship officers, and (v) partnerships between relationship officers and credit underwriting officers. The paper concludes with the business implications that were drawn from U.S. and U.K. practices that are considered important in shaping the lending and credit underwriting practices of Japanese financial institutions when the new collateral scheme is introduced in Japan.
Keywords: lending against enterprise value, relationship officers, credit underwriting officers, cash flow, cash flow-based lending.
