IRGA 2021 Tragem

Mutual funds are portfolios of financial assets managed by banks or asset management companies on behalf of end investors seeking to invest their savings. The value of assets under management in mutual funds worldwide was 54,900 billion USD at the end of 2019, including 26,700 billion USD in the United States and 2,200 billion USD in France. These funds have specific investment objectives (for example, investing in the stocks of small, socially responsible companies). The proposed project aims to

• protect mutual fund investors by enabling them to verify that the funds’ management objectives are being met; this objective will be achieved, in particular, by significantly increasing the frequency with which investors are informed about assets under management compared to what is required by regulation
• improve the functioning and transparency of financial markets by reducing manipulation by mutual fund managers

 
The project aligns with the themes of CERAG’s “Risk Anticipation and Management” research area and is part of the campus’s Idex initiative, building on the CDP RISK research project, which concludes in 2021. The project is interdisciplinary, bringing together faculty members in the fields of Management Sciences (finance) and Information and Automation Sciences (expertise in optimal control and observation of complex dynamic systems from the GIPSA-lab). The funding request is for a collaborative doctoral thesis co-supervised by Isabelle GIRERD-POTIN of CERAG and Didier GEORGES of GIPSA-lab. The research will also benefit from the scientific support of Ollivier TARAMASCO, who has expertise in both Applied Mathematics and Finance. The successful candidate must possess a solid background in finance and either control theory or applied mathematics.

Mutual funds are subject to varying regulations depending on the country. They are required to disclose information on their portfolio composition at varying intervals (quarterly in the United States, semi-annually in the European Union). Disclosing fund compositions at infrequent intervals provides investors with imperfect and manipulable information and can undermine the stability of the financial system as a whole, thereby leading to a significant increase in systemic risks.
Using minimal, non-manipulable data—specifically, the daily returns of mutual funds combined with the returns of assets within the investment universe—we will develop algorithmic methods to reconstruct a portfolio’s daily composition and the decisions regarding changes to that composition over time. This will make it possible to verify the accuracy of the information provided to investors (does the portfolio consistently possess the characteristics reported by the manager?) and prevent manipulation by mutual fund managers on the dates when portfolio compositions must be disclosed (such as “window dressing” or “pumping”). 
 
This type of problem falls into the category of large-scale inverse problems, which involve reconstructing parameters, states, behaviors, or decisions from a large volume of data. Staying within this category of problems, based on the daily knowledge of fund composition established in an initial step using only the funds’ returns, we will proceed by seeking to identify the type of rationality that fund managers exhibit in their decisions. In particular, this will involve identifying their risk aversion or risk appetite parameters and determining which return-risk models and risk measures are implicitly optimized by the chosen portfolio.
 
The use in finance of tools derived from computer science and applied mathematics (notably optimal or asymptotic observers and classification techniques) constitutes an original interdisciplinary approach: the research literature—emerging but still limited—shows that our project addresses a timely issue in the context of financial risk mitigation.
 
To carry out this project, we have access to the Refinitiv Eikon database, to which CERAG subscribes, and to the high-performance computing infrastructure at the GRICAD site. In addition, we explored this topic as part of a master’s internship in 2020.