Axa Research Fund for COVID-19: Launch of Gilles Stupfler’s Project
Gilles Stupfler, Lecturer in Statistics at ENSAI and researcher at CREST was awarded a 2-year grant by Axa Research Fund. Now is the time to launch the project with newly arrived post-doc research assistant Boutheina Nemouchi. Interview.
AXA Research Fund’s Covid-19 flash call received over 500 applications from top-tier universities and research institutions worldwide.
“An Extreme Value Model for the Analysis of the COVID-19 Pandemic and its Impact, and the Mitigation of Future Related Crises”, Gilles Stupfler’s project, was among the ten innovative projects selected. All of them aim to produce a “science-based response to COVID-19, with a focus on vulnerable populations, mental health and crisis monitoring”.
Gilles, how does your specific area of research apply in a pandemic context?
Gilles Stupfler: My research applies to that context because it more generally applies to getting an understanding of the scale and consequences of extreme events that we very rarely see, such as extreme climate events or financial crises. What makes this particular event (the COVID-19 pandemic) particularly challenging is that it is still unfolding, and so its consequences are not fully understood. Managing its impact and using the available information effectively is therefore very difficult.
How would you present the project “An Extreme Value Model for the Analysis of the COVID-19 Pandemic and its Impact, and the Mitigation of Future Related Crises” in a few sentences?
G.S. : My project will develop adapted techniques for modelling the impact of risks related to highly disruptive events, such as the COVID-19 pandemic, in the absence of relevant data on such rare events. The available statistical techniques for analyzing extreme events in a non-stationary, time-dependent world remain scarce; this project addresses this problem by following two strands of work that push the boundaries of current knowledge about extreme value analysis and risk assessment.
Could you elaborate on these two strands of work?
G.S. : The first strand will involve modelling the large economic and financial consequences of the global spread of COVID-19, given auxiliary, geographical, sanitary and climate information. I plan to do this in what is called, in statistics, a non-stationary regression framework. This will be done using various risk measures, such as expectile and extremile-based risk measures. These provide sensible alternatives to the more traditional quantile-based risk measures that would be used in financial and actuarial practice, such as the Value at Risk and the expected shortfall.
The second strand of work will focus on the case when the data are functions, for example of time, to construct estimators that will allow one to analyze and possibly to predict the aftermath of the COVID-19 pandemic. These estimators will be called non-parametric estimators, a statistical term, in the sense that they progressively learn the structure of the data with minimal input from the user.
What do you expect the outcome of this project will be? What will you deliver?
G.S. : What I expect from the project is that the above two strands of work will provide a diverse toolbox for the assessment and mitigation of extreme risk due to rare events such as the COVID-19 pandemic or a disruptive climate event.
What I hope is that this will give new insight into the amount of risk containing contained in rare events with a geographical or temporal spread.
At a more technical level, the math that will be developed during the project shall draw from regression methods, time series, functional data analysis, and I expect that extreme value specialists will be interested in taking all these kinds of results from all of these kinds of fields to expand extreme value analysis further.
The methods will be implemented and simulated on real data from a variety of fields: climatology, epidemiology of course, insurance and finance. With the support of my colleagues, I’ll strive to include the methods in open-source software, including, if this makes sense, by updating an R package called ExtremeRisks that Simone A. Padoan (Bocconi University) and I have worked on during a different project on extreme risk assessment. I hope that practitioners in risk management will be interested in using these tools as complements to the standard industry tools they routinely use, such as the Value at Risk and Expected Shortfall.
You were awarded a 2-year grant for €100,000, what expenses will it cover?
G.S. : This covers the full-time salary of a two-year postdoctoral research assistant, plus computing equipment. ENSAI and the CREST lab have accepted to chip in extra support for travel and other academic activities, so I’m grateful to them as well!
Boutheina Nemouchi just joined ENSAI as a post-doctoral fellow to work on this project with you. What will be her missions? Who else will work on the project?
G.S. : I am indeed very happy to welcome Boutheina Nemouchi. Boutheina will work with me as a postdoctoral research assistant for two years. She holds an MSc degree from the University of Constantine 1 (Algeria) and a PhD degree from the University of Compiègne (advisor: Salim Bouzebda).
Boutheina’s PhD project focused on conditional empirical processes and U-processes; given that the project will for the most part focus on conditional extreme value analysis, I expect that Boutheina will strongly contribute to the theoretical aspects of the project. Boutheina will also help in disseminating our research, for instance by attending seminars and conferences where she will talk about our work.
My other main collaborators on this specific project will be, first, Abdelaati Daouia (Toulouse School of Economics), with whom I have an established research collaboration. His expertise in conditional extreme value analysis and generally regression models will be very welcome!
Second, we should also get a helping hand from Simone A. Padoan (Bocconi University), whose expertise in computational aspects will really help for the release of the procedures developed in the project in an open-source R package. I already have had the opportunity to work with Simone on the package ExtremeRisks I was mentioning before, he has been fantastic!
Find out more about Gilles Stupfler’s project and research at ENSAI.
Photo: Axa Research Fund