PhD-candidate: Simon Donker van Heel
Start: Fall 2023
Many quantities we care about in economics and finance change continuously and cannot be observed directly. Market risk, the state of the business cycle, the pace of climate change. We only observe noisy data from which they must be inferred.
My research treats these moving quantities as time-varying parameters and develops filters that track them through time. Each time new data arrives, the filter updates its estimate with a gradient-based step, the same idea that trains machine learning models, adapted to the structure of economic and financial data. I work where time series econometrics meets optimisation, and I aim to make these filters more accurate and more stable than existing methods.
Some of the work is theoretical, for instance proving mathematically that a filter stays reliable even when the model behind it is wrong. Some is practical, with methods that give sensible risk forecasts and lower trading costs in portfolio management.
These questions matter wherever decisions depend on tracking change, from financial risk and economic forecasting to climate monitoring.
