Financial Engineering Workshops - Giuseppe Brandi (Northeastern University - London)

Abstract: Financial markets are paradigmatic complex systems, in which the collective behaviour of heterogeneous interacting agents gives rise to emergent statistical regularities that transcend simple diffusion. Among these, multiscaling stands out as a particularly informative signature: a process is multiscaling when its moments scale with non-linear, moment-dependent exponents, a property linked to intermittency, heterogeneous volatility and long-range dependence. Multiscaling can, however, arise from fat-tailed distributions, from temporal dependencies, or from both, and disentangling these sources is essential before interpreting it as evidence of genuine dynamical complexity. In this talk, we address this question for the rough Bergomi model, in which fractional Brownian motion with a small Hurst parameter H generates strikingly irregular volatility paths and multiscaling that strengthens as H approaches zero. We propose a two-stage testing procedure based on the generalised Hurst exponent. We first test for the presence of multiscaling against matched fractional Brownian motion surrogates. We then identify its source using shuffled surrogates, which preserve the return distribution while destroying temporal correlations, combined with distance-based permutation tests that are robust to asymmetric null distributions. After validating the procedure on the Multifractal Random Walk and Fractional Lévy Stable Motion, we find that multiscaling in the very rough regime is driven predominantly by fat-tailed returns. At moderate roughness, as tails thin and volatility clustering emerges, temporal dependencies become the main contribution. These results draw a line between genuine path-dependent complexity and distributional effects, with implications for financial modelling, derivative pricing and risk management.

Bio: Giuseppe Brandi is an Assistant Professor in Data Science at Northeastern University London, where he is Programme Lead for the MSc AI and Business Innovation dual degree and teaches courses in Machine Learning and Data Science. His research spans complex networks, data science, and quantitative finance, with a methodological emphasis on complex systems and econometrics. He develops and applies statistical frameworks, including multiscaling analysis, tensor factorisation, network-based measures of financial interconnectedness, and machine learning, to study the behaviour and complexity of financial and economic systems. Extending this toolkit to climate risk, his more recent work brings quantitative finance and econometric methods to bear on the modelling of climate-related financial exposures. He maintains international research collaborations and works with industry partners including Jupiter Asset Management.

Faculty Events | Website | View More Events