Latest research insights

Advancing stochastic dividend and stochastic volatility research for practitioners

Bridging academic rigour with practical implementation
We explore option pricing and risk management frameworks where both the underlying asset's volatility and its dividend yield follow stochastic processes. Our work moves from theoretical foundations to working code, helping practitioners implement models that reflect real market dynamics rather than convenient assumptions.
Affine dividend models with mean-reversion
Rough volatility extensions
Machine learning for model calibration
Joint stochastic process estimation

Rigorous methodology

Every model we publish is derived from first principles and validated against market data before we recommend it.

Working implementations

We provide reproducible code alongside theoretical results, so you can verify, adapt, and extend our methods.

Practitioner feedback

Our research agenda is shaped by conversations with trading desks and risk teams who test our models in production.

Hero Image

Research areas and methodologies

What we investigate

Our research programme centres on four interconnected areas. First, we develop affine dividend models that capture the term structure of equity income while preserving analytical tractability for option pricing. Second, we extend rough volatility frameworks to account for the persistent, non-Markovian behaviour observed in realised variance series. Third, we apply machine learning techniques to accelerate model calibration without sacrificing the interpretability that risk managers require. Finally, we advance joint estimation methods that treat dividend and volatility dynamics as correlated stochastic processes rather than independent inputs. Each project produces both peer-reviewed publications and reference implementations that practitioners can adapt to their own frameworks.
Researcher analysing stochastic model outputs on multiple screens

Published work

Papers, working notes, and model implementations from our faculty and community
Affine dividend term structures

Affine dividend term structures

A working paper deriving closed-form characteristic functions for option pricing under mean-reverting dividend yield processes with time-dependent parameters.
Python NumPy SciPy
Rough Heston calibration

Rough Heston calibration

An implementation study comparing Fourier-based and machine learning approaches to calibrating rough volatility models to equity index smiles.

Python TensorFlow Matplotlib
Neural networks for implied vol

Neural networks for implied vol

A methodological note on using feedforward networks to interpolate implied volatility surfaces while enforcing no-arbitrage constraints.

PyTorch Pandas Plotly
Correlated dividend-vol dynamics

Correlated dividend-vol dynamics

A simulation framework for pricing path-dependent exotics under joint stochastic dividend and volatility models with realistic correlation structures.
C++ Python OpenMP
Variance reduction for exotics

Variance reduction for exotics

A comparative study of antithetic variates, control variates, and importance sampling for barrier and Asian options under stochastic volatility.
Python NumPy Jupyter
Model risk in dividend forecasts

Model risk in dividend forecasts

An analytical review of how deterministic dividend assumptions distort value-at-risk calculations for equity derivative portfolios.

R Python LaTeX

Research updates

Recent breakthroughs, conference presentations, and new methodological developments in stochastic modelling for dividends and volatility
Researcher presenting rough volatility findings
Conference March 2026

Rough volatility at Quant Congress Asia

Our lead volatility researcher presented new findings on rough Heston calibration stability at the annual quantitative finance conference in Singapore. The session examined how fractional Brownian motion parameters interact with dividend yield dynamics during stressed market periods, drawing on two years of index option data.

Benchmark charts comparing calibration methods
Publication January 2026

Machine learning calibration benchmark

We released a comprehensive benchmark comparing neural network surrogates against traditional optimisation for stochastic model calibration. The study includes reproducible code and evaluates speed-accuracy trade-offs across five equity index datasets, offering practitioners guidance on when machine learning adds value versus when it introduces unnecessary opacity.