Anouk Rensen, a Master’s student from Erasmus School of Economics, has gained the second prize in the CFA Quant Awards, a competition celebrating outstanding research in quantitative finance by students. Rensen earned this distinction for her bachelor thesis, titled “Dynamic Clustering in Multi-Factor Copulas with Hidden Markov Models”.
Modeling time-varying dependence in high-dimensional financial markets
Anouk’s thesis (supervised by Assistant Professor Anastasija Tetereva) is about improving how dependence between many financial assets is modeled by allowing groups of firms to change over time rather than remaining fixed. Traditional multi-factor copula models reduce complexity by clustering firms, but these clusters are usually static, based on industry classifications or a one-time data-driven method such as k-means. This is problematic because firms evolve through changes in strategy, mergers, acquisitions, or shifts in market focus, which can alter how their returns co-move with others.
To address this, Anouk Rensen proposes a dynamic multi-factor copula model in which firm cluster memberships are time-varying and inferred from the data. Cluster transitions are governed by a hidden Markov model, where firms are more likely to move between similar clusters and where past membership probabilities influence future assignments. This framework allows the dependence structure to adapt over time while maintaining persistence in group membership, avoiding unrealistic, frequent reshuffling.
The model is applied to daily returns of S&P 100 stocks from 2015 to 2024 and is compared to two common static benchmarks: industry-based clusters and static k-means clusters. Both in-sample and out-of-sample results show that dynamic clustering consistently delivers better predictive performance, including improved forecasts of joint return behavior. Importantly, the paper finds that a model with fewer dynamic clusters can outperform a model with more static clusters, highlighting the efficiency gains from allowing groups to evolve over time. The improvements are strongest during periods of market stress, high volatility, and elevated dispersion in returns, suggesting that dynamic clustering is particularly valuable when dependence structures are most unstable.
About the CFA Quant Awards
The Quant Awards is an initiative led by several CFA Societies, local chapters of the CFA Institute (a global non-profit association of investment professionals). Entries are judged on their applicability, relevance, innovation, accuracy, completeness, and presentation. Winners receive a financial prize: €1,000 for third place, €2,000 for second, and €3,000 for first. An official award ceremony is set to take place in Brussels later this year.
Downloads
- More information
You can download the thesis “Dynamic Clustering in Multi-Factor Copulas with Hidden Markov Models” above.
For more information, please contact Ronald de Groot, Media & Public Relations Officer at Erasmus School of Economics, rdegroot@ese.eur.nl, mobile: +31 6 53 641 846.
