Today I had the pleasure to present the paper “Dynamic Partial Correlation Models” – co-authored with Prof. André Lucas – at the 16th International Conference on Computational and Financial Econometrics (CFE 2022).
The conference program was very rich and all presentations were engaging and stimulating. We presented our paper (remotely) within the Dynamic Conditional Score Model session.
In our work we argue that – compared to other dynamic conditional correlation models – dynamic partial correlation model has several advantages:
- Unlike the matrix equations in Creal et al. (2011), Opschoor et al. (2018, 2021), and Hafner and Wang (2021), there is no complicated correlation constraint;
- Precise stationarity, ergodicity and invertibility conditions exist;
- The parameters can be estimated recursively for a given value of ν, therefore, our set-up is perfectly scalable to higher dimensions;
- The aymptotic theory of the MLE is available;
- The model works nicely in practice, in particular for beta hedging;
- In a controlled simulation setting we show that the new partial correlation model outperforms the considered benchmarks
Creal, D., Koopman, S. J., and Lucas, A. (2011). A dynamic multivariate heavy-tailed model for time-varying volatilities and correlations. Journal of Business & Economic Statistics, 29(4):552–563: https://www.tandfonline.com/doi/abs/10.1198/jbes.2011.10070
Opschoor, A., Janus, P., Lucas, A., and Van Dijk, D. (2018). New heavy models for fat-tailed realized covariances and returns. Journal of Business & Economic Statistics, 36(4):643–657:
Opschoor, A., Lucas, A., Barra, I., and Van Dijk, D. (2021). Closed-form multi-factor copula models with observation-driven dynamic factor loadings. Journal of Business & Economic Statistics, 39(4):1066–1079.
Hafner, C. M. and Wang, L. (2021). A dynamic conditional score model for the log correlation matrix. Journal of Econometrics. https://www.sciencedirect.com/science/article/pii/S0304407621002153