Glad to share that the paper “Dynamic partial correlation models” I have co-authored with André Lucas (Vrije Universiteit Amsterdam and Tinbergen Institute, The Netherlands) has been published on the Journal of Econometrics.
You can read the paper in full open access here.
In the paper we introduce a novel scalable framework for dynamic conditional correlation matrices. We base our approach on a recursive structure of dynamic bivariate partial correlation models.
By leveraging this recursive structure and the principles of perturbed stochastic recurrence equations, we demonstrate the stationarity, ergodicity, and filter invertibility of the multivariate system using conditions derived solely from bivariate slices of the data. Additionally, we establish the consistency and asymptotic normality of the maximum likelihood estimator for the model’s static parameters.
In both simulation studies and empirical asset pricing applications to US stock returns, our proposed model outperforms benchmarks such as -cDCC and multivariate -GAS, in both simulation scenarios and in-sample and out-of-sample asset pricing analyses of US stock returns.
Please feel free to share your thoughts in the comments below, or drop a line here.
To cite: Enzo D’Innocenzo, Andre Lucas (2024) Dynamic partial correlation models, Journal of Econometrics, Volume 241, Issue 2, 105747, ISSN 0304-4076. DOI: 10.1016/j.jeconom.2024.105747