Stock market networks: The dynamic conditional correlation approach

We demonstrate the economic relevance of minimum spanning trees (MSTs) constructed from dynamic conditional correlations (DCC) for a sample of S&P 100 constituents. An empirical comparison of MST properties shows that using the standard approach of rolling (or sliding-window) correlations yields trees that are more robust, have higher densities and exhibit higher industry clustering than MSTs based on DCC. Our results suggest that these properties are achieved at the expense of the smoothing of market dynamics, which is better preserved by DCC. The DCC approach offers a new perspective for the analysis of complex systems such as stock markets.

► We explore stock market networks using minimum spanning trees on S&P 100 stocks.
► Dynamic conditional correlations are suggested instead of usual rolling correlations.
► Rolling correlations lead to more robust, dense trees with higher industry clustering.
► Dynamic conditional correlations do better in preserving market dynamics.
► Both methods have limitations; their use should match particular application.