The Chinese Correction of February 2007: How financial hierarchies change in a market crash

We analyzed 546 stocks in the Singapore Stock Exchange (SGX) and 1173 stocks in the Hong Kong Stock Exchange (HKSE) in 2006 and 2007, to understand how financial hierarchies on these two markets change over market corrections and crashes. To do so, we introduced the digital cross correlation as a measure of the comovement tendencies between stock pairs, and also the method of partial hierarchical clustering to iteratively identify strongly-correlated clusters of stocks. From daily prices over the 2006–2007 period, we found the existence of clusters of local stocks as well as clusters of Chinese stocks traded on the two markets. We further discovered the Chinese clusters organizing into a Chinese supercluster, interacting less strongly with a supercluster dominated by local clusters. Going down to 30-minute prices within two-month overlapping time windows over 2006 and 2007, we found dips in the number of clusters before market corrections and crashes, followed by peaks in the number of clusters afterwards. On the SGX, we also found the stronger intra cluster correlation weakening, and the weaker inter cluster correlation strengthening before the February 2007 Chinese Correction. These features are in qualitative agreement with a chemical reactions picture in which clusters of stocks ‘react’ to form large superclusters of stocks that ‘dissociate’ during market crashes. Finally, on the SGX we found broad humps in the intra cluster and inter cluster correlations for the May/June 2006 market correction and the February 2007 Chinese Correction, but a sharp peak for the July 2007 Subprime Crisis. This suggests that the earlier events were endogeneous to the SGX, while the latter event was an exogeneous shock.