Abstract: The research on the correlation between financial markets has been valued by scholars. The occurrence of extreme events such as financial crisis increases the tail correlation between markets. The TailCoR model is a new method for measuring the tail correlation, which decomposes the correlation into linear and nonlinear components and performs well in small samples. In order to describe the dynamic correlation between financial markets, we propose dynamic TailCoR model, and decompose the correlation into dynamic linear and nonlinear correlation coefficients based on the dynamic TailCoR model. The dynamic linear correlation coefficient is estimated based on DCC-GARCH. The remainder of the dynamic TailCoR divided by the dynamic linear correlation coefficient is defined as the dynamic nonlinear correlation coefficient. Finally, based on the dynamic TailCoR model, the tail correlation of the stock returns of the four large domestic banks is studied. It is found that the tail correlations increase significantly in two periods, namely, in 2008 and from 2015 to early 2016, and the tail correlation rise is mainly caused by the nonlinear correlation. The research methods proposed in this paper give dynamic nonlinear correlation metrics, which can be well applied in investment and risk measurement.

Key words: tail correlation, nonlinear correlation, TailCoR model, banks