Are correlations constant over time? Application of the CC-TRIGt-test to return series from different asset classes

A new test for constant correlation is proposed. Based on the bivariate Student-t distribution, this test is derived as Lagrange multiplier (LM) test. Whereas most of the traditional tests (e.g. Jennrich, 1970, Tang, 1995 and Goetzmann, Li & Rouwenhorst, 2005) specify the unknown correlations as piecewise constant, our model-setup for the correlation coefficient is based on trigonometric functions. Applying this test to assets from different financial markets (stocks, exchange rates, metals) there is empirical evidence that many of the correlations vary over time.