Testing for jumps in GARCH models, a robust approach

Assuming that the observed return series consist of a standard normal ARMA-GARCH component plus an additive jump component, we propose a new test for additive jumps in an ARMA-GARCH context. The test is based on standardised returns, where the first two conditional moments are estimated in a robust way. Simulation results sug1gest that the test has very good finite sample properties, i.e. correct size and much higher proportion of correct jump detection than Franses and Ghijsels's (1999) test. We apply our test on the YEU-USD exchange rate and find twice as much jumps as Franses and Ghijsels's (1999) test.