ARCH and Bilinearity as Competing Models for Nonlinear Dependence.

This paper consider whether the wide acceptance of ARCH models may be at the expense of other nonlinear processes, such as bilinear models. The authors first pose a joint test for ARCH and bilinearity. A nonnested test is then suggested. The tests are then applied to three series. When GARCH models are taken as the null hypothesis, the authors fail to reject it. However, when bilinearity is taken as the null, it is rejected in two cases. Also, an out-of-sample forecasting exercise shows that the GARCH model is superior. The results, therefore, indicate a strong preference for the GARCH model.