Tests for Symmetric and Asymmetric Nonlinear Mean Reversion in Real Exchange Rates
提出基于平滑转换自回归模型的新检验方法，用于检测实际汇率时间序列的均值回归，并区分对称与不对称情况。对17种兑美元和14种兑德国马克的实际汇率月度数据检验发现，比传统迪基-富勒检验更有力地拒绝单位根原假设。
New tests, based on smooth transition autoregressive models, for mean reversion in time series of real exchange rates are proposed. One test forces mean reversion to be symmetric about the integrated process central case, while the other permits asymmetry. The tests are applied to monthly series of seventeen real exchange rates against the U.S. dollar and fourteen against the deutsche mark. They reveal stronger evidence against the unit root null hypothesis than does the usual Dickey-Fuller test.