The operationai significance of the Lyapunov exponent and the correlation dimension for the measurement of chaos in economic time series of medium size length (200 observations) is investigated. In particular, models that are a mixture of a linear model, with a strong autoregressive component, and either a chaotic model or a white noise model are investigated. The empirical time series is the real exchange rate between Japan and the US. The results indicate that the implementation of the Lyapunov exponent for time series of 200 observations is not without problems and that for the JP/US real exchange rate an autoregressive model with white noise errors is more plausible than a model with chaotic disturbances according to the correlation dimension. However, the evidence in favor of the• stochastic model is not very strong and a nonlinear component may be present in the data.