Many small island developing states in the Caribbean tend to depend on tourist arrivals as a main engine of economic growth, foreign exchange and job creation. International tourist arrivals are quite sensitive to news and the general economic cycle. Ideally, tourism planners in these small states would like to anticipate downturns in order to formulate and implement timely policy responses. This paper provides an assessment of the various approaches to forecasting tourist arrivals during a crisis using data from the tourism dependent island of Barbados. The results presented in the study suggest that structural time series models and regime switching models produce forecasts that are relatively unbiased. In general, however, relatively simple models (AR models) tend to produce more accurate forecasts (albeit biased) over the various downturns examined in the study.