Abstract
This paper evaluates the forecasting performance of econometric models and time series model in the context of predicting visitors’ inflow to Japan from some selected Western and Asian countries using annual data on visitors’ arrival to Japan from 1960─2009. The models considered include the static OLS, error correction model（ECM）, autoregressive distributed lag model（ADLM）, time varying parameter model（TVPM）and autoregressive integrated moving average model（ARIMA）. The aggregate forecasts for the two regions over all forecasting horizons clearly indicate that the time varying parameter model outperforms the other econometric models and the univariate time series model in the forecast competition. The superiority of the TVP model suggests that by taking into account the possibility of structural changes in model building, forecasts accuracy will significantly improve.
