Total tourist arrival forecast: aggregation vs. disaggregation

Total tourist arrivals are the sum of disaggregate subcomponent arrivals by country of origin. We use seven time-series models to assess whether the aggregate approach that directly forecasts the total tourist arrivals outperforms the disaggregate approach that produces the total arrival forecast as an unweighted sum of its subcomponent forecasts. Based on Hong Kong's monthly tourist arrival data, we find (a) the seasonal autoregressive integrated moving average model outperforms the other non-seasonal and seasonal models under the aggregate approach, and (b) forecast performance can be improved by the disaggregate approach.