Discussion and Conclusion

Discussion

Firstly, employing aviation data would help estimate tourist arrivals when the official data are lagged or sparse. Introducing exogenous variables, such as the Google Trends index and Covid Stringency Index, proveD to be effective in reflecting the recent tourism challenges and subsequent recovery. Google Trends index, as auxiliary information, can reflect tourists’ willingness to travel, confirmed by the ratio approach in Palau and Vanuatu and by SARIMAX in Vanuatu, Tonga, and the Solomon Islands.

Secondly, combining forecasts from different methods is often better than single forecasts. When comparing models by their errors (RMSE), linear combinations (relative performance weights) outperform single forecasts and other combination methods in three islands, and the least squares combination occupies the remaining two islands. A detailed look at the weights, especially relative performance ones, reveals the performance of each model. The constrained least-square estimation also reflects each model’s performance over a period.

Taking Palau as an example, where relative performance outperforms other models, the ratio’s weights are above 0.55 on average during Covid-19. In contrast, the ratio’s model is constant at 0.676 in the constrained least-square estimation. Both results suggest that the ratio approach is better, possibly because it implicitly relies on the assumption that airlines dynamically adjust their schedules depending on travel demand.

Future Suggestions

The limited availability of aviation data makes modeling official statistics in a time series setting a challenging exercise. An extended timeframe will provide more evidence to confirm the statistical relationship between better aviation data and official tourism statistics. Additionally, the ratio approach implicitly treats the travel distance in the nominator as a constant, and flight-based data with origin and destination information could fill that gap and yield more accurate predictions.

The employment of Google Trends index for tourism research in small islands has potential, but it still lacks systematic evidence. Previous research exploits search engine data in developed countries, such as Google Trends data in markets where Google dominates the search engine market and Baidu in China. The selection and transformation of keywords and the lag between searching and traveling in PICs remain unclear, and more attention needs to be devoted to this field.