Previous Competition
Background
The COVID-19 pandemic has profoundly impacted the tourism industry. Most tourist destinations and tourism-related businesses are facing unprecedented challenges during the crisis, and more crucially, planning for the post-crisis recovery. The pandemic is still ongoing, and when the tourist market will recover remains unknown. Scientific forecasting of the full scale of the impact on the tourism industry and the market recovery is critical for strategic planning of tourist destinations and tourism-related businesses.
There has been a long tradition in the field of forecasting to organise prediction competitions, such as the series of the Makridakis Competitions (also known as the M-Competitions) which started in 1982. The aim is to advance the theory and practice of forecasting. The outcomes of the competitions were published in a leading forecasting journal: International Journal of Forecasting. So far there has been only one organised tourism forecasting competition, in which two teams participated 10 years ago (Athanasopoulos, Hyndman, Song and Wu, 2011), and there has been no forecasting competition focusing on tourism recovery from a global crisis.
At the moment, the Curated Collection of Tourism Demand Forecasting of Annals of Tourism Research is calling for contributions to a special issue on “Tourism Demand Forecasting: From Crisis to Recovery”. This is a perfect opportunity to launch a new tourism forecasting competition and attract scholarly interests in this line of research.
Aim of the competition
Given the above background, the aim of this hotel performance forecasting competition is three-fold:
- To advance the methodology of tourism forecasting and contribute to the development of this field of research
- To inform the tourism industry and destination management and marketing organisations of the good forecasting practice and the predicted impact of COVID-19 on tourism
- To promote Annals of Tourism Research as the leading and main outlet for state-of-the-art tourism forecasting research through the Curated Collection on tourism demand forecasting.
Organisation of the competition
Organisers:
-
Haiyan Song and Gang Li
Participants:
- Three regional teams (Asia and Pacific, Europe, and Africa), based on 18 members of the International Association for Tourism Economics.
Scope of the competition
• Procedures
There are two stages of forecasting in relation to two purposes:
First stage— ex post forecasting of tourist arrivals before COVID-19: Based on data up to the end of 2018, each participating team will predict tourist arrivals in the 20 given destinations across all regions throughout 2019 (see Annex 1 for the full list). The purpose of the first-stage forecasting is to identify the most accurate forecasting method(s) in “normal” times.
Second stage—ex ante forecasting of tourist arrivals during and after COVID-19: Based on the latest available data, each participating team will predict tourist arrivals in the 20 destinations up to the end of 2021. The purpose is to identify the most accurate forecasting method(s) and procedures in a crisis situation.
• Variables to be forecast
Tourist arrivals in each of the 20 destinations, including the total number of tourist arrivals, and tourist arrivals from five key source markets. In total, 120 series are to be forecast.
• Data frequency
Quarter data will be used for all forecasts.
Outline of the rules
• Accuracy evaluation
This competition will focus on point forecasts only. The mean absolute scaled error (MASE) proposed by Hyndman and Koehler (2006) and widely used in previous forecasting competitions is used as the forecast error measure for this competition (see Athanasopoulos et al. 2011, p831 for the formula).
A weighting scheme of 40%: 60% will be applied to the results of the Stage 1 and Stage 2 of forecasting, respectively. Within each stage, equal weights will be applied to different series and different forecasting horizons.
• Choice of forecasting methods
For the first-stage ex post forecasting prior to COVID-19, any forecasting methods are permitted, including time-series, econometric, artificial intelligence and big data methods. Each team should consider at least three methods and choose the best one for the competition evaluation. In the case of an econometric model being used in the first stage of the forecasting, actual values instead of predicted values of the explanatory variables should be used during the forecasting period. Where an intervention variable is introduced to capture the effect of a one-off event during the model estimation/training period, the specific event and the specification of the dummy variable need to be explained in a separate document. Based on an initial dataset up to December 2018, rolling forecasts of one-, two-, three, and four-quarter-ahead for 2019 are produced. Therefore, there will be four one-quarter-ahead forecasts, three two-quarter-ahead forecasts, two three-quarter-ahead forecasts and one four-quarter-ahead forecasts for each series, and the average forecast accuracy under each horizon is to be calculated. Based on the overall performance (i.e., across four horizons, six arrivals series per destination and 20 destinations in total), each team choose the best performing model overall, the results of which will be submitted for competition evaluation. The best performing model should also be used to produce baseline forecasts (2020Q1 – 2021Q4) of the second-stage forecasting.
For the second-stage ex ante forecasting during and after COVID-19, a baseline and three scenarios need to be set up. The baseline should assume no COIVD-19, based on the data up to the end of 2019 using the best performing model from the first stage; three scenarios of COVID-19’s impacts include mild, medium and severe conditions. The scenarios need to be clearly defined, and the scientific procedures and evidence of scenario forecasting need to be explained in detail. Any models (e.g., time-series, econometric and artificial intelligence models) and techniques are allowed for scenario forecasting. The evaluation will be based on the most accurate set of scenario forecasts among the three over the forecasting period of Quarter 1 to Quarter 4, 2021.
References
Athanasopoulos, G., Hyndman, R. J., Song, H., & Wu, D. C. (2011). The tourism forecasting competition. International Journal of Forecasting, 27(3), 822–844.
Hyndman, R. J., & Koehler, A. B. (2006). Another look at measures of forecast accuracy. International Journal of Forecasting, 22, 679–688.
Outcome of the competition
• First place: Asia Pacific Team
Members: Richard T.R. Qiu, Doris Chenguang Wu, Vincent Dropsy, Sylvain Petit, Stephen Pratt, and Yasuo Ohe
• Second place: Europe Team
Members: Anyu Liu, Laura Vici, Vicente Ramos, Sauveur Giannoni, and Adam Blake
• Third place: Africa Team
Members: Nikolaos Kourentzes, Andrea Saayman, Philippe Jean-Pierre, Davide Provenzano, Mondher Sahli, Neelu Seetaram, and Serena Volo
Related publications
Kourentzes, N., Saayman, A., Jean-Pierre, P., Provenzano, D., Sahli, M., Seetaram, N., & Volo, S. (2021). Visitor arrivals forecasts amid COVID-19: A perspective from the Africa team. Annals of Tourism Research, 88, 103197. https://doi.org/10.1016/j.annals.2021.103197
Liu, A., Vici, L., Ramos, V., Giannoni, S., & Blake, A. (2021). Visitor arrivals forecasts amid COVID-19: A perspective from the Europe team. Annals of Tourism Research, 88, 103182. https://doi.org/10.1016/j.annals.2021.103182
Qiu, R. T. R., Wu, D. C., Dropsy, V., Petit, S., Pratt, S., & Ohe, Y. (2021). Visitor arrivals forecasts amid COVID-19: A perspective from the Asia and Pacific team. Annals of Tourism Research, 88, 103155. https://doi.org/10.1016/j.annals.2021.103155
Song, H., & Li, G. (2021). Editorial: Tourism forecasting competition in the time of COVID-19. Annals of Tourism Research, 88, 103198. https://doi.org/10.1016/j.annals.2021.103198
Song, H., Li, G., & Cai, Y. (2022). Tourism forecasting competition in the time of COVID-19: An assessment of ex ante forecasts. Annals of Tourism Research, 96, 103445. https://doi.org/10.1016/j.annals.2022.103445