Ex Ante Tourism Forecasting Assessment

Abstract

Although numerous studies have focused on forecasting international tourism demand, minimal light has been shed on the factors influencing the accuracy of real-world ex ante forecasting. This study evaluates the forecasting errors across various prediction horizons by analyzing the annually published forecasts of the Pacific Asia Tourism Association (PATA) from 2013 to 2017, comprising 765 origin–destination pairs covering 31 destinations in the region. The regression analysis shows that the variation in tourism demand and gross domestic product (GDP), covariation between tourism demand and GDP, order of lagged variables, origin, destination, and forecasting method all have significant effects on the forecasting accuracy over different horizons. This suggests that tourism forecasting should account for these factors in the future.

Combination forecasting using multiple attribute decision making in tourism demand

“…First, beyond grey models, it will be interesting to include more parametric and nonparametric models in a combination set because a diversity of models may improve the synergy of combination forecasting (Song et al , 2009). Another possible research area is to examine the effect of explanatory variables on forecasting accuracy, such as the gross domestic product as suggested by Liu et al (2020), content analysis in social media by Thomaz et al (2017), Google Trends index by Volchek et al (2019) and online word-of-mouth at a destination by Williams et al (2015). Finally, as suggested by Li and Jiao (2020), by using Big Data mining and analytics, the combination forecasting used here will be applicable to instant micro-level forecasting.…”

Bayesian bootstrap aggregation for tourism demand forecasting

“…The one‐sample Wilcoxon test suggests that neither ordinary bagging nor BBagging significantly improves the performance of GETS for the quarterly forecasts. It is not surprising that the performance of GETS tends to improve and become more stable as the sample size increases from 16 periods in the annual forecasts to 68 periods in the quarterly forecasts (Liu et al, 2020). Thus, the superiority of BBagging in handling high forecast variability in small samples is not fully reflected in the quarterly forecasts.…”

“…Ordinary bagging in quarterly forecasting cannot beat GETS, which contradicts the findings of Athanasopoulos et al (2018), perhaps because the sample size in this study is much smaller than theirs. Although GETS becomes more stable in larger samples (Liu et al, 2020), ordinary bagging can be much more accurate and beats GETS when the model is fitted with large samples. The advantages of GETS over ordinary bagging should be examined in future studies.…”

“…The GETS model has also been used to investigate the influence of the 2008 Global Financial Crisis on the tourism and hospitality industry in Asia (Song et al, 2010; Song et al, 2011; Song & Lin, 2010) and the UK (Page et al, 2012), and to examine the global impact of terrorism attacks on inbound arrivals (Liu & Pratt, 2017). Liu et al (2020) summarised the performance of the ADL‐GETS model in a large‐scale ex ante forecasting project. They found that the variation in the tourism demand and gross domestic product (GDP) of source markets and the covariation between tourism demand and GDP have significant effects on the forecasting accuracy of the ADL‐GETS model over different horizons.…”

Timing matters: crisis severity and occupancy rate forecasts in social unrest periods

“…Among the 42 comparisons across different hotel classes and horizons using RMSE, 38 occasions favor intervention models. Different forecast accuracy measurements may have varying results (Liu et al, 2020). The different performance of MAPE and RMSE in this study is related to the structure of RMSE, where each forecast error is weighted by the magnitude of itself by squaring the error.…”