Forecasting Tourism Demand in the Short Term: the Case of Andalusian Hotel Establishments

Policy makers and businessmen demand real-time information and short-term forecasts about bed-nights occupied and the occupancy rate in hotel establishments (the latter being the percentage ratio of the number of bed nights occupied to the number of bed-nights on offer in the same period, usually one month). So, when these forecasts are obtained using single equations time series models, implicit forecasts of bed nights on offer, i.e., bed capacity, can also be derived. These implicit forecasts of bed capacity, which we call indirect forecasts, yield a higher RMSE than the “direct“ forecasts when a univariate time series model is used. To face this problem within the single equation time series modelling framework, the only one feasible due to sample size limitations, we propose obtaining forecasts by the following method. First, a set of “direct” forecasts are obtained for each one of the three variables involved (bed nights occupied, bed capacity, and occupancy rate) using transfer functions and ARIMA models, according to the case. When obtaining direct estimations, alternative models are considered taking into account aggregation problems and different leading indicators related to expectations of the business people. Second, an “indirect” forecast is obtained for each variable using the “direct” forecasts corresponding to the other two variables and taking into account the relationship which defines occupancy rate. Third, a set of “combined” forecasts, defined as the quadratic mean of direct and indirect forecasts, is obtained. This paper shows that: i) the combined forecasts of the bednights occupied and the occupancy rate perform better than the corresponding direct and indirect forecasts under the RMSE criteria, while the direct forecast is the best option for bed capacity; ii) the selected forecasts for the three variables, according to this criteria, fulfil the relationship established by the definition of occupancy rate.