| Summary: | Demand forecasts are the most important piece of information used to make revenue management decisions for lodging properties. High demand forecasts may lead to increases in room rates and stay restrictions while low demand forecasts may result in price decreases and easing of stay restrictions. A number of demand forecasting methods, both long-term (more than 90 days prior to a target date) and short-term (within 90 days of a target date) were modelled and compared for the lodging properties at a major North American ski resort. Long-term forecasting methods included random walk, multiplicative Holt-Winters, ARlMA (autoregressive integrated moving average), linear regression, and nonlinear regression. Short-term models included the five long-term forecasting methods as well as additive pickup and a regression-based booking curve model. In terms of long-term forecasts, the nonlinear regression method was found to be superior while capacity was trending upward but after a capacity shock (unexpected loss in capacity) the random walk method proved optimal. In terms of short-term forecasts, the regression-based booking curve model was optimal in-sample and data was not tested out of sample. Further, the long-term nonlinear regression model and short-term regression-based booking curve model explicitly defined seasonal periods. These statistically defined seasonal periods should help management set seasonal rate targets as well as better understand typical booking patterns among periods. === Business, Sauder School of === Marketing, Division of === Graduate
|
|---|
