Long lead-time forecasting of UK air passengers by Holt–Winters methods with damped trend

Planning decisions for air transport infrastructure require forecasts of air passenger traffic at relatively long lead-times – of the order of ten years. The Civil Aviation Authority (CAA) has collected data on air transport, including a long time series of monthly total UK passenger numbers, from 1949 to the present day, which we analyse in this paper. This series exhibits strong, regular growth and pronounced, approximately multiplicative seasonality. We estimate the historical growth using Holt–Winters decomposition and produce long lead-time forecasts. The long data series allows us to evaluate the out-of-sample forecasting performance of this and other forecasting procedures – a modification to the Holt–Winters method greatly improves forecasting performance for long lead-times. An assessment of uncertainty in the predictions is essential for planning decisions. The trend is the most important component to forecast for long lead-time prediction. The modification to the Holt–Winters procedure allows us to vary the trend used for these predictions, and so estimate the sensitivity of our forecasts to assumptions about the future trend. Persistent extreme growth is required for the most extreme forecasts, and the historical trend suggests that this is unlikely to continue for the length of time required.