Forecasting is contingent upon predictability, where a result is expected to be consistently observed, and uncertainty, which is the level of potential deviation from expected results. The prediction of a future outcome, such as the traffic level (e.g. port or airport terminal, transit ridership) logically experiences a decline in predictability and a proportional increase in uncertainty as longer time frames are being considered. This exercise falls into three main dimensions:
- Forecasting. Commonly using a trend of past observations and trying to infer this trend into the future. It assumes that the parameters related to an outcome remain constant and, as such, that its extrapolation is a relatively simple exercise using time series methods (e.g. moving average and compounded annual growth). These models also express uncertainty as growing confidence intervals. Once uncertainty reaches a level that is higher than predictability, forecasting ceases to be a relevant exercise, which requires scenario building. Depending on the level of predictability of what is being forecasted (some trends are more volatile than others), a timeframe of 5 years is usually considered within the realm of reasonable forecasting. In addition, forecasting is subject to biases that can under or over-estimate trends.
- Scenario building. Tries to predict future outcomes by changing an array of parameters, with each change part of a specific scenario. A common strategy is to present the future within the realm of low, medium, and high growth scenarios and apply these scenarios to forecasting by changing the annual growth rate. Scenario building does not remove uncertainty but may improve predictability by considering a set of possibilities. There is no formal time frame for scenario building, but 10 to 15 years can be considered acceptable.
- Speculations. At some point, the level of uncertainty is such that any prediction enters the realm of speculation. Quantitative methods lose much of their relevance, and under such circumstances, the future must be seen mainly as a speculative exercise about what may or may not take place. A common problem is that forecasting often makes predictions within a speculative time frame using a quantitative methodology to claim a high level of predictability while this predictability is very low. Any forecast looking more than 10 years into the future should be considered speculative.
Scale has an important impact on predictability as forecasting traffic for a single terminal is much more uncertain than for a region. Some outcomes are obviously easier to forecast than others as they have shown a greater level of stability and predictability in the past. For instance, demographic trends tend to be stable, shifting slowly and not subject to radical changes.
An important challenge resides in the planning time frame of megaprojects such as port or airport infrastructure. The delay between the decision to go ahead with the construction and the beginning of operations can easily last 5 years or longer. During that time, traffic expectations assumed by forecasting may have substantially changed. Further, the lifespan of most transport infrastructures can span decades, making predictions about their future use speculative.