Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
29 Forecasts of airport aviation activity have become an integral part of transportation planning in the United States. The type and method of forecasting can depend importantly on the pur- pose for which the forecast is being made, and distinctions between short-term/long-term and constrained/unconstrained demand can lead to significant differences in the associated activity forecasts that are produced. Such differences, how- ever, do not necessarily mean that one is more correct than another. In practice, most airport and regional and state efforts use fairly simple methods to produce forecasts. Data availability and budget constraints often dictate what forecasting tech- niques are employed. Another factor that affects how airport forecasts in the United States are prepared is the set of rules and guidelines for preparing airport master and system plans set down by FAA. The primary methods used to produce airport aviation activity forecasts reflect these constraints. The market share method is a top-down approach, where it is assumed that activ- ity at a particular airport is related to growth in some aggregate external measure (typically a regional, state, or national avia- tion growth rate). Some studies that use this method look only at the very recent past and use a small number of historical data points to establish the numerical relationship between airport activity and the selected external factor. Although such an approach is not generally recommended, it must be recognized that the market share method is used extensively in aviation forecasting, particularly when the historical data are suspect or when past history does not correlate well with other observ- able factors. Additional research would be warranted on the reliability of historical activity data and on how to gather data more effectively in the future. This issue is of particular con- cern at smaller nontowered airports. In cases where reliable historical data can be gathered, econometric modeling has been an effective tool for generat- ing forecasts of airport activity. Although econometric model- ing is potentially a very sound and effective method, there are many ways in which the specific model that is estimated can go wrong. A more detailed study of the many airport forecasts produced with this method could be undertaken to identify the most common sorts of statistical or data problems that affect aviation forecasts, which in turn would provide some useful guidance for future modeling efforts. Another available technique is time series extrapolation. Simple trend analysis, such as year-over-year or month-over- month extrapolation, can be a useful approach, particularly for short-term forecasts. More sophisticated time series methods such as exponential smoothing and BoxâJenkins analysis have not been used often in aviation forecasting. However, for short-term forecasts where there are complex time relation- ships relating to seasonality and trend, these time series meth- ods may be valuable. Further study that extends the analysis of Pitfield (1993) to allow direct comparison of out-of-sample, time series predictions with those from a conventional econo- metric model of aviation activity could be valuable in assess- ing the conditions under which one method might be preferred to another. Assessing forecast uncertainty and accuracy are two sepa- rate but related issues that are often neglected in airport activ- ity forecasting efforts. With regard to uncertainty, most stud- ies provide only point estimates of forecasted values, although it is common to also present alternative âhighâ and âlowâ esti- mates. Although this can provide a reasonable range of esti- mates, there are additional sources of uncertainty related to the statistical properties of the models employed that are often neglected entirely. It is frequently argued that having access to reliable data is more important than the specific statistical model that is employed; however, further investigation into the significance of model-related statistical uncertainty for avia- tion forecasting efforts would be warranted. Finally, research on just how well aviation forecasts proj- ect future airport activity would also be warranted; optimism bias and the possibilities for combining competing forecasts are important topics that to date have not been sufficiently investigated. Such research could be carried out by surveying past forecasts and comparing their projections with currently observed activity levels. These issues are particularly relevant for long-term aviation forecasts that are used to support major decisions regarding capital investments and potentially large expenditures of public funds. CHAPTER FIVE CONCLUSIONS AND SUGGESTIONS FOR RESEARCH