Scenario Forecasting for Travel Demand Modeling
BackgroundQuantified values of population, households and employment covering the metropolitan area at a fine level of geographic detail are the principal data elements in calculating locally generated person-travel demand over the region’s transportation network. Additional socioeconomic variables are also required to account for the considerable volume of commercial vehicle traffic, especially trucks, having origins and destinations within or beyond the metropolitan region.
Calculating future travel demand requires two robust methodologies to forecast the needed socioeconomic inputs. First, an aggregating methodology that estimates future small-area population, household and employment values using Census data along with stated assumptions regarding expected land use. Second, a disaggregating methodology that appropriately assigns determinants of national and global economic growth or decline at the metropolitan scale.
Numerous government agencies typically have land use and transportation planning jurisdiction over the modeling region. These agencies declare, with varying degrees of precision and certainty, future expectations for population, employment and economic activity in the course of preparing their long-range plans. Depending on their respective missions, these organizations may have diverging outlooks on a future that is essentially unknown beyond the next few years. Some agencies (e.g. a State DOT) as steward of tangible assets, may take a risk-averse view and prefer to make long-term planning decisions based on conservative forecasts derived from familiar conditions and previous outcomes. Other agencies (e.g. a regional MPO) may be charged with pursuing structural changes to remedy persistent social and economic problems and will prepare forecasts that reflect the outcome of sweeping policy reforms across the region.
Similarly, federal agencies (along with some private financial services) prepare and publish economic projections intended to guide national policy-making and evaluate the investment risk associated with planned government actions. As with the local agencies, these national or global-scale publications can produce a wide range of forecast values depending on their intended use. Some may use econometric techniques to predict short and medium-term risk exposure, while others extrapolate long-term historical averages decades into the future.
For purposes of travel demand modeling, these two forecasting sentiments can be reconciled through a technique of constructing alternative future scenarios. The scenarios, themselves, are distinguished by a detailed specification of the environmental (i.e. demographic, economic, social) influences that are expected to establish a pattern and intensity of business activity, thereby resulting in a plausible arrangement of land use at a prescribed future date. Each future scenario can be further distinguished by describing any supporting or countervailing government interventions assumed to bring about a desired policy objective. Alternative future scenarios are usually thematically formed by articulating varying levels of capital investment, pricing (e.g. taxes, tolls, user fees), or technological innovation. It is also necessary that the forecasting methodology account for the impact that government actions may have on environmental variables such as household composition, labor force participation or technology adoption.
In any case, each unique combination of assumed environmental influences and government interventions will affect both the rate and geographic pattern of population and employment change going forward. There are two prevailing approaches to quantifying these changes at the level-of-detail needed for travel demand modeling. One approach is to rely on empirical observation of historic trends, before-after studies and local expert opinions to manually apply changes across the region on a case-by-case basis. This approach can almost always withstand local scrutiny provided local conditions are sufficiently considered and reflected; but it is also painstakingly slow, error-prone and difficult to reproduce. A second approach is to apply econometrics and spatial statistics to automatically dampen or enhance the development potential of a geographic location in response to influences and interventions specifically stated in the scenario definition. This approach has the advantage of producing multiple scenarios in shorter time, allows for systematic error-tracking, and is capable of being modified and replicated by others.