Statewide Models: Activity-based Models
This page summarizes Section 6.2.2 of NCHRP Synthesis 514.
Tour-based and activity-based models models predict which activities are conducted when, where, for how long, for and with whom, and the travel choices they will make to complete them. Having this type of detailed model provides many benefits.
# Current Models
The decisions in these states to pursue advanced person travel models were based upon analytical issues thought to be difficult to address with trip-based models. Moreover, each had a highly motivated and accomplished champion within the agency committed to the significant advances required to achieve such goals. Thus, there are too few statewide activity-based models from which to generalize development and data requirements and trends from, made more difficult by the fact that two of the three implement them within a larger (and even rarer) integrated economic-land use-transport modeling framework. However, lessons learned from the implementation of comparable models in urban areas, as well as their experiences, can be instructive.
# Data Requirements
Model developers typically assert that the same data required for best practices person trip-based models will suffice for the development of activity-based models. Thus, it is likely that the types of data and resources required to develop them depend more upon the desired levels of spatial, temporal, and behavioral resolution desired, and the anticipated level of market segmentation, than the choice of modeling paradigm. The choice models included in such models are considerably more sophisticated than those found in aggregate models, even though most could also be implemented within a trip-based modeling framework. One of the more common additional data requirements is the classification of households and businesses by occupational categories, in addition to industry classification. This permits more precise matching of households with appropriate jobs in workplace location models. Such relationships can be gleaned from the Longitudinal Employment and Household Dynamics database, available from the U.S. Census Bureau. However, processing of these data requires considerable familiarity with their contents and limitations.
There are about two dozen operational activity-based models across North America, with a halfdozen more known to be under active development. The San Diego model was transplanted to the Miami region, at the cost of about $1 million. Thus, the cost of building an activity-based model at the statewide level, which would likely incorporate the enhancements described in the previous section for trip-based models, likely ranges from one to several million dollars. It further depends upon the desired functionality, amount of new model estimation required, new features, and data collection required to accomplish it.
An alternative to implementing a full activity-based model all at once is to incrementally move toward it. A trip-based model can be updated in steps, spreading the cost over several years, obtaining some benefits earlier, and enabling staff to gain familiarity and confidence with each part before making further changes. This enables the benefit of each improvement to be measured and understood, and the ability to drop those new features or formulations that fail to deliver them. Such a gradual transition typically starts with the implementation of synthetic populations and replacement of the trip generation module with a daily activity patterns model. Subsequent modules are then replaced over time. The costs associated with gradual transition are most likely no different in total, or slightly higher, than the development of a full activity-based model. They can be spread over time, however, and possibly redirected as agency requirements change.
There are many potential benefits that can be gained from such an approach, but in practice the most compelling appear to be their superiority over traditional models for pricing and equity analyses (Donnelly et al. 2010). The microsimulation framework enables full information about the traveler and trip to be retained throughout the model, as well as variation of rules and parameters across different geographic and market segments. The resulting model output is a simulated travel diary for the entire population, which can flexibly be mined and summarized based upon the analyses at hand. Whether such benefits are compelling enough to justify their investment depends upon the goals of and resources available to each state.
- Donnelly, R. & Moeckel, NCHRP Synthesis 514 Statewide and Megaregional Travel Forecasting Models, Transportation Research Board, Washington, DC, 2017. http://www.trb.org/NCHRP/Blurbs/176702.aspx
- Donnelly, R., G. Erhardt, R. Moeckel, and W. Davidson, NCHRP Synthesis 406: Advanced Practices in Travel Forecasting, Transportation Research Board of the National Academies, Washington, D.C., 2010.