In offering some cautions about traffic models, we have so far kept to the pleasant notion that, within the limits of their assumptions, they are accurate.
Let us complete our discussion by acknowledging that often they are not.
To estimate trip production, we have relied on indices of average trips per household per car or average trips per retail establishment of a certain size. These are crude assumptions. After all, trips per household are subject to great variance: some households of given size and number of cars drive much more than others. Sometimes the reason for variation is economic:
in some dense neighborhoods, the high cost of maintaining a car or to take certain trips gives travelers the incentive to switch modes. In other places, the reason is cultural. In immigrant neighborhoods where family links are extensive and spread far and wide, or in neighborhoods catering to some age groups (young, alternative lifestyle places), households may make more or fewer trips than average, to destinations that the simple models do not correctly forecast.
One answer is to increase the number of variables in the model: add more economic and social factors. In the effort to do so, much research is
being done and important advances have been made. But effort has also gone into creating ever more elaborate models, which weave together ever larger numbers of variables, with ever more speculative functional relation- ships (with what percentage likelihood does a lower middle-class household with members holding three jobs under recession conditions choose multi- occupancy versus single-occupancy trips to varied zonal destinations?) but imperceptible improvements in accuracy. The more that the models seek to incorporate the economic and social determinants of travel choice, under an ever changing economy and society, the more difficult the modeling becomes.
What is most reliable in a transportation model, we have found, is the physical relationship between road (or transit vehicle) capacity and trip volume. We know pretty well how many cars can fit on a road or how many passengers into a subway car. We know less well the social and economic determinants of travel demand. It is troubling to have to make a model ever more elaborate in estimating human behavior, when the economic and sociological disciplines are themselves in flux and full of internal disagree- ment. As transportation planners, we may just have to reduce expectations and accept that transportation models are tools for exploring alternatives, and far from being the outcomes of a science of travel behavior.
There are, nonetheless, important opportunities to make travel models better, and thereby to make better infrastructure decisions, including bridge decisions. The most important is the development of geospatial tools that can tell us where people or vehicles are at various times of the day. To be sure, there are great hurdles. Researchers must assure that peoples’ privacy is protected and that they do not fall prey to malevolent observers monitor- ing their movements. But the technology does open up the possibility that transportation models will come to be calibrated less from simple assump- tions about human behavior, and more from real-time data on trip volume, trip origins and destinations, and infrastructure capacity.
Further Reading
The Inrix company’s rankings of road segments with terrible traffic con- gestion is found on its website, www.inrix.com. A widely used textbook on the details of the four-step process is Juan de Dios Ortúzar and Luis G. Willumsen, Modeling Transport, 4th edition (Hoboken, NJ: John Wiley, 2011). A good introduction to the general reader is Edward Beimborm, Rob Kennedy, and William Schaefer, Inside the Black Box: Making Transportation Models Work for Livable Communities, published by the Center for a Better Environment and the Environmental Defense Fund (n.d.) and available for download at various sites on the Internet. Still another convenient source, copyrighted and posted by Oregon State University, Portland State Univer-
sity, and the University of Idaho, is the Transportation Engineering Online Lab Manual, which can be found by searching under that title; see its section on Traffic Demand Forecasting. The National Capital Transportation Plan- ning Board’s magazine Region introduced Washington’s use of transportation modeling in its 2003 issue (vol. 42), pp. 19–25. For examples of a Trip Generation Manual, see the volumes by that name produced by the Institute of Transportation Engineers in Washington, DC.
123