Iowa travel model performance, twenty years later

Limitations of conventional sequential, static network modeling are often cited in transportation planning literature. Recently, the Travel Model Improvement Program (TMIP) has been instituted to address some of the well-known deficiencies in the state of travel model practice. Among these are an understanding of travel as human behavior, the ability to accurately model policy decisions, incorporation of sensitivity to changes in transit or other modes, and the application of technology to transportation. Because most travel models are continuously updated or replaced, we rarely have the desire, or opportunity, to examine the original constructs for accuracy after horizon years have come and gone - most of the shortcomings identified in the literature are accepted on a theoretical basis. This paper attempts to quantify the deficiencies in travel models by presenting two case studies of actual 1970's models developed for small urban areas to forecast present day travel. By revisiting anticipated results 20 years later, we differentiate the effects of problems in forecasting the land use transportation relationship with other model problems, and ask questions such as, `how accurate are the long range models?' `how good should they be?' and `for what reasons do they fail or succeed?' We conclude by answering these questions in the limited context of the two Iowa models.