Jason Lemp

Understanding and Accommodating Risk and Uncertainty in Toll Road Projects: A Review of the Literature

Forecasting traffic and toll revenues for new highway projects involves great uncertainty because of the inherent uncertainty in the models used to make forecasts. As private investment becomes more common in project financing, quantifying the levels of risk and uncertainty associated with such projects becomes critical. This paper represents a review of many key studies and reports dealing with uncertainty in traffic and revenue forecasts for highway projects. These studies found that tolled projects tend to suffer from substantial optimism bias in forecasts, with predicted traffic volumes exceeding actual volumes by 30% or more about half the time. Moreover, projects with greater uncertainty tend to overestimate Year 1 traffic volumes more and stabilize at lower final traffic volumes. But after one controls for added optimism bias in traffic forecasts (compared with nontolled projects), there is little difference in uncertainty levels between tolled and nontolled forecasts. A typical way to address uncertainty in traffic forecasts is through sensitivity testing via variations in key inputs and parameters. A more extensive and less arbitrary version of this, Monte Carlo simulation, can provide probability distributions of future traffic and revenue, although it tends to require many simulations, demanding greater computational effort and time, unless networks are streamlined. Nonetheless, if reasonable assumptions for model input and parameter distributions can be made, Monte Carlo simulation generates a variety of useful information and establishes the actual likelihood of loss (rather than more basic win–lose indicators from a limited set of stress tests).

From Aggregate Methods to Microsimulation: Assessing Benefits of Microscopic Activity-Based Models of Travel Demand

Two competing approaches to travel demand modeling exist today. The more traditional four-step travel demand models rely on aggregate demographic data at a traffic analysis zone (TAZ) level. Activity-based microsimulation methods use more robust behavioral theory while focusing on individuals and households. Although the vast majority of U.S. metropolitan planning organizations continue to rely on traditional models, many modelers believe that activity-based approaches promise greater predictive capability, more accurate forecasts, and more realistic sensitivity to policy changes. Little work has examined in detail the benefits of activity-based models relative to more traditional approaches. To understand better the trade-offs between these two methodologies, results produced by both were modeled in an Austin, Texas, application. Three scenarios are examined: a base scenario, a scenario with expanded capacity along two key freeways, and a centralized-employment scenario. Results of the analysis revealed several differences in model performance and accuracy in terms of replicating travel survey and traffic count data. Such distinctions largely emerged through differing model assumptions. In general, activity-based models were more sensitive to changes in model inputs, supporting the notion that aggregate models ignore important behavioral distinctions across the population. However, they involved more effort and care in data manipulation, model calibration, and application to mimic behavioral processes better at a finer resolution. Such efforts help ensure that synthetic populations match key criteria and that activity schedules match surveyed behaviors, while being realistic and consistent across household members.

Empirical Investigation of Continuous Logit for Departure Time Choice with Bayesian Methods

Numerous models of travel timing have been calibrated and reported in the literature. Some studies have treated time as a discrete variable by using familiar discrete choice methods, whereas others have treated time in a continuous fashion. Both approaches offer distinct advantages. Here a continuous logit model of work tour departure time choice is estimated; this model offers the advantage of a continuous-time response. A random utility maximization structure is used to capitalize on the key advantages of both main approaches to the modeling of travel timing. Bayesian techniques are used to estimate model parameters, and estimation results suggest a variety of predictive densities for departure times across different individuals. In addition, ordinary least squares regression models are used to estimate travel times and their variance across times of day for the auto and transit modes. These network variables are used to inform estimation of the continuous logit model of departure time. The results are meaningful for multiple applications, and the continuous logit can readily be extended to a two-dimensional choice construct, such that the departure and return times can be modeled simultaneously. In addition, Bayesian estimation techniques allow for the utility function to take any number of forms, which may offer greater predictive ability.

Applications of Integrated Transport and Gravity-Based Land Use Models for Policy Analysis

Forecasting urban futures is of great importance to ensure the adequate provision of public and private services and to implement policies that guide demand while mitigating negative impacts. This study produces year 2030 predictions of land use and travel conditions across the Austin–Round Rock statistical metropolitan area of Texas by integrating a gravity-based land use model (modeled on Putmans Integrated Transportation–Land Use Package specification) with a standard travel demand model. The land use model cycles through job, household, and land consumption estimates across traffic analysis zones before feeding forward into a contemporaneous model of travel patterns. To understand the implications of different policies better, three scenarios were generated: a business-as-usual scenario, a congestion pricing plus carbon tax scenario, and an urban growth boundary (UGB) scenario. Results reveal how these transportation and land use policies may shape our land and travel futures, and they illuminate challenges and pitfalls of the gravity-based approach to land use (including the accompanying land consumption model). Of particular interest is that the imposition of road pricing (roughly 5¢/mi) had almost no discernable effect on land use predictions, yet resulted in the same predicted reduction in regional vehicle miles traveled as the UGB policy (roughly 15%).

Anticipating Welfare Impacts via Travel Demand Forecasting Models: Comparison of Aggregate and Activity-Based Approaches for the Austin, Texas, Region

A great disparity exists between the direction of travel demand forecasting by researchers and the travel demand models used by transportation planning organizations. Activity-based models of travel demand have become increasingly studied in the academic realm, and significant advances have been made in recent years. However, travel demand forecasting tools used in practice have lagged and rely on traditional, aggregate four- or five-step approaches. One reason behind the divergence in methods is the lack of work that directly compares performance of the two approaches. This research provides such a comparison, with an emphasis on calculations of traveler welfare. A traditional, aggregate model and an activity-based microsimulation model of travel demand were developed in parallel by using the same data for Austin, Texas. The models were applied for a base scenario and for several policy scenarios to test model performance and sensitivity to inputs. The spatial distribution of traveler welfare implied by these scenarios illuminates a variety of key differences in the models’ performance and suggests that the activity-based model enjoys a greater sensitivity to inputs. Additional outputs demonstrate the level of segmentation that can be attained in model outputs using microsimulation methods. The comparative analysis of these two competing approaches to travel demand forecasting also offers some insight into the practical benefits of an activity-based approach.

A Bivariate Multinomial Probit Model for Trip Scheduling: Bayesian Analysis of the Work Tour

As tour-based methods for activity and travel participation patterns replaces trip-based methods, time-of-day (TOD) choice modeling remains problematic. In practice, most travel demand model systems handle tour scheduling via joint-choice multinomial logit (MNL) models, which suffer from the well-known independence of irrelevant alternatives (IIA) assumption. This paper introduces a random utility maximization (RUM) model of tour scheduling called the bivariate multinomial probit (BVMNP). This specification enables correlations across TOD alternatives, both outbound and return (on a tour) and over time slots (in a day). The model is estimated in a Bayesian setting on work-tour data from the San Francisco Bay Area (with 28 time slots). Empirical results suggest that a variety of individual, household, and tour characteristics have reasonable effects on scheduling behavior. For instance, older persons typically pursue work tours at earlier times of day, part-time workers pursue their work tours later, and those with additional activities and tours tend to arrive slightly later and leave much earlier than those undertaking only a single tour, everything else constant. The model out-performs a comparable MNL, while offering reasonable implications under a variety of road-tolling scenarios.

Capturing Latent Household Preferences in Daily-Activity Pattern Choices: Application to Activity-Based Model of Houston–Galveston Region in Texas

One fundamental feature of most operational activity-based models (ABMs) developed in the United States is the concept of a day activity pattern, which, in its broadest sense, is a way of characterizing all the activities undertaken in a day at the individual level. This pattern often includes a sequence of models that generate activities of different types for each individual. In many cases, these models treat all individuals in a household independently, while in other cases, specialized techniques capture intrahousehold relationships. This paper presents a new technique for capturing some of these intrahousehold relationships in a daily-activity pattern (DAP) model via terms for latent household preferences while also allowing for other specific relationships to emerge in more standard ways. The model is estimated on data from the Houston–Galveston region and serves as an extension to the DAP model framework being used in the ABM system under development for that region. The results suggest that strong household level preferences exist. For many activity pattern types, these household preferences are more important than the preferences of the individual. In addition, the modeling framework is concise, can easily accommodate households of any size, and can predict the exact number and type of mandatory tours (i.e., work, school, and university activities) for individuals in a single model.

Issues in Expanding the Denver Regional Activity-Based Model for Statewide Modeling in Colorado

This paper discusses the activity- and tour-based model developed for the Denver Regional Council of Governments in Colorado with respect to two major issues that underlie its expansion to a statewide travel model: managing short- and long-distance travel within a single model and accounting for overnight travel. In a statewide model, the issues are interconnected to the modeling of closed and nonclosed tours. Closed tours are those that begin and end at the same location during the travel day (i.e., home). Although nonclosed tours typically are not modeled in regional activity- and tour-based models, or they are modeled with relatively simple procedures, they are more important in statewide models to accommodate overnight travel. Some long-distance travel involves an overnight stay at a location other than the traveler’s home. Such nonclosed tours, which begin or end the day at a location other than home, must be modeled explicitly in a statewide model. In this paper, required adjustments are identified for key activity- and tour-based model components, and the impacts of the explicit modeling of nonclosed tours for the tour mode choice model are presented.