Cast a broad net in evaluating traveler behavior: Managed Lanes analysis finds evidence of "theory blindness" that can impact model accuracy.

The report, submitted to the TRB Annual Meeting, found that traveler behavior often conflicted with expected behavior.

Date Posted
09/30/2019
TwitterLinkedInFacebook
Identifier
2019-L00909

Unrevealed Preferences: Unexpected Traveler Response to Pricing on Managed Lanes

Summary Information

A research paper submitted to the TRB Annual Meeting by researchers at Texas A&M University sought to examine the demand for Managed Lanes (ML). MLs are roadways that charge drivers a dynamic toll but that typically allow for faster, more reliable travel than on the unmanaged General Purpose Lanes (GPLs). Typically, tolls for MLs are calculated using models that assume that travelers choose their route based on costs and time savings.

In order to validate this approach, the researchers used data from the Katy and NTE freeways in Texas. Both routes have MLs that usually operated at or near free-flow speeds and GPLs that were routinely congested at peak hours. The data were collected from Automatic Vehicle Identification (AVI) sensors operated by the Texas DOT along the freeways, and were able to detect vehicles' unique transponder IDs to allow drivers' choice of roadway on a granular basis. The dataset contained nearly two million unique travelers making over 24 million trips. The authors note that this is significantly larger than most previous studies on travelers' revealed preference (RP).

The data suggested that the frequency of a traveler's freeway use had only marginal impact on their choice of lane. With the exception of travelers who made very few trips per month (between 1 and 3), drivers tended to exhibit similar rates of ML usage on each road. Additionally, it was found that frequent commuters typically received the same value for toll dollars as those who traveled infrequently. This indicates that commuters are not significantly more adept at judging the cost-benefit ratio of a ML toll than other drivers. Finally, a random sample of individual driver profiles showed that the ML's time savings and toll price had almost no impact whatsoever on a traveler's decision to use a ML.

Lessons Learned

The authors note that the conventional transit industry model for managed lanes assumes that the inputs to traveler decisions are technically complex, for example historic congestion levels, trip purpose, and cost-benefit analysis. However, the results of the study indicate that the decisions made are so statistically noisy that, instead, the best predictor of future behavior is simply a traveler's previous behavior. As such, it can be said that most travelers are not choosing to take one lane or the other with each trip, but rather default to taking a specific lane absent some sort of particularly compelling external influence.



The authors suggest re-framing models of driver behavior to de-emphasize "optimization" of lane choice. Approximately ten percent of travelers choose to use the MLs even in situations where they were negatively efficient (i.e., they had both a toll and greater congestion than the GPLs), indicating that optimization of travel is not necessarily a critical priority for drivers. The authors note that decision fatigue may be an influence, and should be considered: infrequent travelers who are unfamiliar with the route may find it too difficult to determine whether the toll is "worth it," while frequent commuters may prefer to avoid the mental load of calculating which route to take each day.



Finally, the authors also suggest that further research be done to understand traveler choice, specifically to develop a cohort-centric approach to understanding and forecasting managed lane use over time. In particular, they note that conventional wisdom can result in "theory-induced blindness," in which analysts narrow in on a predetermined explanation at the expense of broader understanding.

Goal Areas

Keywords Taxonomy: