Institute of Transportation Studies at UC Berkeley

An earlier empirical study of San Francisco Bay Area freeways concluded that HOV lanes unfavorably affect freeway traffic by creating congestion. That study attributed the observed congestion to HOV lanes and tentatively recommended their elimination over the full lengths of the freeways it examined; and even from all Bay Area freeways. It recognized, however, that its analysis is fragmentary and recommended further work to solidify its conclusions. This is logical since the study lacks a spatiotemporal analysis to pinpoint where and how congestion first forms (at bottlenecks).

The present report re-examines the same set of freeway sites in spatiotemporal detail to understand more deeply how HOV lanes are affecting traffic. It enriches the data from the original study with data from neighboring detector stations, to identify: first the locations (bottlenecks) where queues are triggered; and second the role that HOV lanes play in this phenomenon. This study includes an even more detailed analysis of high-resolution video data from a bottleneck where the HOV lane initially seemed to be having an unfavorable effect.

To our surprise, we found no compelling evidence that the HOV lanes were triggering delays and queues on the freeway sites in the earlier study. In all cases queues formed first at bottlenecks and, save for one questionable case, formed for reasons unrelated to the HOV lanes. This was true even on additional days that we studied. Moreover, data did not conclusively show that HOV lanes were reducing bottleneck flows or prolonging the queues; no adverse effects could be confirmed. To the contrary, and quite remarkably, the HOV lane seemed to increase the capacity of the bottleneck that was videotaped, even though that lane was underutilized. (The video data show that higher than normal discharge flows arose in the remaining lanes when the HOV lane was underutilized – enough even to compensate for that lane’s underutilization. Reassuringly, this effect had been predicted in earlier simulations.)

Although the sites assessed in this and the original study may not contain bottlenecks where the HOV lane is contributing to problems, the present study recognizes that such bottlenecks exist. They are just less prevalent than originally suspected. Fortunately for society, an HOV lane can be useful in most of these cases: theory and simulations indicate that if an HOV lane is rescinded near a problematic bottleneck but is preserved on the entire queued freeway stretch upstream, the lane will neither affect the bottleneck’s discharge flow nor the delay to vehicles that pass through the bottleneck. For typical freeways with four or more lanes, delay to all vehicles would change little. The HOV lane would allow HOVs to bypass most of the congestion without significantly increasing total vehicle-hours of travel.

This means that even in problematic cases, HOV lanes can usually be preserved and allowed to perform their intended societal function: reducing people-hours of travel without significantly increasing vehicle-hours of travel by favoring HOVs where freeway queues arise. To avoid increased vehicle delays, however, and perhaps even to increase bottleneck capacities, some HOV-lane installations should be modestly altered near bottlenecks. This report also describes field studies necessary to determine where and how such alterations should be deployed.