Differentiated Road Pricing, Express Lanes, and Carpools: Exploiting Heterogeneous Preferences in Policy Design

Nathaniel Baum-Snow:

Small, Winston, and Yan's paper provides valuable new insights on the potential social welfare gains associated with implementing various new urban highway lane pricing and carpooling options. The authors apply modern econometric techniques to estimate a partial equilibrium model of highway travel demand. Using the resulting estimates, the authors perform a detailed welfare analysis of different high-occupancy vehicle and toll lane policies. California State Route 91 (SR91) is an ideal case to evaluate. The road has few exits, considerable variation in congestion delays for different times of day, and variation across observations in the cost of the HOT lane. The structural approach lends itself well to the full welfare analysis of various policy alternatives performed at the paper's end. The authors choose their empirical specification to be flexible enough to inform us about distributional consequences of different HOV, toll lane, and HOT policies in addition to their mean effects. Finally, the authors deserve considerable credit for their data collection efforts. Without these unique data, this analysis would not have been possible. This paper should be of great use to urban transportation policymakers considering the implementation of highway congestion tolls and HOV lanes.

The authors cite a body of work documenting that road congestion delays have been rising rapidly since the early 1980s. At the same time, the fraction of the U.S. population commuting by car continues to increase, reaching 91 percent in 2000. Commuting by car has risen even in the face of large investments in public transit infrastructure. On the face of it, this may seem like a golden opportunity for carpool lanes. At very low cost, they have the potential to provide large congestion reductions, thereby increasing commuting speeds for many commuters. However, as seen in table 10, the fraction of auto commuters who carpool has been falling rapidly. In 1980 it stood at 20 percent, while by 2000 it had fallen to just 12 percent nationally. The decline in carpooling is seen for central city and suburban residents alike. It also holds in the Los Angeles Consolidated Metropolitan Statistical Area (CMSA), where the SR91 experiment took place, though to a lesser extent. In Los Angeles carpooling rates fell from 17 to 15 percent of drivers between 1980 and 2000. Over this period, the fraction of commuters using public transit declined by less in both absolute and percentage terms, despite considerable public investment in both transit and HOV lane infrastructure.

Table 10.

Trends in Carpooling, 1980-2000^a

How can one explain steeply declining carpooling rates despite longer congestion delays and more drivers on the roads? A few clues are revealed by examining income levels and travel times. Table 11 presents average commuting times and total income levels (in 1999 dollars) of people who drive alone and people who carpool. It shows that carpoolers are considerably poorer than solo drivers and that their average commuting time is greater. This pattern holds in central cities and suburban areas, persists over time, and also holds for the Los Angeles CMSA. Two explanations may be important in accounting for these patterns. Poor people may have longer distance commutes and as a result may be more likely to find carpooling (and carpool lanes) to be useful. In addition, while carpooling reduces highway travel time where there are carpool lanes, it increases nonhighway travel time due to the need to circulate to pick up or drop off passengers. Population and employment decentralization can only have increased the fixed-time cost of carpooling. As such, the driving force behind more prevalent carpooling among the poor may be savings on pecuniary costs of car ownership and gasoline rather than time savings.

In the SR91 case, which involves longer than average commutes, about eight minutes is the maximum time savings available from carpooling under any HOV or HOT lane scenario considered. Picking up and dropping off a carpooler, or even coordinating the departure of two people who live or work in the same location, can easily take this much time every trip. The suggestive evidence...