Two new academic abstracts set up a striking debate, on the question of the cost-benefit analysis of rail projects. From Peter Gordon and Paige Elise Kolesar:
Rail transit systems in modern American cities typically underperform. In light of high costs and low ridership, the cost-benefit results have been poor. But advocates often suggest that external (non-rider) benefits could soften these conclusions. In this paper we include recently published estimates of such non-rider benefits in the cost-benefit analysis. Adding these to recently published data for costs and ridership, we examine 34 post-World War II U.S. rail transit systems (8 commuter rail, 6 heavy rail and 20 light rail). The inclusion of the non-rider benefits does not change the negative assessment. In fact, sensitivity analyses that double the estimated non-rider benefits and/or double transit ridership also leave us with poor performance readings. Advocates who suggest that there are still other benefits that we have not included (always a possibility) have a high hurdle to clear.
Meanwhile, from Robert Cervero and Erick Guerra, both of UC Berkeley:
The debate over the costs and benefits of rail passenger transit is lively, deep, and often ideological. As with most polemical debates, the truth probably lies somewhere in the middle of extreme views. Some rail systems have benefits that outweigh their costs, while others do not. Applying a commonly used transit-fare price elasticity to 24 of the largest light and heavy rail systems in the United States and Puerto Rico, assuming a linear demand curve, and accounting for a counterfactual scenario, we find that just over half of the systems have net social benefits. Although Los Angeles’ rail system does not “pass” our back-of-the-envelope cost—benefit analysis, as the network expands, it will begin to mimic the regional spatial coverage and connectivity of its chief competitor—the auto-freeway system—and approach the fare recovery rates of other large, dense American cities.
I don't have time to dig into the papers, but I think the difference in tone of the abstracts is interesting in itself. The issue of network completeness identified by the second abstract is of course critical. Discuss.
(Hat tip: Murray Henman. Post composed via wifi on the Amtrak Cascades, which is excellent!)