It is by having well connected networks that serve a large proportion of the travel market that countries such as France, Japan and Germany can have such high ridership levels in intercity rail travel (100 million passengers in France, 120 million in Germany, over 300 million in Japan).

The question Glaeser is considering is whether HSR makes economic sense (whether the benefit exceeds the cost). His (provisional) answer is that it doesn’t. The “alternatives” here are “build HSR” and “don’t build HSR” and the analysis suggests that the latter alternative is the better one.

Is it doing nothing? Then at what point does the rising cost of congestion justify construction of something?

At the point where the benefit of that thing (from congestion reduction and/or whatever else it may be) exceeds the cost. If the cost of building anything would exceed the cost, then yes, the best alternative is to build nothing.

Why would he choose this corridor to examine? Why not begin with the most natural place to construct true HSR — the Northeastern Corridor — or the state moving fastest toward building its own true HSR network — California?

Er, how is the Northeast Corridor “the most natural place to construct true HSR?” I can think of several reasons why political and economic obstacles to “true HSR” may be greater in the NE corridor than for a Dallas-Houston line. The NE corridor already has HSR-lite rail service (the Acela), but there is no direct rail service at all between Dallas and Houston. The relatively small improvement in travel times of “true HSR” over the Acela may not be worth the huge investment that would be required for new ROW and track. The potential benefit of going from no reail service at all to “true HSR” may be much greater than the potential benefit of going from HSR-lite to “true HSR.” Texas is mostly flat, land is cheap, and there is much less existing development. So land cost, environmental and NIMBY obstacles to “true HSR” may also be much weaker in Texas than in the NE. The same kind of considerations would apply to a comparison between Texas and California.

Well, Glaeser was able to use Dallas’ low share of commuters taking transit to knock the corridor’s estimated ridership down by half. Transit’s share of commuting in Los Angeles is nearly three times that in Dallas. In San Francisco, transit’s share, at 32.2 percent, is more than seven times larger than in Dallas. Presumably this difference had something to do with his choice.

Your argument here is also totally confused. Yes, on the basis of transit’s share of commutes, we would expect higher ridership on a SF-LA route than a Dallas-Houston route. But San Francisco has a smaller workforce than Dallas, which would tend to erode that effect. And the proposed HSR route for SF-LA is 432 miles, almost twice the route length for Dallas-Houston. So the construction, maintenance and operation costs for a SF-LA line would be much higher than for a Dallas-Houston line. So SF-LA would probably need a much higher ridership than Dallas-Houston for the benefits to exceed the costs. Is the higher share of commutes in SF and LA enough to provide that additional ridership? You don’t even try to answer that question. Given that Glaeser’s analysis suggests that Dallas-Houston fails the cost-benefit test so egregiously (costs exceed benefits by between 3-to-1 and 6-to-1), it does not seem plausible that your quibble about commute shares would invalidate his basic conclusion.

And swap $12b for $2b in the Fixed Costs part of the formula, and even under Glaeser’s over-simplified formula where all Fixed Costs have to be justified by first party benefits, the case is much easier to make.

Ian, I disagree that offering an alternative analysis “wouldn’t be that difficult” for all the reasons laid out above. This is a highly nuanced calculation where the outcome is greatly affected by even small changes to just one or two assumptions. The criticisms Ryan made are important and should be taken seriously.

Your last sentence claims Glaeser couldn’t do much worse. Worse would be if he hadn’t written down any model at all and had just thrown together a bunch of random speculation with out any quantitative calculation, what you did here.

It seems obvious that running the analysis on a city-pair with awkward airport locations and less frequent service would be a good place to start.