According to the latest edition of the Transportation Energy Data Book from Oak Ridge National Laboratory's website (edition 26):
In 2005, domestic airlines on average consumed 20.5% more energy per passenger-mile than Amtrak, while cars consumed 27.2% more than Amtrak. Looked at the other way round, Amtrak consumes 17.0% and 21.4% less energy per passenger-mile than airlines and cars, respectively. [One passenger-mile is one passenger traveling one mile.]
Those percentages are derived from these Oak Ridge figures (British Thermal Units or BTUs per passenger-mile, 2005 data), organized here most to least efficient:
Commuter rail: 2,743
Rail transit: 2,784
Certificated air carriers: 3,264 (excludes international services)
notes of interest:
Amtrak consumed 14.6 trillion BTUs in 2005, which was 8.2% less than 15.9 trillion in 2003 and 19.3% below Amtrak's peak year of energy use (2001, with 18.1 trillion BTUs).
Amtrak in 2005 consumed 65,477,000 gallons of diesel fuel and used 531,377,000 kilowatt hours. [Both figures exclude consumption by commuter railroads for which Amtrak provides services.]
This indicates that 62.3% of Amtrak energy is diesel fuel vs. 37.7% electricity.
[Note that about half of Amtrak's ridership is on the more efficient electric powered routes in the Northeast.]
The tables from this document you may find most useful are:
Table 2.12 Passenger travel and energy use, 2004
Table 2.13 Energy intensities of highway passenger modes, 1970-2005
Table 2.14 Energy intensities of nonhighway passenger modes, 1970-2005
Table 9.10 Historical Amtrak figures including car-miles, train-miles, etc.
Table 9.11 Summary statistics for commuter rail operations, 1984-2005
Table 9.12 Summary statistics for rail transit operations, 1970-2005
Table A.15 Intercity Rail Fuel Use
[Note: Table 2.12 has 2004 data because 2005 data is not yet available for some modes. As the relevant footnotes explain, the airline statistics in Table 2.12 include "1/2 of international scheduled services" whereas those in Table 2.14 do not include any international services. This report also has considerable freight data.]
Several comments regarding what is the most energy-efficient way to travel . . . [By Christopher Parker]
First, the decisions are different if you are an INDIVIDUAL or a society. For us individuals, if we travel on something that's already going there, we are using NO fuel and causing NO pollution - unless we displace someone else who then has to drive.
Figures on energy efficiency are different, depending on if you measure seat-miles or people-miles. Airlines do better than Amtrak at cramming every last seat full of people. Automobiles are typically occupied by one person - so they get twice as efficient if two people are in them.
Much of the environmental impact of driving comes when building the car in the first place, which is added on top of the figures above. There is a lot of value in going car-free, and for society, in enabling people to be car-free. Thus renting a car every weekend is preferable to owning one.
Another measure, not often talked about, is the amount of land required. On this railroads are much superior to roads. I'm not sure if airports (which are very big) have enough footprint to cancel not needing any land at all once you are in the air.