This Appendix to "Fuel-Efficiency of Travel in the 20th Century" by David S. Lawyer, is about 3 times longer than the text part of the article. It's on the Internet and thus need not be printed with the article.
April 2004: Added subsection on buses
March 2004: Used more recent data for Amtrak (for 2000 and 2001) showing a sharp drop in fuel efficiency.
August 2003: I discovered that the government "statistics" for air had apparently erroneously counted fuel used for air freight as being used for passenger transport. Mea Culpa for not being more vigilant. Of course the BTU/PM of TEDB for auto is "too low" for 1970-1989, and not "too high" as I erroneously typed in 2002.
July 2003: Originally (in June 2002) I estimated that the energy-efficiency in 2000 was 4 times that of 1900. It turns out to be over 5 times. The July 2003 version fixed that error. In 2002, I had both overestimated energy-efficiency for 1900 and underestimated it for 2000. For 1900, I failed to realize that old coal likely had a higher energy content (better quality) than today's new coal and erroneously used the data for new coal. It's still not clear exactly what the BTU/pound was for coal in 1900, so there could be future corrections. Since I didn't have the data for 2000 (in my 2002 version) I failed to take into account improvements in energy-efficiency so I used a low estimate of 30 pmpg for 2000 when it should have been 33 pmpg. See [3]. I was just too hasty in making estimates and didn't refine them until July 2003.
[Ayres]: Energy Sources -- The Wealth of the World, by Eugene Ayres and Charles Scarlott, McGraw-Hill, NY, 1952. On p. 139, Fig. 1, is a graph of power plant fuel rates (energy-intensity) between 1902 and 1950.
[BusFacts-1935]: Bus Facts. 1935. p.9: chart: "Passenger-miles in United States by Various Means of Transport" (1890-1935). See also Bus Facts, 1966 (34th. edition) p. 6 "Intercity Travel in the United States 1929-1969". Bus Facts was published by NAMBO = National Association of Motor Bus Operators.
[BusFacts-1948]: Same as above but for 1948
[Census-1912]: U. S. Department of Commerce and Labor, U. S. Bureau of the Census: Central Electric Light and Power Stations and Street and Electric Railways, 1912. Washington: Government Printing Office (GPO) 1915 (series title of: "Special Reports"). This includes data for 1902, 1907 and 1912 and represents a merger of the two special census reports: 1. Central Electric Light and Power Stations. 2. Street and Electric Railways. They were issued separately in 1905 (1902 data) and 1910 (1907 data) but merged in 1915 with 1912 data included.
Not all the information from the 1905 and 1910 editions made it into the merged volume in 1915. For some 1902 data see also "Abstract of the Twelfth Census of the US" pp. 385+, 405+. US GPO 1904. For some 1902-1917 (and 1918) data see: "Proceedings of the Federal Electric Railways Commission" US GPO 1920, vol. 3 pp. 2222, 2229, etc.
[Energy in Am. Economy]: Energy in the American Economy, 1850-1975 by Sam H. Schurr and Bruce C. Netschert. The John Hopkins Press, Baltimore MD, 1960.
[Hilton]: The Electric Interurban Railways in America, by George W. Hilton; Stanford University Press, 1960. p. 186 shows line mileage from 1889-1959.
[Hirst]: Energy Intensiveness of Passenger and Freight Transport Modes 1950-1970 by Eric Hirst. Oak Ridge National Laboratory (then part of the U.S. Atomic Energy Commission, now part of the U.S Dept. of Energy), 1973 (ORNL-NSF-EP-44). This 39 page report is on the same topic as this article. With a limited scope of only a 20 year snapshot within the 20th century, it utterly failed to pick up the long term trend of increasing energy efficiency. In fact, per Fig. 3 (p. 21) it looks like all energy efficiencies are declining except for railroads.
[ICC-Graph]: "ICC Graphical Supplement to Monthly Reports, Series 1937 no. 4. Interstate Commerce Commission (ICC), 1937. Chart: "Steam Railway Traffic in Relation to Population, Years 1890-1936"
[Stat. Abstract 1930]: Statistical Abstract of the United States, (US GPO) 1930, p. 423, Table No. 459: "Electric Railways". This has data for 1907-1927 (in 5 yr. increments). But footnote 4 indicates a serious problem with kwh figures prior to 1927).
[Chomitz]: "A Survey and Analysis of Energy Intensity Estimates for Urban Transportation Modes" by Kenneth M. Chomitz (Institute of Transportation Studies, University of California, Irvine, California 92717, December 1978.
[TransitFacts]: "Transit Fact Book" 1943- (annual) but has data back to 1920. By the American Transit Association (ATA). Starting with the 1974-75 edition, it was by the American Public Transit Association (new author). In 2002 it was "Public Transportation Fact Book" by the American Public Transportation Association (new name).
[TEDB]: Transportation Energy Data Book (annual) by Stacy Davis, Oak Ridge National Laboratory, U.S. Dept. of Energy. The most significant table is: "Energy Intensities of ... Passenger Modes" in Ch. 2. The latest edition is on the Internet. Ch. 2 is at Transportation Energy Data Book, Ch. 2.
[Trans in Am.]: "Transportation in America" (annual) by the Eno Transportation Foundation, Washington DC. (Formerly Transportation Facts and Trends, by the Transportation Association of America) Some of their "statistics" are now found in the "Statistical Abstract of the US"). The 18th edition of Trans. in America is available with a "Historical Compendium 1939-1999" which covers years that other editions ignore. The table "Domestic Intercity Passenger-Miles by Mode" shows the modal split for intercity travel (but excludes international air travel).
[UrbanTransAndEnergy] "Urban Transportation and Energy: The Potential Savings of Different Modes", Congress of the United States, Congressional Budget Office. Washington DC, US GPO, Dec. 1977. This uses data from other older reports, so it uses the poor fuel efficiency autos before federal mandates made them improve. It presents low, medium, and high estimates which often differ by a factor of 3. That's very confusing. But the Appendix contains much useful information.
This article's title, "Fuel-Efficiency ...", could also be called "Energy-Efficiency ..." since most all of the energy for passenger travel comes from fuel. Fuel-efficiency expresses efficiency in units such as passenger-miles per gallon of fuel. However, energy-efficiency would use units such as passenger-miles per million BTU, where BTU (British Thermal Unit) is a measure of heat content. Since there are about 125,000 BTU in a gallon of gasoline, knowing fuel-efficiency allows the immediate calculation of energy-efficiency and conversely.
Energy-intensity is just the inverse of energy-efficiency. Energy-intensity of travel is often expressed in units of BTU per passenger-mile. To get energy-intensity from energy-efficiency just divide 1 by the energy-intensity.
Thus the 3 different names: fuel-efficiency, energy-efficiency, and energy-intensity are basically the same concept (mean the same thing), but are expressed in different units of measurement. Of course, outside of the United States, "kilometers" will be often used instead of "miles" and heat values are more likely to be in kilo-calories (kcal) instead of BTU. --------------------------------------------------------------------------
See: Elements of Railroad Engineering by William G. Raymond. John Wiley, New York, 1923. p.157: Based on tests in 1904, simple freight locomotives consumed from 3.5 to 5 pounds of coal per horse-power hour. 4 lbs/hp-hr is only an efficiency of about 5% assuming a heat content of coal of 13.1k BTU/lb (US Bureau of Mines). Internal combustion engines have an efficiency of about 25%. Both efficiencies are at steady speed and full power. Under actual operating conditions, the actual efficiencies will be significantly less.
The approximate 6-fold increase in pass-mi/gal (equivalent) when railroads converted to diesel, tends to support this.
The heat rate is the amount fuel used to generate a kilowatt-hour (kwh) of electricity. It's often given in units of BTU/kwh. But in olden days (before 1920) when coal was the principle fuel used, it was often given in units of lb-coal/kwh. For 1900, I've used the estimate of 6.0 lb-coal/kwh. This includes consideration of hydro-electric power so the figure for fossil-fuel generation would be higher at perhaps 7.25 lb-coal/kwh. Today for fossil-fuel power plants, it's equivalent to between .8 and .9 lb-coal/kwh. But about 10% of the power is hydro-electric today which makes it between .7 and .8 lb-coal/kwh. This makes todays electric generation about 8 times more efficient.
In 1900, the electric railways generated most of their own power instead of purchasing it from central power stations. So the next sections covers railway-owned power stations while the section after that covers central power stations. Since certain data about electricity generation by central power stations, is absent from the data for electric railways, it's important to examine both types of power generation. After about 1923, the electric railroads purchased most of their power from central power plants, etc. See [Stat. Abstract 1930]. By 1927, 68% of power was purchased.
For 1900, I estimate 6.0 lb-coal/kwh for such power stations. This estimate is obtained from estimates using [Census-1912] for 1902. Based on the amount paid for "hired power" and compared to the cost of generating their own power, it looks like about 85% of electric railway power was self-generated. So we need to find the lbs-coal per kwh for this generation (and assume that the efficiency of generation of purchased power was about the same). While the census data gives the kwh generated it doesn't give the amount of fuel used but only the amount paid for the fuel. So use the price that the Central Electric Light and Power Companies paid for coal ($2.06/ton, calculated) and assume that the electric railways paid the same. This gives 5.5 lb-coal/kwh. To bring this figure to 1900 (from 1902), add .4 per linear interpolation, resulting in 5.9. See [Ayres]. Using linear interpolation on the 6.85 figure reported by Ayres results in 7.25 lb-coal/kwh for 1900 (coal fueled central power plants) and not 5.9. But the 7.25 figure ignores hydro-power so "compromise" at 6.0
The discrepancy (between 7.25 and 5.9) may be in part due to:
1. Some of the generation by the railways was by hydro-electric which
resulted in less coal consumption. In this case, the electric
railways get credit for hydro-power. Except that if purchased
power) had a higher hydro content, it's not accounted for here.
2. The electric railroads may have generated power more efficiently than
the Central power stations. The amount of power generated by the
electric railways was about the same as by the central stations, so
one can't use an "economy of scale" argument to imply that central
stations should be more efficient. Since the electric railroads
were a newer industry than central power stations, they may have had
more modern (and more efficient) generators.
Note that if we accept the 7.25 figure (no hydro) and assume that 1/3 of the energy was generated by hydro-electric power (the estimate for central power stations see the next subsection), then the result is 4.8 lb-coal/kwh. So using 6.0 implies that the use of hydro power by the electric railways was less than average. But this seems logical since the primary business of electric railroads was not to generate electricity. So they likely were less likely to develop hydro resources than those whose only business was to generate and sell electricity. Industrial users that needed cheap electricity would be expected to locate where cheap hydro-power was plentiful but electric railroads had to locate where people needed transportation.
Another dilemma is this. Is it fair to count hydro so as to increase the energy-efficiency statistic? One could ask: How much coal would have been saved had there been no electric railways? Good sources of hydro power are limited and thus the existence of the electric railway doesn't stimulate the development of hydro power by all that much. If the electric railways hadn't existed, then there would have likely been a higher percentage of electricity generated by hydro. Thus would have resulted in a savings in coal well beyond that saved by electric railways not consuming coal via electric power. In the short run, if an electric railway decreases it's load, then the reduction in generation will come entirely from coal power, since hydro generation is almost free (once it's in place). There's no clear answer to this dilemma but the estimate of 6.0 lb-coal/kwh does take hydro into account (although perhaps not fully as explained above).
The "central power stations" included locations where hydro-electric power was generated. They didn't include generation by non-electric companies such the electric railways.
For the efficiencies of electric power generation (given in terms of energy-intensity and called a "heat rate") see Historical Statistics of the United States, Colonial Times to 1970, by the U.S. Bureau of the Census, Washington DC, 1975. Table S 95-107. This only goes back to 1920 and fails to show the heat rate for 1920-1924, but this may be estimated from other data in this table. Coal was the dominant fuel in the 1920s and the pounds(coal)/kwh is shown for 1920 onward.
Before 1920, an estimate for 1900 is in [Energy in Am. Economy] p. 180 (bottom). For 1900, it claims 6.85 pounds(coal)/kwh, but there's no explanation of how this was derived. The same figure is shown for 1902 in: [Ayres] p. 139, Fig. 1. This shows lbs/kwh as a graph between 1902-1950. Again, no source is given but the 1902 date is a dead giveaway that it's estimated from the 1902 Census data. So if it's 6.85 for 1902, linear interpolation would give 7.25 for 1900.
Another source for 1902 is "Entering the Seventh Decade of Electric Power" by C. E. Neil in the "Edison Electric Institute Bulletin", Sept. 1942, p.6. He claims a heat rate of 92.5K BTU/kwh which gives 7.06 lb-coal/kwh using 13.1 kBTU/lb-coal. It's about the same as Ayer's 6.85 figure and likely from the same source.
In the census data for central power stations: The tons and cost of coal burned is shown. For other fuels, only the cost is shown so that one may make an estimate of the BTU by assuming a price per BTU equal to that for coal. Since coal fuel dominates, errors for the other fuels are not all that harmful. Another major problem is that the aggregate kwh shown includes both hydro and steam generated power. What percent is hydro?
In the census of "Electric Light and Power Stations" the installed capacity is shown for both hydro and steam generation. If one assumes that the split between hydro and steam is 24% vs. 76% (directly proportional to the installed capacity for 1902) then one obtains a heat rate of 5.9 lbs/kwh for coal. This assumption for the split is likely erroneous since steam is used for peak loads. Much of the steam capacity would only be used during the evening peak loads while hydro might be better utilized (except for drought periods). Peak loads then were likely more extreme than they are now since there was a higher preponderance of lighting which mostly happened in the evening hours. In order to get the 6.85 lbs/kwh for coal plants, 34.4% of power generation should be hydro (not 22%). This implies (for central power stations) that the utilization of the installed steam generating capacity was only 60% of that for hydro. This may be reasonable since the generators (both hydro and steam) only generated 23.5% of what they would have, had they been generating 24 hours/day at full capacity (assuming a 90 % efficiency of converting mechanical work to electricity since the installed capacity was given in horsepower and not in kilowatts).
In 1900, the electric railways got about 7.5 pass-mi/gal [21] and the steam railroads perhaps 5.7 [10]. A weighed average (based on distance traveled) gives 6.3 pmpg (harmonic mean). See [6] and [ICC-Graph] for distance.
For 2000 TEDB implies about 35 pass-mi/gal for the auto (using 124 k-BTU/gal). Small trucks (including SUVs) got 28.2 pmpg. Taking the harmonic mean (weighted by distance traveled) results in 32.7 pmpg for passenger transport in autos, SUVs, and small trucks. See Highway Statistics 2001, Table VM-1 Since 32.7 pmpg for autos in 2000 is about 5 times 6.3 pmpg in 1900, energy efficiency today is about 5 times better than in 1900 (neglecting non-auto modes). But including the non-auto modes doesn't change things much, as will be explained shortly.
For air, see [13]. which estimates 38 PM/gal for 2000. Note that per [TEDB] Table 2.11: "Passenger Travel and Energy Use" (in Edition 22), PM by air is about 1/8 of that by auto-SUV-air so take the harmonic mean weighted average of 38 PM/gal for air and 32 PM/gal for auto-SUV yielding 33.3 PM/gal for auto-SUV-air.
General aviation (private planes) per TEDB only gets about 9.5 pmpg but only constitutes 0.6% of intercity travel per [Trans in Am.] Intercity bus travel is claimed to obtain almost 130 pmpg and provides 2.2% of intercity travel. But the weighted harmonic mean of these is 35pmpg which implies that these two almost cancel out each other.
Since rail and bus transit, as well as Amtrak, get roughly the same pmpg as the auto (at 35 pmpg), little is changed by taking them into account. Urban buses only get about 26 pmpg while Amtrak gets about 41 pmpg but only constitutes 0.6% of intercity travel.
So the final resulting energy-efficiency today comes out to be 33.4 PM/gal which is a little over 5 times the 6.3 PM/gal estimated for 1900. Of course, one could say "roughly 5 times" since the estimates used for 1900 are rough and the "5 times" depends on these estimates.
The numbers in this paragraph follow from data in the previous and next paragraph of the main part of this article (which have citations). While total travel is 200 times greater (in passenger-miles), each pass-mi traveled uses only 1/5 the fuel, thus implying that total fuel consumption has increased 200/5 = 40-fold.
5. (not used)
The average travel distance is pass-mi divided by the population in 1900. To get pass-mi, multiply the estimated pass-mi/kwh (4.5) [21] by the kwh used. Unfortunately, the electric railway census data for 1902 [Census-1912] only gives the kwh generated, ignoring the kwh purchased and the kwh sold. Electric railways often sold some of the power they generated. But the census does show the cost of kwh purchased and the revenue from kwh sold. If one can estimate the price (per kwh) for buying and selling, then one can then find the number of kwh bought and sold. Then kwh (used) = kwh (generated) + kwh (purchased) - kwh (sold). The estimate using the above method is (in billion kwh): 2.37 (used) = 2.26 (generated) + 0.34 (purchased) - 0.23 (sold). Note that most of the power was generated so that errors in estimating the prices are not all that significant.
For the selling price assume that the electric railways sold power for the same price that the central power plants charged (on average). This was 3.34 cents/kwh and included all types of customers. For purchasing electricity, assume that they paid 30% more than the operating and maintenance costs of self-generation or 1.18 cents/kwh. The 30% is added to account for capital and administrative overhead. Since the electric railways were big customers, it seems reasonable that they would be able to purchase power at about half the price that they sold it for (mainly to commercial and residential customers). More research is needed on this.
The result of the above is (4.5 PM/kwh x 2.37 Bkwh)/76 M (population) = 140 miles/year per person of travel by electric railways but this was for 1902. Streetcar ridership was then rapidly growing and nearly doubled between 1890 and 1902. See "Proceedings of the Federal Electric Railways Commission" US GPO 1920, vol. 3 p. 2223, Chart C-107: "Riding Habit". Linear interpolation would decrease the 140 by 8% to make it valid for 1900. This gives 129 miles/year (about 130).
As a sanity check on this indirectly estimated value note per "Riding Habit" that in 1902 the typical person made 61 trips on electric railways. Dividing 140 miles by 61, results in an average trip length of 2.3 miles. For 1900, it would be a little less, say 2.0 miles. This figure seems reasonable considering that most of the lines were short city streetcar lines and not interurban lines.
The following estimate indicates that roughly 14% of the line length was interurban in 1900. [Hilton] p. 186: Table 6 "Statistics of Electric Interurban Railways 1889-1959" shows 2,107 miles in service in 1900. Compare this to the 21,902 miles of track for electric railways (both urban and interurban) in 1902 (see [Stat. Abstract 1930]) Assume it was 20,000 in 1900, based on linear interpolation of the total miles of track operated between 1890 and 1902. Since it's claimed that conversion to electricity was mostly complete by 1900 the change from 1900 to 1902 mostly represents construction of new lines (and not conversion of horsecar lines to electricity). Also note that in 1907 at least 2/3 of the line length was single track. [Stat. Abstract 1930] If the percentage of single track was the same in 1902, this implies that the 20,000 miles of track represented at least 15,000 miles of line. So only about 14% of the line mileage was interurban in 1900. (Is there better data in [Census-1912]?)
The above has estimated a statistic for 1902 that in 1902 was thought to be unknown. From the 1902 version of [Census-1912] p.9 it states: "It is quite possible that taking all street railways together, the passenger mileage, or number of passengers carried 1 mile, per mile of track operated, is larger to-day than it was in 1890, but statistics on this subject are wholly lacking."
The figure of 130 miles per person may seem low but in 1900 most people didn't live in cities that had streetcars and if they did, many didn't live near a streetcar line or didn't travel along the streetcar route or saved fares by walking or bicycling. See Population So most people who were actually using streetcars were using them much more than just 130 miles per year.
In 2000 the typical person traveled perhaps 16,000 miles. See Pocket Guide to Transportation by the Bureau of Transportation Statistics (US). Table: Per Capita Passenger Travel and Freight Transportation. For the 1995 NHTS survey one may add local travel to long-distance travel to get about 17,500. But the 2001 survey shows about 15,500 (no adding). Was there double counting in the 1995 survey? The US population is found in the Statistical Abstract of the US.
Another source is [TEDB]. Using table 2.11 from Ed. 22 and correcting to include international air travel per table 12.1, the result is about 16,000 miles/year per person. This is also roughly 50 times 330 miles (in 1900).
8. (not used)
For steam rail [ICC-Graph] shows 48 billion PM for 1920 and 28 for 1930, a 20 billion PM loss. (about 40%). For the auto, [BusFacts-1935] shows a 350 billion PM gain from 60 Billion in 1920 (about a 6-fold increase). [Stat. Abstract 1930] shows little change in the number of electric railway passengers between 1917 and 1927 so assume that the pass-mi was the same in 1920 as in 1925. Then per [BusFacts-1935] electric railway pass-mi declined from 50 billion to 40 (a 20% drop) from 1925 to 1930. While the number of passengers didn't decline that much, interurban mileage declined by 32% over the decade (per [Hilton]) so the decrease in trip length accounts for much of the decline.
Using "Bituminous Coal Annual", 1948, p.116, table 51 gives car-mi per ton of coal. Then multiply by the ICC's data on passenger-miles per car-mi. Then multiply by .8 to account for car-mi of dining and baggage cars, etc. This is based on the same methodology as used in http://www.lafn.org/~dave/trans/energy/PM-gal_36-63.html.
ICC's Bureau of Statistics, published an annual: "Revenue Traffic Statistics of Class I Steam Railways in the United States" (Statement M-220). Statistic Used: Passenger-miles per car-mile
Another source is: American Transportation in Prosperity and Depression by Thor Hultgren. National Bureau of Economic Research (NBER), Studies in Business Cycles, No. 3, 1948: p. 233, chart 94: "Passenger-miles per Ton of Coal or Equivalent Consumed in Passenger Service". This shows a lower fuel-efficiency than the above method, which is likely due to not subtracting the 25% share of fuel used to transport mail on passenger trains (based on car-mi of "other" cars). But increasing the PM/gal by 35% to account for this results in values that are somewhat higher than those by the 1st method.
The same statistics as the above (but not the same values) are found at the website: NBER statistics: Transportation and Public Utilities http://www.nber.org/databases/macrohistory/contents/chapter03.html The figures here have significantly wider variations from month to month than in Hultgren's book. The chart by Hultgren seems to be plotted with data points for every month, which implies that there should be no discrepancies in the data. But for some months one set of data may be 20 % higher than the other set. For other months it's just the reverse. For some months they almost agree. Could Hultgren have used a "moving average" where the value shown for each month is an average of values for previous months?
See also: US Interstate Commerce Commission, Bureau of Transport Economics and Statistics: Use and Cost of Railway Fuel and Problems in Fuel Statistics, Statement 4428 by Alexis P. Bukovsky (mimeographed 1944) p. 72, chart V: "Rates of Consumption of Fuel and Power by Locomotives and Rail Motor-Cars - Class I Line-Haul Railways, 1916-1943". This shows a decrease in energy-intensity from 18 lbs-coal per passenger car-mi in 1923 to only 14.5 in 1930. But between 1916 and 1923 it only decreased by 0.5.
For 1900, assume the the energy intensity was 20 lb-coal per car-mi and that there were an average of 15 persons/car. This is equivalent to 6 pass-mi/gal (after multiplying by the .8 factor). Between 1920 and 1935 the persons/car decreased from 19.76 (wartime conditions) to 10.93 (depression conditions). In 1925 it was 14.78. Also, the typical size of the cars was smaller in 1900. the pass-mi/car-mi is multiplied by .8 to account for the dining and baggage cars, etc. Conversion between coal and gasoline is at 9.3 lb-coal per gal of gasoline. The result is 5.7 pass-mi/gallon.
Prior to 1940 there is no survey data. However, there is a lot of anecdotal reports, many of which are on the Internet. A major problem is that some of these are taken from advertisements by the auto makers and thus are suspect for being hype. Also, examples that people mentioned in print likely tend to be from people who were bragging about what good mileage they got. They were likely from people who took better care of their cars and drove better so as to get better mileage than the typical driver.
1900-1910: The Antique Automobile Club of America has a website which claims 35 mpg for 1900. For the 1903 Cadillac (10 hp, 1 cyl., max speed 30 mph), 25 mpg is claimed. Roads were mostly dirt which contributed to poor economy. So in the early years of the auto it could be 20 mpg, about the same as a hundred year later (2000).
1910-1930: Per Antique Automobile Club of America: 20 mpg for 1919 and also for the Ford model T. But a website by UAW Local #387: www.local387.org/ford_model-t.htm says it was 13-21 mpg for the model T (from 1909 to 1927). This seems more realistic. Since Ford was considered to be more economical than average, it seems that for the 1910-1930 period it was under 15 mpg. In 1915 a Studebaker got 13 1/2 mpg coast-to-coast over the gravel Lincoln Highway, but other cars did worse (Studebaker News, Oct. 1915).
1920-1950: see [Ayres]. He claims p. 131 that the mi/gal for automobiles has shown no improvement from 1920-1950. But he suspects that there was higher automobile occupancy in earlier times. He claims on p 132 that "Statistics on miles per gallon are unreliable and confusing." He says that the records of car fleets show lower mi/gal than national mi/gal estimates.
1940-1955: [Energy in Am. Economy] p. 656, Table D-37 which shows per the Bureau of Public Roads (in car-mi/gal): 1940: 15.3, 1945: 15.0, 1950: 14.5, 1955: 14.5. Also reported here are somewhat higher estimates by someone from General Motors in a 1957 paper.
1940-2000: US DOT: Federal Highway Administration: Highway Statistics. See issues for 1985, 1995, etc. that summarize previous years. For example: "Highway Statistics, Summary to 1985 pp. 229-232 Table "Annual Vehicle-Miles of Travel and Related Data 1936-1985". The "data" is obtained by estimates from individual states and there is apparently a lag of a number of years before improvements in fuel-efficiency fully show up in the tables. This is because states sometimes use fuel consumption figures to help estimate vehicle-miles, using the mpg figures reported in highway statistics. If all estimates of vehicle-miles were made based on fuel sales, then all changes in fuel-consumption would be erroneously attributed to changes in the number of vehicle-miles (and not to changes in mpg). Then the mpg figure reported would never change (even thought actual mpg did change). It's of course not this bad, but this explains the reason for the delay. The dramatic improvement in mpg in the mid and late 1970's didn't show up in the statistics much until the 1980s.
1970-2000: See [TEDB] The BTU/PM for automobiles prior to 1990 are too low due to using the erroneous data on automobile occupancy supplied by NPTS [22] I believe that there hasn't been much change in automobile occupancy since the 1970s so use the figures on BTU/vehicle-mile for trends over time. See [14]
To get pass-mi, multiply by the persons/car (average). Today it's about 1.6. In the past it was likely higher but not nearly as high as claimed by NPTS due to double counting as explained elsewhere. If it were 1.67 (1 2/3) in the 1950s the result would be about 25 pass-mi/gal.
Sources of "statistics":
1935-1955:
[Energy in Am. Economy] p.
550
1950-1970:
[Hirst] pp. 21, 34-5
1970-2000:
[TEDB]
A major error made by TEDB is the neglect to allocate part of the aircraft fuel to cargo (mail and freight). If one uses [TEDB], edition 22, 2002, table 12.1 and assumes that one ton-mi of cargo is equivalent to 4 pass-mi one obtains the equivalent of almost 38 pass-mi per gallon of automobile gasoline for the year 2000.
In comparing the energy-efficiency of air travel with the auto, it should be compared with intercity automobile travel which is more energy efficient due to higher vehicle occupancy and higher vehicle miles/gallon.
A major problem is how to allocate aircraft fuel between freight and passengers since passenger flights often carry mail and freight (other than the baggage of passengers). [Hirst] p. 35, considered 400 lbs. of mail/freight equivalent to one passenger. This may be too low so I used 500 lbs.
See also: [Energy in Am. Economy] p. 550, Table XXVI: "Petroleum Products Consumed Compared to Work Performed by Aircraft, Selected Years 1935-1955". Since passenger travel on jet aircraft in 1955 (and prior) was insignificant, one many neglect the reported jet fuel use and assume that the jet fuel reported was almost all for military aircraft. The high efficiency for 1945 was due to planes being full due to wartime conditions. International flights were significantly less energy-efficient prior to 1955, but by 1955 there was only a little over 10% difference.
Here are the pass-mi/gallon (of aviation fuel figures) based on one ton of cargo being equivalent to 4 passengers for fuel consumption purposes.
Passenger-miles per gallon (aviation gasoline)
Year 1935 1940 1945 1950 1955
Scheduled Domestic 12.3 16.8 27.5 21.0 23.3
Scheduled International 7.6 11.7 19.8 16.5 20.9
The reason why international flights were less energy-efficient than domestic flights is because long distance non-stop flights (such as trans-oceanic international flights) must carry a huge load of fuel. On a long flight, the weight of the fuel needed for each passenger may weigh ten times as much as the passenger. Today, in part due to higher load factors on international flights, there isn't much difference in energy-efficiency between international and domestic flights. In the early days of aviation, domestic flights often made intermediate stops for refueling which reduced fuel use due to lower fuel weight while airborne.
See [TEDB] The TEDB figure indicates (for 1990's) that the auto is more fuel-efficient than transit while showing intercity rail to be much more efficient than the auto. Both of these are misleading. Splitting the auto data into 2 parts (urban and intercity) will show that the auto is roughly the same energy-efficiency as mass transit and intercity rail.
My estimation for autos uses: miles/gal: urban 20, intercity 26; persons/auto: urban 1.4, intercity 2.0. BTU/gal is 125,000. This gives: urban 28 PM/gal; intercity 52 PM/gal; average 35 PM/gal. Note that intercity is about 50% higher than the average value of 35 PM/gal.
These numbers are estimated based on:
1. For mi/gal: EIA estimates about 22 mi/gal in a recent pamphlet [TEDB] used 21.6). "Highway Statistics, 1990" by FHA, DOT in table VM-1 shows about 1/3 of vehicle-miles is non-urban. Looking over EPA tables of mi/gal leads to the estimates of 20 (urban) and 26 mi/gal (intercity) which averages to about 22 when weighted by urban/intercity vehicle-miles. The "average" should actually be done by using the harmonic mean method but the error introduced by a using the arithmetic mean is small compared to other errors.
2. For persons/auto the "1990 Nationwide Personal Transportation Survey" (NPTS) by FHA, DOT (p. 7-5) shows 1.62. pass-mi/veh-mi for autos. Using a figure of 1.6, this could be split up into 1.4 for urban and 2.0 for intercity similar to the above. This split is based on the following:
Old (1970) studies for Los Angles and San Francisco give persons/auto as 1.44 (studies agreed). For work trips, the Los Angeles study showed 1.10 persons/vehicle while for the entire US this figure was 1.17 (US Census of Population 1970). For 1990 the US figure is 1.10. So one may reason that Los Angeles was 20 years ahead of its time regarding auto occupancy and that the rest of the nation in 1990 was about what Los Angeles was in 1970. So this implies that one might use 1970 data for Los Angles to approximate what the US was in 1990. Doing this gives an occupancy of 1.44 which is rounded to 1.4 since: 1. Such reasoning is only a crude estimate 2. The author has personally sampled occupancy and thinks that 1.44 may be too high.
3. Using 1. and 2. above and 125,000 BTU/gallon for gasoline gives:
overall auto BTU/PM = 125,000 / (22 x 1.6) = 3,550
intercity auto BTU/PM = 125,000 / (26 x 2.0) = 2,400 (about 33% lower)
urban auto BTU/PM = 125,000 / (20 x 1.4) = 4,460 (about 25% higher)
USA Railroad Passenger-Miles per Gallon 1936-1963 by David S. Lawyer (the author)
See: [Trans in Am.] "Intercity Travel by Modes". In 1955, rail had only 4.0% of the intercity travel market. Per "Yearbook of Railroad Facts" (annual). Association of American Railroads, in 1955 there were about 6000 steam locomotives vs 31,500 diesel-electric units. By 1960, there were only 261 steam locomotives in service.
For the fuel-efficiencies see: "Amtrak and Energy Conservation in Intercity Passenger Transportation" by Stephen J. Thompson (Congressional Research Service, Report to Congress) Sept. 3, 1996. (http://www.cnie.org.nle/eng-11.html) See also [TEDB]. This shows the energy of Amtrak to be only about 8% better than the automobile (for 2000). Note that in 2001 per [TEDB] Amtrak's fuel efficiency significantly dropped making the auto about 15% more energy efficient.
But the automobile figure is for both city and intercity driving. To fairly compare it to Amtrak, one must correct the auto figure so it reflects intercity use. For intercity, there are about 2 persons/auto as compared to 1.6 overall, and the auto gets more miles/gallon (about 26 as compared to 22 overall). [14] After making this correction, Amtrak appears to be a little less-efficient that the auto for 2000 and even worse in 2001.
See also CRS Report: 96-22 - Amtrak and Energy Conservation: Background ... This 1996 update by the Congressional Research Service, agrees with the above conclusion that Amtrak energy-efficiency is about the same as the auto. The Amtrak energy intensity they used from "Transportation Energy Data Book" was too low due to the failure to consider that each BTU of electricity takes a few BTUs of fuel to generate. This resulted in the energy-efficiency of Amtrak's electric trains being grossly overstated. But while the data may be flawed, the conclusions of the CRS report turn out to be the same as mine.
The Transportation Revolution, 1815-1860 by George Rogers Taylor, (Vol IV of the Economic History of the United States) Holt Rinehart and Winston, New York, 1951: p.74 claims that in 1840 Europe had 1,818 rail miles vs. about 3000 for the US. Table on p.79 shows rail mileage. For mileage in 1920 see American Railroads by John F. Stover. University of Chicago Press, 1961, p. 224 (table). This info is also in Yearbook of Railroad Facts (annual), Association of American Railroads, Washington, DC. Statistical Abstract of the United States, 1930, p. 422 shows 32,548 line miles of electric railways for 1917 and 31,264 for 1922.
[Energy in Am. Economy] p.84+: "Oil, The Beginnings of the Industry".
See: American Railroads by John F. Stover; University of Chicago Press, 1961, p. 162. In the 1860s major railroads were switching from wood to coal as a fuel. See also [Energy in Am. Economy] pp. 52, 59. Wood was the major fuel of railroads until about 1870.
See "Panorama of Transport" (an annual) by Eurostat (of the European Commission, part of the European Union, EU). Luxembourg, 2003. A figure on p. 33 shows automobilisation (or motorisation) for both Western Europe and the U.S. On pp. 69-70 the modal spilt (in pass-km) is shown for 1970-1999.
A major source of data in English is:
EDMC Handbook of Energy & Economic Statistics in Japan by The Energy
Data and Modelling Centewr, The Institute of Energy Economics, Japan
and published by The Energy Conservcation Center, Tokyo, Japan. The
2001 edition was used for this section: p. 110 (Volume of
Transportation by Mode), p. 112 (Energy Intensity by Mode)
Travel by truck and by sea are not shown. In 2000, the sum of these modes was only about 1% of total travel.
Billion Passenger-kilometers
Year Auto Bus Rail Air Truck Sea Total
1965 62 80 235 3 11 3 416
1970 277 103 289 9 19 5 703
1975 329 111 324 19 16 7 804
1980 414 110 315 29 16 6 891
1985 511 105 330 33 16 6 1000
1990 727 110 385 52 16 6 1296
1995 806 97 400 65 14 6 1388
1999 858 88 385 79 12 5 1428
The second Rail figure is just the first one mutliplied by 3 to account for the fuel needed to generate electricity for electric railroads. The first "Rail" figure is the one reported from Japan. The EMDC data source from Japan notes in the "Explanitory Notes" that the efficiency of electric power plants was taken into account for the "Energy Balance Table" but doesn't mention the transportation Energy-Intensity table. The fact that the "Final Energy Consuptionnn by Source" on p.39 fails to account for power plant efficiency tends to imply that the transportation Energy-Intensity table (p. 112 in the 2001 edition) didn't do it either.
kcal/passenger-kilometer
Year Auto Bus Rail Rail Air Truck Sea
1965 561 113 57 * 1507 1354 191
1970 373 116 48 144 1026 1012 220
1975 518 129 45 135 870 1343 203
1980 544 122 48 144 795 1151 214
1985 518 125 46 148 705 1341 174
1990 488 145 48 144 550 1528 202
1995 560 155 49 147 569 1257 253
1999 584 165 51 153 441 1854 488
* For 1965 it's assumed that a signficant part of the energy came from diesel trains. Thus multiplying by 3 is not valid. From 1965 to 1999, the average reported energy intensity steadily increased from 190 kcal/pass-km to 417 kcal/pass-km. But the reported value is somewhat in error due to neglect of the losses in generating electricity for electric trains. So the actual relative increase is not quite as great as the above would imply.
For 1900, 7.5 pass-mi/gal is based on 4.5 pass-mi/kwh and a generation energy intensity of 6.0 lb-coal/kwh. [2] with 1 gal. gasoline = 10 pounds coal (in heat value). See [23]
The 4.5 pass-mi/kwh is a rough estimate, based in part on the figures of 4.87 for 1927 and 4.62 for 1930. These two values are calculated based on the kwh reported by American Transit Association's: "Transit Fact Book", 1948, p.27: "Electric Power", and the passenger-miles reported in [BusFacts-1948]. In 1900, most of the the electric railways were short and not interurban. Therefore they made more frequent stops than in later years and likely used a little more energy per passenger mile. At the same time, some early streetcars tended to be light and small which would tend to make them more fuel-efficient.
Another estimate of pass-mi/kwh (for the 1970's) is [Chomitz] On p.29 Table 8, one can calculate: PM/kwh = 6.1 (Cleveland), 2.0 (Philadelphia, "Red Arrow") and 1.9 (TNJ, Newark, New Jersey). It's mentioned that the Cleveland line is a "showcase" system and is not typical but that the other two use heavy cars and are likely not typical either. So an average would be about 4 which is not far from the 4.5 used for 1900.
For 1970, 4 PM/kwh is about 44 PM/gal at the power plant heat rates between 1960-2000. This is about double that of the automobile prior to the 1970s (14 mi/gal with 1.7 people). Since many electric railways were interurban, a comparable auto trip would consist of both urban and rural driving and thus overall auto energy-efficiency is used for this comparison.
As a sanity check, the separate edition for 1902 of [Census-1912] p.229 shows an average of 2.14 kwh/car-mi based on 307 selected electric railways. Multiplying this by the assumed 4.5 PM/kwh results in 9.6 passengers per car (on average). Given that streetcars were much smaller in those days, and that the streetcars were apt to be nearly empty when they started from the end of the line near the outskirts of the city, 4.5 PM/kwh seems reasonable.
Sources of "statistics":
1907-1927
[Stat. Abstract 1930]
1920-
[TransitFacts]
1925-1935:
[BusFacts-1935]
1950-1970:
[Hirst] pp. 14, 37
1970-2000:
[TEDB] for "Rail transit"
Only the [Hirst] and [TEDB]report energy-intensity. [BusFacts] only reports Pass-mi which allows calculation of energy-intensity using TransitFacts and the power plant heat rates. It might seem that we have better data for 1925-35 and 1950-2000 but this is not necessarily so as will be explained.
For 1950-2000 the following are reported per above in BTU/Pass-mi
Hirst TEDB
1950 3900 1970 2453 (Note discrepancy for 1970)
1955 3800 1975 2962
1960 3900 1980 3008
1965 3900 1985 3461
1970 4100 1990 3453
1995 3818
2000 3105
Note that for 1970 [Hirst] reports 4100 but [TEDB] reports 2453. This discrepancy indicates that the values shown above may not be very accurate. What is the reason for this discrepancy? Both used [TransitFacts] as their source of information so there shouldn't be any discrepancy. However in 1970 [TransitFacts] only reported energy use and non Pass-mi (but did report the number of passenger). Thus to use their data to get say Pass-mi/kwh one needs to estimate the average trip length. The discrepancy is mostly due to differences in estimated trip length.
But now, let's look at the New York City subway system (heavy rail). In 1970 about 88% of electric railway energy was used by heavy rail per [TransitFacts] Table: Trend of Energy Consumption ... and about 75% of this was for the New York City subways per [Chomitz] p.24 (Table 4). Thus about 2/3 of the energy for the electric railways in the U.S. was used New York City.
For many years an average trip length for New York heavy rail was estimated (erroneously) to be 7 mi. This was used in "Public Transportation, Operational and Financial Status in the Tri-State Region, 1977" Sept. 1979 by the Tri-State Regional Planning Commission, One World Trade Center, New York, NY, p. 15 (Table I: Annual Public Transportation Trips ...), and p. 18 (Table IV: Estimated Passenger Miles Traveled ...).
The method of making this estimation is described in "City and Suburban Travel" Issue 162, 1975, pp 2-5: Passenger-Miles and Passenger-Trips. It's like the so called "gravity" model with all distances being the same. The explanation below has been slightly modified to insure flow conservation. The heavy rail system counts the number of passengers which enter each station at the entrance turnstiles, but doesn't know how far they travel. What percentage of passengers entering a given station (say station A) are headed for say station B? If we can estimate this for all stations then we know the flow of passengers between all pairs of stations and can readily find both passenger-miles and average trip length since we know the distance between all pairs of stations.
But how to estimate the percentage that go from A to B? Well, there's a simple (but erroneous) way to do this. Each station has a known market share which is the percent of total passengers which enter that station in a fixed period of time (say a year). For any station, assign the distribution of destinations to be directly proportional to the market share of each destination station. Let's say station B has 3% of market share. So we might assume that 3% of the people entering as all other stations go to station B. This method must be corrected so as to assign all flow since for passengers entering station A, the market shares of all other stations don't add up to 100%. So for station A, just find a correction factor which multiplies the market share of all other stations (other than A) so the the adjusted-market-shares of non-A stations add up to 100%. These are just the market shares of the other (non-A) stations if one neglects the existence of station A in calculating them.
But unfortunately, no consideration is given to the fact that in most cases, passengers have a preference for shorter trips so as to save time. Thus, the estimates obtained using the above method are likely to be too large. Such errors are especially relevant to systems such as New York where it takes time and effort to transfer from one train to another so as to go a longer distance.
So one hypothesis is that [TEDB] used the above distance estimate (at least for New York) trip length. [Hirst] seems to have used an average trip distance of only 3.5 miles (assuming he considered revenue passengers only). So which estimate is correct? One might expect that the true value is in between the two estimates. But there is still another consideration: the figures for the amount of electric energy used.
For New York heavy rail, it appears that the amount of energy use reported is that of the energy supplied to the traction substations less 10% to account for losses in conversion of AC to DC electricity plus the heating losses in the 3rd rail. This seems to correspond to what is shown in [TransitFacts] considering that about 75% of this figure is for New York City. However the amount of electricity uses for non-traction purposes (signals, escalators, ventilation, drainage, lighting, supplying transit offices, etc.) is about 10% of that used to supply the traction substations. Thus the total energy used is roughly 20% higher than reported. It's not clear what portion of this additional energy should be allocated to transportation of passengers. Both the [TEDB] and [Hirst] figures neglect this "overhead" 20% energy.
In view of this, the [Hirst] estimates may be closer to the truth than the [TEDB] estimates. More work on this topic of electric railway energy intensity is needed.
Automobile efficiency also depends on automobile occupancy. The US Federal Highway Administration publishes a "Nationwide Personal Travel Survey" (=NPTS) every 10 years or so. The first (called .. Transportation Study) in 1970 was seriously in error (on the high side) regarding automobile occupancy. Later surveys are also suspect. The 1970 study erroneously reported 1.4 persons/car (average) on trips to work while the US census reported 1.17. Just watching cars pass by on an early weekday morning indicated that the 1.17 figure was the correct one (the author actually did this in the 1970's).
Occupancy data from NPTS for 1977 and 1983 also appear to be in error by a factor of 2. For a comparison of NPTS to census data (but no explanation for the discrepancy) see: 1990 NPTS Travel Mode Special Reports. DOT, Federal Highway Administration, Dec. 1994. Fig. 3: Carpooling and Average Vehicle Occupancy". This figure shows that in 1990 the error disappeared and NPTS agrees with the census.
This error is likely due to taking a simple average of the number of occupants as reported by both drivers and passengers. To get a correct figure of the number of persons per auto, one should only sample the drivers. Even better, one should use a weighed average, assigning the weight 1/n to an observer who report n occupants in a car. But this was apparently not done until 1990. Thus the large errors.
For heat values, I've used the high heats of combustion, which were traditionally used in the past: 125,000 BTU/gallon for gasoline. For coal I've used: 12,500 BTU/pound (power generation) 13,100 BTU/pound (railroads). This results in 1 gallon of gasoline being equivalent to: 10 pounds of coal (power generation) or 9.5 pounds of coal (railroads).
For gasoline, this seemingly conflicts with the figure of 114,000 BTU gallon, reported by Chevron. See Fuel Economy of Gasoline Vehicles. But this discrepancy is more apparent than real.
What has happen is that there are two different values of BTU/gallon for the same batch of gasoline. One is the high (or gross) heat of combustion and the other is the low (or net) heat of combustion. The high value is obtained when, after the combustion, the water in the "exhaust" is in liquid form. For the low value, the "exhaust" has all the water in vapor form (steam). Since water vapor gives up heat energy when it changes from vapor to liquid, the high value is larger since it includes the latent heat of vaporization. The difference between the high and low values for gasoline is significant, about 8 or 9 percent. This accounts for most of the apparent decrease in the heat value of gasoline since Chevron is reporting the low heat of combustion. See Appendix B of Transportation Energy Data Book. They give 115.4k BTU/gallon as the low heat value for gasoline.
In "Fuels and Combustion Handbook" Ed. by Allen J Johnson, McGraw-Hill, New York, 1951. p. 364 it states that the higher heat value (gross) is the standard ordinarily accepted in the USA. Foreign countries tend to use the lower heat value (net). Would it be better to use the low values? Probably yes, since practical engines exhaust the water as a gas (vapor). The heat values for coal are also given as the high heats of combustion. One uses these heat values to determine how much coal is equivalent (in heat value) to a gallon of gasoline. I've used only the high heat values for both coal and gasoline because it's traditionally done this way.
But one may object to this. Coal is black and mostly carbon. If it were pure carbon, it would not generate any water vapor when burned. This is because a fuel must have hydrogen in it to form water which contains hydrogen (H2O). Thus the heat of combustion for a hypothetical "pure coal" (consisting only of carbon) would have only one heat of combustion. The high (gross) and low (net) values would be the same. So in this case it would be better to use the low heat value of gasoline for calculating a gasoline-to-coal equivalence.
Real coal does contain some hydrogen . But the difference between the high and low heat values is only say 3.5% . See the "Fuels and Combustion Handbook", op cit, p.367. Thus a better comparison would be to use the low heat values for both coal and gasoline. Doing this would make the old coal-based transportation economy of 1900 about 5% less energy-efficient than what I've estimated. But gasoline is a highly refined product as compared to coal so one could argue that we should increase it's BTU value to account for that. This would tend to cancel out the 5% bias mentioned above.
In 1950 the US Bureau of Mines adapted 13,100 BTU/pound for coal but today reports 12,000 to 12,500 BTU/pound. The U.S. Dept. of Energy claims that typical coal today (2000) is only about 12,000 BTU/pound. What was the BTU/pound of coal in the past for railroads and power plants? In footnote 33, p. 184 of [Energy in Am. Economy], it mentions that electric utilities used coal of a lower heat content (12,263 BTU/pound). The 13.1k BTU/pound value is mentioned in this footnote 33 as well as on p. 317 of [Ayres]. Some railroads used anthracite coal which Ayres (p. 317) shows at 12,700 BTU/pound. So I'm assuming that railroads in the past used coal of a higher heat value than power plants.
Per [14]
overall auto BTU/PM = 125,000 / (22 x 1.6) = 3,550
intercity auto BTU/PM = 125,000 / (26 x 2.0) = 2,400 <about 33% lower)
urban auto BTU/PM = 125,000 / (20 x 1.4) = 4,460 (about 25% higher)
Rail transit is an exception. But the energy data they report to the government doesn't include the energy used for lighting and air conditioning of stations. See National Transit Database Internet Reporting, Resource Module. Compare the 72.8 kBTU/vehicle-mi propulsion energy for rail transit (see p. 2-14, table 2.11 in [TEDB] for 2000) to the BTU/vehicle-mi reported for station energy (see p. 62, table A-3 in [UrbanTransAndEnergy]). It looks like station energy is roughly 10% of propulsion energy. This results in rail transit being 25% more energy-efficient than the auto.
The energy-intensities reported for transit may be too low due to possible overestimation of passenger-miles by the transit operating agencies. See NPTS 1990: "Limitations of Data on Transit". I have made no attempt to correct this since it's not clear how much it is in error. Since many persons using transit would travel alone if they went by auto, a fair comparison might use a load factor (persons/auto) of say 1.3 instead of the 1.4 I used. These two biases are in opposite directions and tend to cancel.
Another consideration when comparing the auto to transit is circuity. By auto, one usually travels by the shortest route, but since transit is not located everywhere, passengers often travel further than if they were to go via auto. This favors the auto by say 20%. However, "autos" above excluded SUVs so factoring in this consideration still favors rail by about 10%.
The British Economist, William S. Jevons in his 1865 book, The Coal Question, pointed out that increasing the efficiency of using coal only leads to more consumption of coal.
More recently, this was called the "rebound effect". See Energy Efficiency and the Rebound Effect Does Increasing Efficiency Decrease Demand? - NLECRS Report: RS20981. See also: Does Energy Efficiency Save Energy The first article above claims that increasing efficiency does save energy, but not as much as one would expect due to stimulation of demand. When energy-efficiency is increased there is the price effect and the income effect and both effects tend to increase energy consumption.
The saving of energy due to increased efficiency may lead to a decrease in the price of fuel, due to a decreased demand for it due to the higher efficiency of its use. The decreased price in turn tends to increase to consumption of that fuel. This is the price effect.
Also, since consumers save money due to the increased energy efficiency, they have additional money to spend on things that consume lots of energy (or took energy to manufacture). They might buy a SUV, go on an airplane trip, buy a hot-tub or a high-powered computer, etc. This is the income effect. Due to the savings from energy efficiency, consumers in effect have more disposable income to spend on energy.
Realistically, these effects may not be very pronounced. If a fuel is becoming more scarce as it is depleted, or if demand for it is growing for situations where the energy-efficiency measures were not applicable, then there may be little (if any) decrease in its price. If it costs significantly more to build energy efficiency into a product and the consumer must pay this cost. then the income effect could be minor (or even the opposite of what was described above).
But there are other circumstances that stimulate energy consumption. One is the income effect of technological "progress" resulting in higher productivity which means that people have more money to spend on energy. Another is technological and societal change which creates new "needs" which consume energy. For example, the invention of the railroad and automobile resulted in the creation of new travel needs which was set in concrete and steel by locating places of residence, employment and shopping such that a lot of travel is needed to travel to and from them.
One way to counter the factors simulating the consumption of energy is to place a high tax on it, or even more restrictive, rationing. But whether or not either of these two measures should be done and just how they should be done is beyond the scope of this article.
The thesis of Jevons has been insidiously used by those who seem to oppose energy conservation. The book "Why Energy Conservation Fails" by Herbert Inhaber only presents one side of the question and is thus biased. Yet many of the points he makes are correct. What's wrong with this book is what was omitted from it, such as suggestions for reducing demand by taxes and concern about future generations, etc.
Note that in the late 1980s, the fuel economy of new cars peaked and started to slightly decline thereafter. Nevertheless, energy efficiency of cars on the road continued to increase in the late 1980s and throughout the 1990s due to the fact that most new cars were still more efficient than the older ones that were being junked.
For 1940, Transit Fact Book says that no fuel data is available for Urban Buses. Thus estimates for 1920-1940 are missing. Compare the figures below with Lawyer's 4460 BTU/Pass-Mi estimate for the urban auto, and 2400 for the intercity auto (both for 1990). See [14]
Bus Energy Intensity in BTU/Pass-Mi
-Intercity- ------Urban-------
Year Hirst TEDB Hirst TEDB Lawyer
1945 2650
1950 640e 3100
1955 1100e 3400
1960 1500 3400
1965 1600 3500
1970 1600 1674 3700 2472 (note urban discrepancy)
1975 988 2814
1980 1082 2813
1985 964 3423
1990 962 (2400*) 3794 (4460*) * for autos per Lawyer
1995 964 4310
2000 932 4515
From 1975-2000 [TEDB] shows that the energy-intensity ranges from 932 to 1286 BTU/PM. There seems to be no clear trend. 1000 BTU/PM would be equivalent to 125 PM/gal of gasoline. The bus companies brag (on the Internet) about their energy-efficiency and quote figures in terms of diesel fuel, which results in a PM/gal about 10% higher than it would be in terms of gasoline (due to the larger heat content of diesel fuel per gallon).
The American Bus Association, claims 160 PM/gal for 1999. See ABA - Industry Facts But reducing this to gasoline give about 145 PM/gal. Now a large number of the members of this association are charter bus operators. Charter buses would seem to be more likely to be nearly full than scheduled intercity buses and thus get higher PM/gal. So if we think of intercity buses as including charter buses, then the energy-efficiency is higher. But if we consider only scheduled buses available to the public, then the energy-efficiency may be about 130 PM/gal.
The [Hirst] estimates cite ATA (American Transit Association) as a source. But the ATA's [TransitFacts] doesn't give pass-mi but only shows the number of passengers. Contacts with them reveal that they did not collect Pass-mi data then. However, the number of passengers is reported and Hirst seems to have multiplied this by an assumed 2.5 to 2.6 miles average trip length to obtain his results (assuming he used unlinked trips or total pass-mi). In 2000 APTA (new name of old ATA) p. 75 of 2002 "Fact Book" reports 3.7 miles for average unlinked bus trip length (no transfers).
But urban bus trips have become longer. One reason is that transit agencies often took over long Greyhound (etc.) commute bus routes which were previously considered to be "intercity" but are now classified as "urban". In some cases, the construction of Freeways (and availability of subsidy) led to establishing long bus routes on freeway. Urban sprawl contributed to longer trips. So it Hirst's estimate of about 2.6 miles for the 1950-1970 period may be reasonable, especially for the 1950s.
However Hirst assumed 136,000 BTU/gallon for bus fuel. But [TransitFacts] shows that prior to 1958, bus fuel was mostly gasoline which has 125,000 BTU/gallon. For 1950, about 4 times as much gasoline was used as diesel (at 137.5 BTU/gal) so the average was about 127,000 BTU/gal (and not 136,000 as used by Hirst). This makes the Hirst energy-intensities several percent too high for the 1950s but only a few percent too high for the 1960s as gasoline buses were being phased out. But making a correction for this error fails to eliminate most of the discrepancy.
So what explains the discrepancy for 1970: Hirst: 3700 BTU/Pass-mi TEDB: 2472 BTU/Pass-mi. Even though Hirst's work only goes to 1970 an extrapolation of it indicates that this discrepancy continues well after 1970. When the government began to collect estimates and data from mass transit agencies in the early 1970's what they sometimes got (based on my observations at that time) were estimates which were often biased to make the Pass-mi too high. As time went by, the government began to "require" more accurate estimates. Although this was likely not very well enforced, it did apparently result in better estimates since the reported energy-intensity has continued to increase since 1970. Thus the increase may be partly due to better estimates. The poor energy-intensity is partly due to poor ridership and also to subsidy which permits operations to continue with poor ridership.
See Ekonomicheskie Problemy Razvitiya Transporta (in Russian, translated title: Economic Problems of Transport Development), editor: A. A. Mitaishvili. Moskva, Transport (publisher), 1982. See table 7.2 (on modal shares) p. 78.