Skip to main content

SEPTA and the Economics of Urban Transportation

Josh Hausman '05

This paper examines and tests the view of public transportation presented by Arthur O'Sullivan (2003) in the textbook Urban Economics. To test O'Sullivan's assertions, I will refer to information on the Southeastern Pennsylvania Transportation Authority's (SEPTA's) public transportation system, particularly its commuter rail division. Time and space will limit me to two broad points. First, I shall consider O'Sullivan's argument that transit subsidies are distortionary and at best, a necessary evil.  Second, I shall consider O'Sullivan's discussion of how commuter rail ridership varies with changes in the price, frequency, and speed of service. Data from the Philadelphia area suggests that O'Sullivan's arguments are fundamentally biased against public transportation.

I. Public transportation subsidies: a necessary evil?
It is natural to begin any discussion of transportation with a discussion of externalities. Externalities are benefits or costs not mediated through the market system.  If public transportation reduces pollution, for example, public transportation has a positive externality; the reduction in pollution is a benefit, but it is not a benefit that affects the market for public transportation. Likewise, pollution is a negative externality of auto travel. Along with natural monopoly, externalities are one of the principal justifications for government provision or subsidy of public transportation. This paper will not discuss the less controversial natural monopoly justification, but will instead focus on transportation externalities. O'Sullivan's chapter on auto use is concerned with two negative externalities of private car travel: pollution and congestion. O'Sullivan convincingly argues that a congestion tax could effectively solve the congestion problem. If the tax revenue were properly redistributed (which may be difficult), then redistribution could make everyone better off (p. 264). To make drivers pay the social cost of the pollution they generate, he suggests a system of effluent fees or pollution taxes on new car purchases (p. 281). Yet there is an alternative to these taxes. Public transportation is a substitute for auto travel. Thus subsidizing transit should reduce the number of miles driven and bring the amount of auto travel closer to the social optimum, where the social optimum is defined to be the point at which net social benefit (total social benefit - total social cost) is greatest. However, O'Sullivan opposes transit subsidies. The following argument is typical of what O'Sullivan repeatedly asserts (p. 273):

The fundamental problem of the transit subsidy is that it underprices transit, increasing ridership above its optimal level. In the absence of congestion taxes, a transit subsidy may improve the efficiency of the transportation system, but the subsidy will never be as efficient as a system of congestion taxes. The policy question is whether the benefits of the transit subsidy (the diversion of drivers from underpriced and congested roads) is larger or smaller than the cost (excessive transit ridership).

O'Sullivan's neglect of numerous externalities is troubling. He considers congestion and pollution, but these are only two of many negative externalities associated with auto travel. Janet Pack (1992) does a rough cost-benefit analysis of SEPTA's commuter rail system. Unlike O'Sullivan, she considers the cost of the traffic accidents associated with auto travel. Ironically, O'Sullivan devotes an entire section to "Auto Safety," yet he never mentions that an obvious way to reduce traffic accidents is to decrease the number of miles driven (pp. 282-286). Because they reduce the amount that people drive, public transportation subsidies save lives. Using a detailed analysis by the Urban Institute of the costs associated with fatal and non-fatal traffic accidents, Pack estimates that by preventing traffic accidents, in 1989 SEPTA commuter rail provided a social benefit worth between 7.2 and 16.2 million dollars (p. 51). 

Externalities, and urban transportation externalities in particular, can be difficult to measure. Vukan Vuchic (1999) writes that "an automobile-based environment . . . makes areas less human-oriented and thus decreases the quality of social life, urban environment and sustainability" (p. 2). He cites New York, San Francisco, and Toronto as cities whose intermodal transportation systems endow them with "a high degree of efficiency and livability."  In these cities, not only can one travel by car; one also can conveniently travel by rail or bus (p. 9). Three cities are not proof, but it is likely that highways and parking lots do not create or maintain walkable, dense cities; conversely, it is likely that good public transportation systems do help create such cities. Thus the creation or maintenance of a livable city is a positive externality of a good public transportation system, albeit one that is difficult to measure. This sort of externality is qualitatively different from externalities such as pollution or congestion. Pollution and congestion decrease in proportion with the number of trips made by public transportation instead of by car. In contrast, the creation of a livable city is a positive externality associated only with a well-used public transportation system. Five hundred more or less riders on SEPTA does not affect Philadelphia's livability. The existence of SEPTA does.

There are other public transportation externalities that operate in a similar way. A good example is mobility for those who do not drive. Writers on urban transportation often strangely neglect the effect of an urban transportation system on those who cannot drive or who are too poor to own a car. Around 1990, twenty-two percent of households in the Philadelphia region did not have a car, and as baby boomers age, this number is likely to grow nationwide (Urban Institute, 1991, p. ES-5). Those without cars do not choose between transportation modes; rather, the existence or non-existence of public transportation determines whether or not they have mobility. Taxis can provide convenient transportation for non-drivers, but most people cannot afford to use taxis on a regular basis. For the elderly and handicapped in the Philadelphia region who cannot drive and cannot afford taxis, the loss of SEPTA service could mean that they "simply [would] cease going outside their homes" (Urban Institute, p. 4-12).  SEPTA service is also important for students of all ages in the Philadelphia area (Urban Institute, p. 4-13 - 4-14).  Thirty-six thousand Philadelphia school children use SEPTA to get to school every day (Urban Institute, p. 4-13). Without SEPTA, the Philadelphia public school system would be forced to spend more money on school buses, and their desegregation program would be endangered (Urban Institute, pp. 4-13, 4-15). Giving mobility to otherwise immobile people is a significant positive externality of public transportation. And quite apart from quantifiable social benefits, a transportation system that leaves a significant proportion of the population immobile does not meet basic definitions of fairness. O'Sullivan, however, is generally uninterested in considerations of equity or fairness; therefore, I too will focus on questions of economic efficiency.

Partly because of the increased mobility that public transportation gives many people, Vuchic (1999) and others contend that public transportation subsidies can increase regional economic growth and productivity. The research department of the Federal Reserve Bank of Philadelphia (2001) argues that transportation investments "can improve the functioning of the labor market giving individual workers access to more job locations" (p. 20). Highways, of course, only improve access to employment for those who own cars. Vuchic (1999, p. 7) and Richard Voith (1994, p.15) argue that public transportation helps high density areas function and thus facilitates the benefits that high density urban centers provide to businesses (so-called agglomeration economies). The Urban Institute (1991) concludes that the Philadelphia region would suffer a myriad of negative economic effects if SEPTA service were reduced or eliminated. Doing business in the Philadelphia region would become more expensive as the costs of automobile travel and truck transport increased (Urban Institute). Moreover, the Urban Institute notes that businesses would also suffer from reduced access to labor markets. Higher auto transportation costs also imply a higher cost of living for everyone in the region, and more money spent on cars and gas - goods produced outside the Philadelphia area (Urban Institute). Urban economics tells us that a reduction in spending on locally produced goods will be multiplied into a larger negative effect on a region's economy.

The existence of many externalities is no proof that the government should subsidize mass transit. The market for public transportation could be such that the government subsidies necessary to provide service would be much larger than the benefits of public transportation. Because it compares estimated social benefits from SEPTA commuter rail with the cost of subsidizing SEPTA commuter rail, Pack's (1992) analysis is useful. She considers five different social benefits: fewer automobile accidents, reduced commuting times, less congestion and pollution, benefits of reduced crowding for riders of other forms of public transportation, and welfare gains for commuter rail riders in excess of the fares they pay (so-called consumer surplus) (p. 51). Pack concludes that the net benefits of SEPTA commuter rail (the above listed social benefits - subsidies)  were between 38.8 and 74.8 million in 1989 (p. 51). SEPTA commuter rail receives by far the largest subsidies per trip mile of any SEPTA division, and thus Pack's results suggest that other divisions of SEPTA might have even greater net social benefits (Gambaccini, 2003). Because she neglects the effects of SEPTA commuter rail on the livability and productivity of the Philadelphia region, Pack's results may also significantly understate the net social benefits of SEPTA commuter rail.

Pack (1992) concludes that net benefits from SEPTA commuter rail grew throughout the 1980's; during the 1990's these net benefits probably continued to grow. In 1989 the subsidy given to SEPTA commuter rail was 98.9 million dollars (Pack, p. 51). In 2003 the subsidy was 108.3 million dollars (SEPTA, 2002). Ninety-eight point nine million 1989 dollars had the purchasing power of 146.6 million 2003 dollars (Minneapolis Federal Reserve, 2005). Thus adjusted for inflation, between 1989 and 2003, the subsidy to SEPTA commuter rail fell by over 25 percent. Although the subsidy fell, ridership remained roughly constant (SEPTA, 1998, p. 18). If social benefits are proportional to readership, this suggests that the net social benefit from SEPTA (social benefit - subsidy) increased during the 1990's.

The conclusion that SEPTA currently provides significant social benefits only implies that SEPTA should not be eliminated; it says nothing about what the ideal level of SEPTA service and the SEPTA subsidy should be. As noted above, questions of equity and fairness may (and I argue, should) also determine the level of public transportation service that society provides. If one's goal is economic efficiency, however, one should examine what would happen to benefits and costs with increased or decreased public transportation service. This leads economists to an in depth study of how public transportation ridership responds to changes in the price, frequency, and speed of service.

The next section will show that different estimates of these so-called elasticities of demand can result in quite different policy prescriptions.

II. Elasticities and Public Transportation
O'Sullivan states that demand for public transportation responds little to changes in the price and speed of service (p. 294). He reports that a good estimate for the price elasticity of demand is about -.33. This number means that if a public transportation system were to double its fares, it would lose only one third of its riders. Therefore, although doubling fares would not double fare revenues, it would raise them substantially. If there were x number of riders paying a fare of y before the fare doubled, then after the fare doubled there would be .67x riders paying 2y. The difference in total revenue would be:

(.67x)*(2y) - xy =
1.34x - xy =
.34 xy.

Suppose x is daily ridership and is equal to 100,000, and suppose y is $5. Then by doubling fares, the public transportation agency would increase its daily revenue by .34*100000*$5 = $170,000. Doubling fares again would raise daily revenue by a further .34*.67*100000*10 = $227,800. Thus a price elasticity of -.33 suggests that public transportation authorities could easily reduce their reliance on public subsidies by raising fares. O'Sullivan notes also that this low price elasticity means that reducing fares will not greatly increase ridership.

Estimates for the ridership responses to changes in service speed (so-called time elasticities) have similarly important implications. O'Sullivan estimates the elasticity of demand with respect to line haul time—the percent change in quantity demanded / the percent change in time while in transit—to be .39. Because average speeds on public transportation systems are so low, this number appears deceptively small. If the average speed is 22 miles per hour, as was the case on SEPTA commuter rail in 1986, then a one mile per hour increase in speed is a 4.5 percent increase in speed. Thus O'Sullivan's time elasticity of .39 implies that increasing speeds by ten miles per hour would increase ridership by almost 18 percent.  O'Sullivan reports a much higher elasticity of demand (.71) for time spent waiting for public transportation services; this value is still, however, less than one, and thus if the average time people spend waiting decreases by x percent, ridership will increase by less than x percent. 

Although O'Sullivan presents these elasticity values as absolute facts, there is other evidence on elasticities with quite different implications for government funding of public transportation. In studies of elasticities of demand for SEPTA commuter rail, Voith (1991, 1997) finds that ridership on SEPTA is much more responsive to changes in the price and quality of service in the long-run than O'Sullivan's elasticity values imply (see table 1).   Voith (1991, 1997) calculates separate short-run and long-run elasticities of demand. To see why this is worth doing, consider how the pool of potential commuter rail riders is determined. The better the commuter rail service, the more likely that those commuting from suburbs to the central business district will decide to live near rail stations. Voith (1993) finds that this sort of residential sorting occurs in the Philadelphia area. Thus good SEPTA service increases the pool of potential riders and hence makes it cheaper to provide good rail transportation. Technically, Voith's (1993) insight is that the long run pool of potential riders is endogenous; it is determined by the quality of public transportation.

In the short-run, Voith (1991, p. 360), like O'Sullivan, finds that demand for public transportation responds little to changes in price and service. Indeed, for the short-run elasticity of demand with respect to speed and number of peak trains, Voith (1991) finds that demand responds less (is more inelastic) than O'Sullivan's numbers suggest. Voith's short-run elasticities imply that in the short-run, public transportation authorities can balance their budgets by raising fares and cutting service. His results for the long-run, however, are quite different. Voith finds the responses to service and price changes after one year to be about double what they are in the short-run. In the long-run, demand for public transportation is price elastic; thus if the price increases by x percent, ridership decreases by more than x percent. Long-run ridership is also very responsive to changes in the number of off-peak trains (Voith 1991, 1997). But even in the long-run, Voith (1991, 1997) finds demand to be inelastic with respect to changes in the speed and the number of peak trains. As noted above, given the slow operating speeds on SEPTA, the elasticity of demand with respect to speed is deceptively small.

High elasticities of demand imply that fare increases and service cuts can be suicidal. Raising fares in the short-run may increase total revenue; however, a year later, higher fares will have decreased total revenue. Total revenue falls because there are so many fewer riders. The social benefits of a public transportation system depend on the number of riders; therefore, at the same time as raising fares reduces total revenue, it also reduces the social benefits provided by public transportation—a lose-lose proposition. Cutting service may have similarly perverse effects. Voith (1991) notes that the value of 1.86 for the long-run elasticity of demand with respect to the frequency of off-peak trains suggests that reducing the frequency of off-peak trains will actually reduce ridership per train in the long-run (p. 370). The fewer the riders per train, the higher the costs and subsidy per rider.

By comparing SEPTA and the Port Authority Transit Corporation of Pennsylvania and New Jersey rail line (PATCO), we can see concrete evidence of the higher net social benefits produced by cheaper and more frequent rail service. Both SEPTA and PATCO provide service between Center City Philadelphia and "demographically similar neighborhoods" (Voith, 1994, p. 19). In 1994 the fare on PATCO from Philadelphia to the end of the line was $1.60. The equivalent fare on SEPTA was $3.25 (Voith, p. 19). PATCO ran 33 peak trains into Philadelphia, while SEPTA ran an average of seven trains on each of its commuter rail lines (Voith, p. 19). The result, Voith notes, was that PATCO carried ten times more people per mile of railroad than SEPTA. The Delaware Valley Regional Planning Commission (DVRPC) (2002) reports that 33,234 riders enter Center City Philadelphia on PATCO daily (p. 1). In contrast, DVRPC reports that only 42,788 riders enter Center City Philadelphia from the west on seven SEPTA regional rail lines combined.  Clearly, long-run ridership responds to fare and service differences.

PATCO's higher ridership means it almost certainly provides more net social benefit per rider than SEPTA. To see this, first note that more riders means more intensive use of infrastructure. The result is much lower costs and subsidies per rider on PATCO than on SEPTA. In 1994 the average cost per rider on SEPTA commuter rail was $7.32; on PATCO it was $2.26 (Voith, 1994, p. 21). This meant the average subsidy per rider on SEPTA was $4.37; on PATCO it was only $.89 (Voith, p. 21). As discussed above, even at a high level of subsidy per rider, Pack (1992) finds that SEPTA commuter rail provides large net social benefits to the Philadelphia region. Voith argues that there is no reason to believe that the marginal social benefits of additional commuter rail riders are decreasing rapidly (p. 21). Indeed, the marginal social benefit per rider might be increasing because of the high levels of ridership necessary to realize public transportation externalities such as agglomeration economies and the creation of a livable, walkable city. If one assumes that the social benefits per rider would not change as the number of riders increased, then a reduction in the SEPTA commuter rail subsidy per rider from $4.37 to $.89 would increase the net social benefits per rider by $3.48. 

III. Conclusion
This paper has argued that O'Sullivan's presentation of urban transportation economics is biased against public transportation in many ways. The first section of this paper showed that O'Sullivan neglects many positive externalities of public transportation such as mobility for those without cars and increased regional economic growth. In the second section, I argued that O'Sullivan understates elasticities of demand for public transportation and thus underestimates the ability of public transportation systems to increase their ridership through fare cuts and better service. The comparison of PATCO and SEPTA commuter rail provides anecdotal evidence for this argument. O'Sullivan and many policy makers disagree with this paper's arguments and conclusions. By the time this paper goes to press, a lack of funds may have forced SEPTA to reduce weekday service by 20 percent and raise fares by 25 percent (Downs, 2005). This paper suggests that the savings from these cuts are unlikely to equal the immense costs to the Philadelphia region. In the long-run, these service cuts and price increases could decrease SEPTA's total revenues, worsening SEPTA's financial position and necessitating more service cuts and fare increases in the future. As services are eliminated and fares increase, net social benefits will likely decrease along with ridership.  
Delaware Valley Regional Planning Commission. 2002. 1960-2000 Travel Trends in the Philadelphia Central Business District. Philadelphia, PA.

Downs, Jere. 2005. "SEPTA Gets Money from Road Projects." Philadelphia Inquirer Online. 28 January. Available at

Federal Reserve Bank of Philadelphia (Research Department). 2001. "A Philadelphia Report Card."

Gambaccini, Louis. 2003. Speech to Professor Caskey's urban economics class. April 9, 2003. Swarthmore College, PA. 

Katz, Michael L. and Harvey S. Rosen. 1998. Microeconomics. Boston, MA: Irwin.

Minneapolis Federal Reserve Bank. 2005. "What is a Dollar Worth?" Available at

O'Sullivan, Arthur. 2003. Urban Economics. Chicago: Irwin.

Pack, Janet Rothenburg. 1992. "You Ride, I'll Pay: Social Benefits and Transit Subsidies." Brookings Review. Summer, 10:3, pp. 48-51.

SEPTA (Business Development). 1998. Ridership and Statistics Report. Philadelphia, PA.

SEPTA. 2002. "Fiscal Year 2003 Operating Budget."

The Urban Institute and Cambridge Systematics. 1991. The Economic Impacts of SEPTA on the Regional and State Economy: Final Report. Washington, D.C.: Urban Mass Transportation Administration.

Voith, Richard. 1991. "The Long-Run Elasticity of Demand for Commuter Rail Transportation." Journal of Urban Economics. November, 30:3, pp. 360-372.

Voith, Richard. 1993. "Changing Capitalization of CBD-Oriented Transportation Systems: Evidence from Philadelphia, 1970-1988." Journal of Urban Economics. May, 33:3, pp. 361-376.

Voith, Richard. 1994. "Public Transit: Realizing Its Potential." Federal Reserve Bank of Philadelphia Business Review. September / October, pp. 15-23.

Voith, Richard. 1997. "Fares, Service Levels, and Demographics: What Determines Commuter Rail Ridership in the Long Run?" Journal of Urban Economics. March, 41:2, pp. 176-198.

Vuchic, Vukan R. 1999. Transportation Improvements for Increased Livability in the Philadelphia Region. Philadelphia, PA. (Sponsored by the William Penn Foundation)  


About writing, Josh says:  "For me, writing is like running; it is painful but satisfying. If I am enjoying the writing process, I am usually doing something wrong. Why? Because to write well is to think hard, and thinking hard is often unpleasant.  Whatever the state of my current writing, if not for hours of help from my parents, professors, and friends, it would be much worse. The Elements of Style by William Strunk and E. B. White has also been very useful to me."