U.S. transit ridership falls behind international peers. The cited causes include poor land use, ill-conceived network effects, and low frequency, yet transit ridership also varys tremendously within the United States. Within U.S. transit there are both bright spots and examples of under-preforming systems. The causes of these differences should be examined carefully. Within the U.S., systems vary in the spacing of stations, by intended ridership, intersection priority, and rolling stock design. This analysis is not meant to demean systems but rather to provide viewers with an alternative method of comparing systems.
We can examine how well transit capital is used by comparing ridership per route mile across systems. A system with more riders and shorter overall system will prove more efficient with it’s use of system capital. While length of track is not the primary determinant of total ridership or capital costs, it does serve as a proxy. Smaller systems that move the same as a larger system are an impressive example of how well-designed and -placed transit assets can deliver ridership and revenue. To move trips from cars to transit, we should prioritize transit planning that maximizes ridership.
The following figures use American Public Transit Association data to demonstrate heavy rail transit efficiency through ridership per route mile. Figures show total monthly ridership divided by route mile length for rail systems and monthly ridership divided by fleet size for buses. Ridership, route length and fleet data was collected from datasets compiled by the American Public Transit Association.
New York City’s Subway dominates per route mile ridership just as it excels at ridership overall. The PATH system, another New York Metro Area system, is similarly far more efficient on route usage than any other heavy rail system in the country. However, the results are unexpected among less used systems. Heavy-ridership systems, the Washington DC Metrorail and Chicago L, don’t do not rank as high on ridership per route mile.
Smaller older systems, the MBTA Heavy Rail and SEPTA Heavy Rail, as well as the LA Metro Heavy Rail lines move a surprisingly high number of passengers over relatively shorter systems. The Great Society ‘Metros’ with suburban commuter focuses and sometimes poor urban core access fail to compete on ridership per route mile with the older and more urban focused ‘Subways’. BART, MARTA, WMATA, and Miami Metro all preform similarly when it comes to system use efficiency which means the high total ridership experienced by BART and WMATA might be more associated with system length than system performance.
Ridership data is as of September 20, 2023 and reflect June’s monthly ridership levels.
Toggle system types on and off using the legend.
Commuter rail systems generally move far less people and are used far less per route mile than comparable heavy rail systems. Often lower ridership on commuter rail systems is more a product of less frequent service and lower density catchement areas rather than reduced capacity or ability. Many commuter rail systems, especially grade separated and lines with overhead catenary, have capacities that meet or exceed the most productive heavy rail systems. The TFL’s Elizabeth Line across greater London and Southeast England is one great example, its newly inagurated service is one of the most popular in all of the United Kingdom.
The graph below demonstrates the per route mile ridership per month of most commuter rail systems in the United States. The APTA classifies the eBART as commuter rail and does not provide data on the CT Rail Hartford Line or the New Arrow Service in California. All three services are not shown
Services are split into, commuter rail and regional rail. The main distinction being levels of service, generally regional rail services provide weekend, mid-day, and reverse peak service. However, many ‘regional rail systems’ in the United States provide very poor levels of weekend, mid-day, and reverse peak service.
The graph clearly shows that regional rail services, ie systems running more trains, are more productive per route mile than commuter rail services. Although all of the systems shown below could likely dramatically increase ridership and capital productivity by running more services, the commuter rail systems could especially benefit by transistioning to all day & bi-directional service. All services, especially regional rail systems, could also put their current services to greater use by increasing transit oriented development (TOD), providing all day feeder bus service, and integrating fares with competing regional and local transit options.
Denver’s somewhat new regional rail services provide metro-like service with great ridership productivity results only exceeded by the dominate Long Island Railroad (LIRR). The RTD (Denver’s transit agency) inaugurated regional rail services with overhead catenary, 15-30 min headways, and feeder bus services from the get-go. Yet TOD around RTD regional rail stations is currently lacking and likely reduces the potential of the RTD regional rail services, focusing on converting the large RTD parking lots to TOD could lead to further ridership successes. Nonetheless, existing and future rail planning could learn from the tremendous success of the RTD’s regional rail program.
Ridership data is as of September 20, 2023 and reflect June’s monthly ridership levels.
Toggle system types on and off using the legend.
Light rail in the United States is a more puzzling phenomena. Not much can be said other than Light Rail in the United States is an enigma, services among this list could easily be described as interburbans, light metro, trams, and pre-metro.
A clear ryhme or reasons to ridership productivity also cannot be easily deduced. Overall and per-mile ridership on light rail spans factors including service levels, signal priority, nearby density, local car culture, station design, and rolling stock. These many often changing factors can’t be easily grouped and many light rail operations/policy-makers tend upgrade or roll back these malleable systems on a moments notice.
Yet, the close resemblance in ridership productivity among Light and Heavy rail systems is apparent. Since light rails can take many forms, many can out pace some heavy rail systems. For example, the rail use productivity of Great Society Metros is less than the most efficient light rail systems.
Ridership data is as of September 20, 2023 and reflect June’s monthly ridership levels.
Click on a plot line or search from the drop down menu to highlight a light rail system of interest. You may select multiple systems at once. Reset figure by double clicking or clearing search bar.
Toggle system types on and off using the legend.
Bus systems obviously don’t have tracks and supposedly can switch routes with relative ease (despite this not often happening without lengthy consultation), yet the following figure demonstrates that systems do not equally apply route-making to gaining ridership. Relatively popular bus systems such as WMATA Metrobus are on par with the Maryland MTA on a ridership per vehicle basis and the South Nevada RTC, relatively unknown for their transit, preforms third only to the NYC MTA & San Francisco MUNI on a ridership per vehicle basis. Clearly the Southern Nevada RTC has optimized routes & vehicle use to drive ridership while WMATA has not yet achieved similar results. WMATA’s ongoing bus system redesign may hopefully put WMATA’s fleet to greater use.
Click on a plot line or search from the drop down menu to highlight a bus system of interest. You may select multiple systems at once. Reset figure by double clicking or clearing search bar.
Ridership data is as of September 20, 2023 and reflect June’s monthly ridership levels.
Fares do not cover the expenses associated with operating U.S. transit services. Only a few international systems are able to generate an operating profit. However, within the United States there are wide variances among U.S. transit system’s ability to generate revenue from fares.
Since transit systems with more riders per route mile or riders per vehicle should have increased fare potential per fixed capital cost, it would reason that systems with high capital efficiency would be able to generate a higher proportion of expenses from fares.
The following graphs illustrate regressions of farebox recovery ratios and passengers per route mile. Ridership data is as of September 20, 2023 and reflect June’s monthly ridership levels. Farebox recovery data was collected from the National Transit Database’s 2021 revenues and expenses datasets. The farebox recovery percentage was calculated by dividing farebox revenues by total operational expenses. Some agencies provide more recent farebox recovery numbers. When available more recent values replaced National Transit Database data, these sources are provided at the end of the page.
Other than the logistic & linear regression for heavy rail, the regressions do not point to a convincing relationship. The weak regressions do suggest that higher rail efficiency is better for farebox recovery.
##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.079366 -0.033515 -0.006737 0.031565 0.096452
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.06413 0.04776 -1.343 0.20227
## log(CountofRiders) 0.05039 0.01123 4.489 0.00061 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05424 on 13 degrees of freedom
## Multiple R-squared: 0.6078, Adjusted R-squared: 0.5777
## F-statistic: 20.15 on 1 and 13 DF, p-value: 0.0006095##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.074653 -0.024497 0.007624 0.017213 0.097699
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.01685 0.04655 -0.362 0.72368
## log(CountofRiders) 0.03663 0.01151 3.181 0.00791 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.04725 on 12 degrees of freedom
## Multiple R-squared: 0.4575, Adjusted R-squared: 0.4123
## F-statistic: 10.12 on 1 and 12 DF, p-value: 0.007905Only includes light rail systems with over 5 miles of system length to exclude streetcars. The New York Subway is also excluded.
##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.100210 -0.037297 -0.008107 0.025685 0.143120
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.012958 0.026647 -0.486 0.63
## log(CountofRiders) 0.033539 0.007573 4.429 8.48e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05028 on 36 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.3527, Adjusted R-squared: 0.3347
## F-statistic: 19.61 on 1 and 36 DF, p-value: 8.481e-05##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.106422 -0.065612 -0.006827 0.028663 0.214650
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.08436 0.01838 4.589 9.95e-05 ***
## log(CountofRiders) 0.01369 0.01273 1.076 0.292
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.08337 on 26 degrees of freedom
## Multiple R-squared: 0.04261, Adjusted R-squared: 0.005791
## F-statistic: 1.157 on 1 and 26 DF, p-value: 0.2919Excluding fare free systems.
##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.099155 -0.037206 -0.009805 0.028865 0.144734
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.01932 0.03674 -0.526 0.60386
## log(CountofRiders) 0.03490 0.01153 3.028 0.00581 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05603 on 24 degrees of freedom
## Multiple R-squared: 0.2764, Adjusted R-squared: 0.2462
## F-statistic: 9.166 on 1 and 24 DF, p-value: 0.005812##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.12117 -0.04975 -0.01864 0.03132 0.24021
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.058451 0.014716 3.972 0.000167 ***
## log(CountofRiders) 0.015554 0.005022 3.097 0.002787 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.0756 on 72 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.1175, Adjusted R-squared: 0.1053
## F-statistic: 9.591 on 1 and 72 DF, p-value: 0.002787##
## Call:
## lm(formula = FareboxRatio ~ log(CountofRiders), data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.10835 -0.05751 -0.03086 0.02179 0.82730
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.07138 0.08592 0.831 0.411
## log(CountofRiders) 0.02170 0.05449 0.398 0.693
##
## Residual standard error: 0.1457 on 38 degrees of freedom
## Multiple R-squared: 0.004157, Adjusted R-squared: -0.02205
## F-statistic: 0.1586 on 1 and 38 DF, p-value: 0.6927##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.08338 -0.04152 0.01106 0.03188 0.10256
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9.478e-02 1.665e-02 5.691 7.4e-05 ***
## CountofRiders 3.724e-04 7.851e-05 4.743 0.000384 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05242 on 13 degrees of freedom
## Multiple R-squared: 0.6338, Adjusted R-squared: 0.6056
## F-statistic: 22.5 on 1 and 13 DF, p-value: 0.0003843##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.08412 -0.04497 0.01159 0.03456 0.10204
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.0957748 0.0201640 4.750 0.000472 ***
## CountofRiders 0.0003584 0.0001670 2.146 0.053039 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05454 on 12 degrees of freedom
## Multiple R-squared: 0.2773, Adjusted R-squared: 0.2171
## F-statistic: 4.604 on 1 and 12 DF, p-value: 0.05304Only includes light rail systems with over 5 miles of system length. The New York Subway is also excluded.
##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.081274 -0.034165 -0.009514 0.034476 0.156337
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.0748053 0.0112052 6.676 8.77e-08 ***
## CountofRiders 0.0004868 0.0001386 3.512 0.00122 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05394 on 36 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.2552, Adjusted R-squared: 0.2345
## F-statistic: 12.33 on 1 and 36 DF, p-value: 0.001218##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.11761 -0.06181 -0.01345 0.03021 0.21548
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.082714 0.020405 4.054 0.000406 ***
## CountofRiders 0.002804 0.003046 0.920 0.365854
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.08385 on 26 degrees of freedom
## Multiple R-squared: 0.03155, Adjusted R-squared: -0.005698
## F-statistic: 0.847 on 1 and 26 DF, p-value: 0.3659Excluding fare free systems.
##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.087513 -0.038020 -0.006903 0.008547 0.159630
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.0527314 0.0163582 3.224 0.00363 **
## CountofRiders 0.0010633 0.0003729 2.851 0.00882 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05693 on 24 degrees of freedom
## Multiple R-squared: 0.253, Adjusted R-squared: 0.2219
## F-statistic: 8.128 on 1 and 24 DF, p-value: 0.008816##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.10218 -0.05420 -0.01568 0.03623 0.22275
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.832e-02 8.909e-03 8.790 5.14e-13 ***
## CountofRiders 4.230e-04 8.967e-05 4.717 1.14e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.07034 on 72 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.2361, Adjusted R-squared: 0.2255
## F-statistic: 22.25 on 1 and 72 DF, p-value: 1.141e-05##
## Call:
## lm(formula = FareboxRatio ~ CountofRiders, data = regress_hold)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.10617 -0.05586 -0.02892 0.02043 0.82873
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.085793 0.057834 1.483 0.146
## CountofRiders 0.003718 0.010630 0.350 0.728
##
## Residual standard error: 0.1457 on 38 degrees of freedom
## Multiple R-squared: 0.003209, Adjusted R-squared: -0.02302
## F-statistic: 0.1223 on 1 and 38 DF, p-value: 0.7285