Abstract
The USA leads the developed world in motor vehicle fatalities, presenting a critical public health threat. We examined whether an increasing share of mass transit use, relative to vehicle miles traveled on public roads, was associated with reduced motor vehicle fatalities. We used annual city-level data for the USA from 1982–2010 provided by the Fatality Accident Reporting System, the Texas A&M Transportation Institute, the Census Bureau, and the National Oceanic and Atmospheric Administration to estimate a structural equation model of the factors associated with mass transit miles and motor vehicle fatalities. The final analytic data included 2,900 observations from 100 cities over 29 years. After accounting for climate, year, and the economic costs of driving, an increasing share of mass transit miles traveled per capita was associated with reduced motor vehicle fatalities. The costs of congestion to the average commuter and gas prices were positively associated with increasing the share of mass transit miles traveled. The economic costs of driving increased over time, while both the fatality rate and the share of mass transit miles traveled decreased over time. Increasing the share of mass transit miles traveled may be associated with fewer motor vehicle miles traveled. Increasing mass transit uptake may be an effective public health intervention to reduce motor vehicle fatalities in cities.
Keywords: Motor vehicle, Transportation, Mortality, Automobile driving
Introduction
Mass transit has been linked to economic development, reduced traffic congestion and emissions, and better access to affordable transportation options.1 In a poll of the American public’s spending priorities, mass transit ranked 10th, higher than support for funding highways and bridges.2 Aside from the purported benefits of mass transportation, such as job growth and lower emissions, driving is substantially more dangerous than riding mass transit. The World Health Organization predicts that road fatalities will be the fifth leading cause of death worldwide by 2030, and the USA leads most industrialized nations in numbers of fatalities per capita.3 Compared to more than 30,000 deaths from fatal crashes on public roads each year, there are fewer than 150 fatalities on all mass transit combined (a 200-fold difference), and most of those fatalities are due to collisions of nonpassenger trains with cars or persons that were unauthorized to be on the tracks.4
Most research on motor vehicle fatalities has focused on measuring the benefits of vehicle safety, driver behavior, and law enforcement in reducing traffic fatalities.5–14 For example, there have been numerous studies evaluating the benefits of seat belts, motorcycle helmets, and other safety devices and evaluating the effectiveness of improving licensing requirements, traffic enforcement, and laws against drunk driving, cell phone use, and texting.5–19 However, some people may drive more recklessly with increased vehicle safety.20 Distracted driving is a major road hazard, but the effectiveness of laws prohibiting cell phone use and other forms of distraction are mixed.17,18,21–28 By offering a much safer alternative to driving, and thus reducing the exposure to the risk of the roadways, mass transit circumvents the challenges of trying to transform people into safer drivers.
The relationship between mass transit use and road fatalities is complex.29 Calculating an estimate of the potential benefits of mass transit in reducing motor vehicle fatalities is important, despite these challenges, because mass transit systems are usually taxpayer funded and have high capital and operating costs. Thus, their creation or expansion may be controversial, especially in times of economic distress. Establishing a relationship between increased utilization of mass transit and road fatalities may generate greater political urgency to act in support of mass transit. Therefore, this study examined whether an increasing share of mass transit use, relative to vehicle miles traveled on public roads, was associated with reduced motor vehicle fatalities.
Methods
Data Sources
Our study merges various sources of city-level data for the period 1982–2010. Comprehensive data on road fatalities is provided by the Fatality Accident Reporting System (FARS). This database records every vehicular/pedestrian accident occurring on public roadways in the USA where there is at least one fatality resulting from the accident within 30 days. Information is compiled from various records, including police accident reports, vehicle registration and driver licensing files, vital/death certificates, and coroner, hospital, and emergency medical reports.30
We merged annual, city-level data from the Texas Transportation Institute (TTI) urban mobility report with the FARS using year and city identifiers. The TTI collects mass transit utilization and commuting data for 100 urban areas in the USA.31 The cities include the most populous urban areas in the USA, ranging from Eugene, OR, to New York City. Ridership data includes number of annual passenger miles traveled and annual unlinked passenger trips for each city and incorporates data on all forms of mass transit available for city residents. Details about the methodology for the TTI data including details on data sources and calculation of measures can be found in the appendix of the urban mobility report.31 We also merged data on city population estimates from the US Census Bureau and climate data at the city level from the National Oceanic and Atmospheric Administration.32,33 The final analytic data included 2,900 observations from 100 cities over 29 years.
Measures
The primary outcome variable was motor vehicle deaths per 100,000 population using both the FARS and Census data. The outcome was not normally distributed; therefore, we performed a zero-skewness log transformation. The main predictor variable was the share of mass transit utilization at the city level. We calculated the share of mass transit miles traveled per capita by first dividing the annual mass transit miles traveled per capita into the combined annual miles traveled per capita for both mass transit and motor vehicles and then multiplying by 100. Fuel cost was measured in inflation-adjusted dollars per gallon measured daily and then the average value of daily gas values over a given year was used to create an average annual cost. The inflation-adjusted cost of roadway congestion was measured as an index, which was calculated by TTI based on the cost of daily vehicle hours of delay (which uses data on daily vehicle miles of travel and traffic speed) and cost of wasted fuel (which uses data on fuel consumption). Therefore, this index measure parsimoniously accounts for overcapacity of the road system by measuring a number of interrelated factors such as number of commuters, commercial traffic, lane capacity, fuel usage, and travel time.31 Climate was measured by annual precipitation and temperature at the city level. The climate data have been annualized and adjusted by NOAA for systematic, nonclimatic changes that bias temperature and precipitation trends.33 Year was included to account for annual changes in traffic fatalities, annual changes in economic conditions, and annual changes in travel behavior. We squared year to account for nonlinear, annual variations.
Analytic Strategy
Stata 13 was used to perform all analyses (StataCorp, College Station, TX). We employed a structural equation model (SEM) using maximum likelihood estimation and observed information matrix standard errors. We reported standardized coefficients for the parameter estimates and the entire model fit indices available in Stata. We did not have any missing data in our analytic sample. We chose SEM because the relationships between the variables required a path analysis to adequately model the endogenous relationship between the outcome (motor vehicle fatality rate), the exogenous factors (year, climate), the economic costs of driving (congestion, gas), and the share of mass transit miles traveled. This analytic approach also accounts for correlated error between variables (e.g., our model correlates the shared error variance between congestion cost and gas prices).
We conducted a sensitivity analysis given that more than one out of four mass transit commuters in the USA are located in New York City, and 56 % of NYC households do not own a car.34 We ran the structural equation model with New York/Newark excluded from the analysis. The results were substantively the same as the main analysis where New York/Newark are included—with similar standardized coefficients and statistical significance.
Results
Table 1 provides descriptive statistics for the variables used in the structural equation model. Over the 29-year study period of the 100 urban areas included in the sample, there were on average 4.73 annual motor vehicle fatalities per 100,000 persons with a range of 0 to 21.3 per 100,000. The share of mass transit miles traveled, relative to motor vehicle miles traveled, on a daily basis was 2 % with a range of 0 to 29 % indicating that mass transit travel comprises a small proportion of daily travel behavior in most cities. The price of gas was on average US$1.59 with a range of US$0.93 to US$3.84 over the study period. The cost of congestion to the average auto commuter was US$411.87 with a substantial range over time and across cities of less than US$7 to US$2,197.
TABLE 1.
Description of the study sample: 100 metropolitan areas in the USA with any mass transit, 1982–2010
| Mean | Standard deviation | Min | Max | |
|---|---|---|---|---|
| Precipitation, inches | 3.36 | 2.44 | 0.45 | 39.01 |
| Temperature, Fahrenheit | 55.98 | 8.25 | 33.9 | 79.3 |
| Gas prices, US$ | 1.59 | 0.67 | 0.93 | 3.84 |
| Cost of congestion, average commuter, US$ | 411.87 | 329.41 | 6.41 | 2,197.15 |
| Share of annual mass transit miles traveled per capita, % | 2.01 | 2.86 | 0 | 29.16 |
| Road fatalities per 100,000 population, # | 4.73 | 3.56 | 0 | 21.3 |
Sources of data: Fatality Accident Reporting System, Texas Transportation Institute (TTI) urban mobility report, Census Bureau, and National Oceanic and Atmospheric Administration. N = 2,900
Figure 1 provides the standardized coefficients for the structural equation model. The detailed results from the analysis can be found in Table 2. The model fit was good as indicated by the model fit indices (RMSEA = .04; SRMR = .02). The coefficient of determination was 0.72 suggesting that our model explained a majority of the variance for the outcome. See Table 3 for a complete list of the model fit statistics. The primary finding from the structural equation model was that, after accounting for climate, year, and the economic costs of driving, an increasing share of mass transit miles traveled per capita was associated with a lower motor vehicle fatality rate (beta = −.15, standard error (SE) = .02). The costs of congestion to the average commuter (beta = .43, SE = .02) and gas prices (beta = .15, SE = .03) were positively associated with increasing the share of mass transit miles traveled. The economic costs of driving increased over time, while both the fatality rate (beta = −26, SE = .02) and the share of mass transit miles traveled (beta = −.42, SE = .03) decreased over time. Increasing temperature was associated with a higher fatality rate (beta = .27, SE = .02), while higher levels of precipitation were associated with lower fatality rates (beta = −.04, SE = .02). All paths shown in Fig. 1 were statistically significant.
FIG. 1.
Standardized coefficients for the structural equation model. Sources of data: Fatality Accident Reporting System, Texas Transportation Institute (TTI) urban mobility report, Census Bureau, and National Oceanic and Atmospheric Administration. N = 2,900, maximum likelihood estimation method.
TABLE 2.
Structural equation model: standardized coefficients, standard errors, and statistical significance statistics
| Standardized coefficient | Standard error | P value | |
|---|---|---|---|
| Share mass transit <− | |||
| Congestion cost | 0.43 | 0.02 | 0.00 |
| Gas cost | 0.15 | 0.03 | 0.00 |
| Year | −0.42 | 0.03 | 0.00 |
| Precipitation | 0.19 | 0.02 | 0.00 |
| Temperature | −0.08 | 0.02 | 0.00 |
| Constant | 49.87 | 3.18 | 0.00 |
| Fatalities per 100k <− | |||
| Share mass transit | −0.15 | 0.02 | 0.00 |
| Year | −0.26 | 0.02 | 0.00 |
| Temperature | 0.27 | 0.02 | 0.00 |
| Precipitation | −0.04 | 0.02 | 0.04 |
| Constant | 32.66 | 1.94 | 0.00 |
| Congestion cost <− | |||
| Year | 0.57 | 0.01 | 0.00 |
| Constant | −66.83 | 1.37 | 0.00 |
| Gas cost <− | |||
| Year | 0.74 | 0.01 | 0.00 |
| Constant | −85.91 | 0.86 | 0.00 |
| Variance for error | |||
| Mass transit | 0.83 | 0.01 | |
| Fatalities | 0.82 | 0.01 | |
| Congestion cost | 0.67 | 0.01 | |
| Gas cost | 0.45 | 0.01 | |
| Congestion, gas | −0.07 | 0.02 | 0.00 |
Sources of data: Fatality Accident Reporting System, Texas Transportation Institute (TTI) urban mobility report, Census Bureau, and National Oceanic and Atmospheric Administration. N = 2,900, maximum likelihood estimation method, observed information matrix standard errors
TABLE 3.
Model fit statistics for structural equation model
| Fit statistic | Value | Description |
|---|---|---|
| Likelihood ratio | ||
| chi2 | 62.17 | model vs. saturated |
| p > chi2 | 0.00 | |
| chi2_bs | 4578.06 | baseline vs. saturated |
| p > chi2 | 0.00 | |
| Population error | ||
| RMSEA | 0.04 | Root mean squared error of approximation |
| 90 % CI, low | 0.03 | RMSEA lower bound |
| 90 % CI, high | 0.05 | RMSEA upper bound |
| pclose | 0.88 | Probability RMSEA < = 0.05 |
| Information criteria | ||
| AIC | 164395.56 | Akaike’s information criterion |
| BIC | 164491.12 | Bayesian information criterion |
| Baseline comparison | ||
| CFI | 0.99 | Comparative fit index |
| TLI | 0.98 | Tucker-Lewis index |
| Size of residuals | ||
| SRMR | 0.02 | Standardized root mean squared residual |
| CD | 0.72 | Coefficient of determination |
Discussion
Our study examined the magnitude of the relationship between motor vehicle fatalities and mass transit use. Motor vehicle injuries and deaths are considered a worldwide problem and, certainly, one of the winnable public health battles in the USA.3,35 We used structural equation modeling to assess the complex relationship between mass transit use and motor vehicle fatalities. This analytic approach allowed us to account for several complicating factors including the trend toward lower motor vehicle fatalities, seasonal variations in travel behavior, the relationship between the economic costs of driving and mass transit use, and the relationship between congestion and mass transit use. We found that increased use of mass transit was associated with fewer fatalities from motor vehicle crashes after accounting for climate and the economic costs of driving. Therefore, reduced traffic deaths may be counted among the benefits of mass transit use in addition to already reported benefits such as economic development, reduced traffic congestion, and lower emissions.1 Moreover, mass transit could also be considered a public health policy to consider among the many options to reduce motor vehicle fatalities in addition to already well-established policies that work toward improving vehicle safety, driver behavior, and law enforcement.
Our results should be interpreted within the context of several limitations. First, our data are limited to the use of mass transit in 100 major cities, which means this study cannot be interpreted as representative of the entire USA. One strength of the TTI data was measures of the economic costs of driving, which is a critical factor when people make decisions about the mode of travel. However, TTI did not have data on the cost of mass transit, which is also a critical factor in the decision process. We expect that the cost of mass transit would be weighed against the cost of driving when deciding whether and how much one would use mass transit. However, the pricing structure of mass transit is complex because the base cost may vary by city depending on the number of stops or based on the number of rides. Finally, our model would be stronger if there was historical data at the city level for number of fatalities from mass transit, although aggregate data from the Bureau of Transportation Statistics estimate there are approximately 150 per year for the USA and that many of these are caused by commercial trains hitting pedestrians or cars.4
Conclusion
Increasing the share of mass transit miles traveled, relative to motor vehicle miles traveled, may be an effective public health intervention to reduce motor vehicle fatalities in cities. Transportation policies should encourage the use of existing mass transit systems, potentially incentivize the development of new mass transit services, and lower incentives to drive by increasing the economic cost of driving in areas with available—and oftentimes underutilized—mass transit services.
Acknowledgments
Robert Wood Johnson Foundation’s Public Health Law Research Program partially funded this work.
References
- 1.White House. Department of Treasury with the Council of Economic Advisers. An economic analysis of infrastructure investment. October 11, 2010. Available at: http://www.whitehouse.gov/sites/default/files/infrastructure_investment_report.pdf. Accessed February 18, 2014.
- 2.Smith TW. Trends in National Spending Priorities, 1973–2008. General Social Survey. Chicago, IL: University of Chicago, National Opinion Research Center; 2009. Available at: http://news.uchicago.edu/static/newsengine/pdf/090210.SPEND08.pdf. Accessed February 18, 2014.
- 3.World Health Organization . Global status report on road safety: time for action. Geneva, Switzerland: World Health Organization; 2009. [Google Scholar]
- 4.Bureau of Transportation Statistics. National transportation statistics. US Department of Transportation. Available at: http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/index.html. Accessed February 18, 2014.
- 5.Beck LF, Shults RA, Mack KA, Ryan GW. Associations between sociodemographics and safety belt use in states with and without primary enforcement laws. Am J Public Health. 2007;97(9):1619–24. doi: 10.2105/AJPH.2006.092890. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chen LH, Baker SP, Li G. Graduated driver licensing programs and fatal crashes of 16 year old drivers: a national evaluation. Pediatrics. 2006;118(1):56–62. doi: 10.1542/peds.2005-2281. [DOI] [PubMed] [Google Scholar]
- 7.McCartt AT, Teoh ER, Fields M, Braitman KA, Hellinga LA. Graduated licensing laws and fatal crashes of teenage drivers: a national study. Traffic Inj Prev. 2010;11(3):240–8. doi: 10.1080/15389580903578854. [DOI] [PubMed] [Google Scholar]
- 8.Ivers R, Senserrick T, Boufous S, Stevenson M, Chen HY, Woodward M, Norton R. Novice drivers’ risky driving behavior, risk perception, and crash risk: findings from the DRIVE study. Am J Public Health. 2009;99(9):1638–44. doi: 10.2105/AJPH.2008.150367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Friedman LS, Hedeker D, Richter ED. Long-term effects of repealing the national maximum speed limit in the United States. Am J Public Health. 2009;99(9):1626–31. doi: 10.2105/AJPH.2008.153726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Houston DJ, Richardson LE. Motorcycle safety and the repeal of universal helmet laws. Am J Public Health. 2007;97(11):2063–9. doi: 10.2105/AJPH.2006.094615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Tester JM, Rutherford GW, Wald Z, Rutherford MW. A matched case-control study evaluating the effectiveness of speed humps in reducing child pedestrian injuries. Am J Public Health. 2004;94(4):646–50. doi: 10.2105/AJPH.94.4.646. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Retting RA, Ferguson SA, McCartt AT. A review of evidence-based traffic engineering measures designed to reduce pedestrian-motor vehicle crashes. Am J Public Health. 2003;93:1456–63. doi: 10.2105/AJPH.93.9.1456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hingson R, Heeren T, Winter M. Effects of recent 0.08% legal blood alcohol limits on fatal crash involvement. Inj Prev. 2000;6(2):109–14. doi: 10.1136/ip.6.2.109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wagenaar AC, O’Malley PM, LaFond C. Lowered legal blood alcohol limits for young drivers: effects on drinking, driving, and driving-after-drinking behaviors in 30 states. Am J Public Health. 2001;91(5):801–4. doi: 10.2105/AJPH.91.5.801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Homer J, French M. Motorcycle helmet laws in the United States from 1990 to 2005: politics and public health. Am J Public Health. 2009;99(3):415–23. doi: 10.2105/AJPH.2008.134106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Mertz KJ, Weiss HB. Changes in motorcycle-related head injury deaths, hospitalizations, and hospital charges following repeal of Pennsylvania’s mandatory motorcycle helmet law. Am J Public Health. 2008;98(8):1464–7. doi: 10.2105/AJPH.2007.123299. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.McCartt AT, Geary LL. Longer term effects of New York State’s law on drivers’ handheld cell phone use. Inj Prev. 2004;10:11–5. doi: 10.1136/ip.2003.003731. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.McCartt AT, Hellinga LA. Longer-term effects of Washington, DC, law on drivers’ hand-held cell phone use. Traffic Inj Prev. 2007;8(2):199–204. doi: 10.1080/15389580701238859. [DOI] [PubMed] [Google Scholar]
- 19.Houston DJ, Richardson LE. Safety belt use and the switch to primary enforcement, 1991–2003. Am J Public Health. 2006;96(11):1949–54. doi: 10.2105/AJPH.2005.074385. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Peltzman S. The effects of automobile safety regulation. J Polit Econ. 1975;83(4):677–726. doi: 10.1086/260352. [DOI] [Google Scholar]
- 21.McCartt AT, Hellinga LA, Braitman KA. Cell phones and driving: review of research. Traffic Inj Prev. 2006;7(2):89–106. doi: 10.1080/15389580600651103. [DOI] [PubMed] [Google Scholar]
- 22.McEvoy SP, Stevenson MR, McCartt AT, Woodward M, Haworth C, Palamara P, Cercarelli R. Role of mobile phones in motor vehicle crashes resulting in hospital attendance: a case-crossover study. BMJ. 2005;331(7514):428. doi: 10.1136/bmj.38537.397512.55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Caird JK, Willness CR, Steel P, Scialfa C. A meta-analysis of the effects of cell phones on driver performance. Accid Anal Prev. 2008;40:1282–93. doi: 10.1016/j.aap.2008.01.009. [DOI] [PubMed] [Google Scholar]
- 24.Klauer SG, Dingus TA, Neale VL, Sudweeks JD, Ramsey DJ. The impact of driver inattention on near-crash/crash risk: an analysis using the 100-Car Naturalistic Driving Study Data. Washington, DC: National Highway Traffic Safety Administration; 2006. [Google Scholar]
- 25.Neyens DM, Boyle LN. The influence of driver distraction on the severity of injuries sustained by teenage drivers and their passengers. Accid Anal Prev. 2008;40(1):254–9. doi: 10.1016/j.aap.2007.06.005. [DOI] [PubMed] [Google Scholar]
- 26.Stutts J, Feaganes J, Reinfurt D. Drivers’ exposure to distractions in their natural driving environment. Accid Anal Prev. 2005;37(6):1093–101. doi: 10.1016/j.aap.2005.06.007. [DOI] [PubMed] [Google Scholar]
- 27.Strayer DL, Drews FA, Johnston WA. Cell phone-induced failures of visual attention during simulated driving. J Exp Psychol Appl. 2003;9(1):23–32. doi: 10.1037/1076-898X.9.1.23. [DOI] [PubMed] [Google Scholar]
- 28.Wilson FA, Stimpson JP. Trends in fatalities from distracted driving in the US, 1999–2008. Am J Public Health. 2010;100:2213–9. doi: 10.2105/AJPH.2009.187179. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ewing R, Schieber RA, Zegeer CV. Urban sprawl as a risk factor in motor vehicle occupant and pedestrian fatalities. Am J Public Health. 2003;93:1541–5. doi: 10.2105/AJPH.93.9.1541. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.National Highway Traffic Safety Administration . FARS: analytic reference guide 1975–2010. Washington, DC: National Highway Traffic Safety Administration; 2011. [Google Scholar]
- 31.Schrank D, Lomax T, Eisele B. 2012 Urban mobility report. Texas Transportation Institute. December 2012. http:// http://mobility.tamu.edu/ums/. Accessed March 21, 2014.
- 32.US Census Bureau. Current estimates data. Available at: http://www.census.gov/popest/data/index.html. Accessed March 21, 2014.
- 33.National Oceanic and Atmospheric Administration National Climatic Data Center. U.S. Historical Climatology Network. http://www.ncdc.noaa.gov/oa/climate/research/ushcn/. Accessed March 21, 2014.
- 34.US Census Bureau. 2000 Census of population and housing. Summary file 3: technical documentation. Available at: http://www.census.gov/prod/cen2000/doc/sf3.pdf. Accessed February 18, 2014.
- 35.Centers for Disease Control and Prevention. Motor vehicle injuries. Available at: http://www.cdc.gov/WinnableBattles/MotorVehicleInjury/. Accessed February 18, 2014.