Skip to main content
SpringerLink
Log in

The Effect of Gasoline Taxes and Public Transit Investments on Driving Patterns

  • Published:
Environmental and Resource Economics Aims and scope Submit manuscript

Abstract

This paper analyzes how driving patterns are affected by gasoline taxes and the availability of a substitute for driving—public transportation. We develop a measure of transportation substitutability based on the difference between individuals’ predicted commute times by private and public transit, conditional upon their demographic characteristics and geographic location. Improved substitutability decreases annual vehicle miles traveled (VMT) by inducing modal shifts to public transit, though gasoline taxes are found to have a much larger impact on VMT. Our results imply that a policy that raises gasoline taxes and recycles the revenues into public transit improvements can have even larger impacts on driving patterns than either policy alone.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3

Similar content being viewed by others

Notes

  1. For example, Fullerton and Gan (2005) find that a 1 % increase in the price of an SUV causes shifts in purchasing away from bundles of vehicles with SUVs in favor of bundles with only cars.

  2. For other extensive surveys of the literature, see Dahl (1986), Oum (1989), Dahl and Sterner (1991), Goodwin (1992), Espey (1998), Basso and Oum (2007).

  3. This difference in estimated elasticities can be large; for example, Dahl (1978) finds a-long run elasticity for US households of \(-\)0.78 versus a short-run elasticity of \(-\)0.44. However, both Haughton and Sarkar (1996) and Sipes and Mendelsohn (2001) find smaller discrepancies of 0.1 to 0.2 between long-run and short-run estimates.

  4. Subway and buses run more frequently at peak times in most cities within the US, which may not be true in other parts of the world.

  5. For a detailed description and summary of the dataset, see Hu and Reuscher (2004).

  6. Hotel and airport shuttles, limousines, taxis, private boats, and private airplanes were not included as fitting either of these two definitions; however, this excluded a negligible portion of the survey responses. Out of the 91,742 individuals who reported a method of commute, all of the above represented a combined 118 observations. Walking and biking, which constituted a total of 1.4 % of those surveyed, were also excluded. Likewise, no distinction was made for those who carpool, representing 3.2 % of those who commuted by private vehicle.

  7. While much of the National Household Travel Survey data is publicly available, geographic locations of individuals are only available via a confidential agreement with the U.S. Department of Transportation.

  8. We are unable to use MSA dummies in the regression, as they would be correlated with the MSA price variable. We tested the implications of this by using free monthly MSA gasoline price data available from the DOE for a select sample of six MSAs and including MSA dummies in the regression. The results with month-MSA level prices are statistically insignificant, and they are also not statistically different from the results on the subsample of the same MSAs with yearly level MSA prices making us more confident in using our annual MSA gasoline price data. Unfortunately, time-varying data are expensive and we do not have funding to purchase the time-varying data for the full sample of MSAs.

  9. As our data spans through the terrorist attacks in September 2001, factors such as transit disruptions in New York and Washington, DC, higher gasoline prices, etc., may have affected the results. Thus, we test our results excluding the observations after 9/11 and our results were not statistically significantly different from not excluding them; suggesting the 9/11 disruptions do not drive our results.

  10. Table 1 includes an X for each city that has a subway system. However, we do not control explicitly for the existence of a subway system in our estimation. See Sect. 5 (Footnote 17) for a discussion of why we do not need this control and of how our model does control for spatial variation in the quality of public transit.

  11. At a minimum our results should be robust for the population and areas included in our data.

  12. Certain transportation models calculate driving times by public and private times, though there are several drawbacks to these models. For example, a Stata program that uses Google maps does not calculate optimal public transit times (which may involve travel by combinations of different transit modes). Other existing models only work for one MSA and thus would not allow for a calculation of elasticities given our data constraint on gasoline prices.

  13. Although other factors (such as cleanliness, privacy, and security) may likely factor into the decision on whether to take public transit, it is difficult to quantify these aspects with data. Therefore, our analysis ignores other reasons why individuals may choose not to use public transit.

  14. We also conduct matching techniques instead of predictions in the first stage. These alternate techniques did not provide us with meaningful estimates; results are not shown here.

  15. For more about this, see Sect. 7.

  16. A \(t\)-test reveals these two means are not statistically different from each other.

    Fig. 3
    figure 3

    Distributions of commute times. Notes The fitted private and public commute time graphs show the distribution of predicted commute times for NHTS individuals based on Eq. (1). The observed private commute time shows the distribution of observed commute times for NHTS individuals who commute to work by private modes

    Full size image

  17. Our fitted results include no observations below zero even though we utilize OLS, as none of the observed private or public times are negative or zero.

  18. Utilizing the difference in predicted public and private transit times allows us to capture many factors including speed, frequency and reliability that would affect the decision to take public transit. Existence of a subway system also affects public transit times; however, we do not include this as a regressor, as the first stage prediction regressions are conducted at the PUMA level, where it is a reasonable approximation to consider access to the subway as uniform (including distances to subways would provide within-PUMA variation, however we do not have access to this sort of data). Therefore, our estimates do control implicitly not only for the existence of a subway or rail system in cities that have them, but for some degree of within-MSA heterogeneity in access to those systems as well.

  19. If public transit accessibility affected VMT directly (not just through mode choice) it would mean that individuals with better access drive less without shifting modes, which does not seem likely. While it is possible that there could be unobserved factors that lead people to avoid taking public transit while driving more, we assume that once we have controlled for other factors affecting VMT that this is not the case.

  20. Feng et al. (2005) also find that households with more than one vehicle are more elastic than households with one.

  21. The Durbin-Wu-Hausman test for endogeneity produces a \(p\)-value of 0.000, significantly proving the endogeneity of mode choice and invalidating OLS.

  22. Census Divisions include Pacific, Mountain, West North Central, East North Central, West South Central, East South Central, East North Central, Middle Atlantic, South Atlantic, and New England; we exclude the South Atlantic division.

  23. Fall is the excluded dummy variable. The seasons are generated in the following way—winter: December to February; spring: March to May; summer: June to August.

  24. We do not include census division dummies in the VMT equation, as this masks much of the gasoline price variation in our data.

  25. See Freedman and Peters (1984) and Efron and Tibshirani (1993) for a discussion on the validity of bootstrapping standard errors in multiple stage estimation techniques.

  26. The MPG coefficient remained statistically insignificant even in alternative specifications of the model.

  27. These estimates demonstrate the impact of marginal changes in both gasoline price and public transit accessibility; large changes in public transit (such as from the implementation of a new subway or light rail line) may have large external impacts from new sorting behavior. Our model does not attempt to evaluate large changes in public transit infrastructure; instead we analyze the impact of a marginal change, such as an increase in the number of buses on a currently existing route. We do not expect these marginal changes to have large cascading impacts on sorting behavior. Furthermore, though our model implicitly finds a short-run estimate (in that we hold fixed vehicle stock and housing location), we expect that, in the long-run, major adjustments would be small under marginal changes. Though minor adjustments will likely occur with marginal changes, estimating these sorting adjustments would require much more extensive data which we do not have access to, and hence, this estimation is outside the scope of this project.

  28. Given that the dependent variable in Eq. (2) is logged VMT, when calculating the elasticity of VMT with respect to price, VMT falls out of the equation.

  29. These results may imply that public transit improvements today would be more effective than in 2001, given the large increase in average prices over the past decade. This is supported by analytical evidence: Klier and Linn (2010) demonstrate that the demand response is greater with higher gasoline prices. Analyzing the impact of public transit accessibility using more current price data to resolve this issue would be a worthy goal for future work; however, it is outside the scope of this paper. Therefore, our use of earlier data likely underestimates the impact of public transit on driving in later years.

  30. Even if individuals choose to drive slightly more with a more fuel efficient vehicle (i.e., the rebound effect), this effect will mitigate the decline in consumption but will not completely eliminate it. Gillingham et al. (2013) argue that while the rebound effect exists, it is too small (generally less than 10 %) to overturn energy savings entirely. An older study, Small and Van Dender (2005), also find small rebound effects, both in the long and short-run.

  31. We are not able to use the full sample of the NHTS for this estimation given crucial missing data on gasoline prices and MPG.

  32. See Spiller (2012) for a discussion of how the explicit modeling of vehicle substitution affects the elasticity estimate.

  33. In fact, Lin and Zeng (2012) find that the elasticity of VMT is three times higher than the elasticity of gasoline price.

  34. Though we do not present the results here, they are available from the authors upon request.

References

  • Baltagi BH, Griffin JM (1983) Gasoline Demand in the OECD: an Application of Pooling and Testing Procedures. Eur Econ Rev 22(2):117–137

    Article  Google Scholar 

  • Basso L, Oum TH (2007) Automobile fuel demand: a critical assessment of empirical methodologies. Transp Rev 27(4):449–484

    Article  Google Scholar 

  • Bento A, Goulder LH, Henry E, Jacobsen M, von Haefen RH (2005) Distributional and efficiency impacts of gasoline taxes: an econometrically based multi-market study. AEA Papers Proc 95(2):282–287

    Article  Google Scholar 

  • Bhat C (1997) Work travel mode choice and number of non-work commute stops. Transp Res Part B Methodol 31(1):41–54

    Article  Google Scholar 

  • Bhat CR, Sardesai R (2006) The impact of stop-making and travel time reliability on commute mode choice. Transp Res Part B 40(9):709–730

    Article  Google Scholar 

  • Blum UCH, Foos G, Gaudry MJI (1988) Aggregate time series gasoline demand models: review of the literature and new evidence for west Germany. Transp Res A 22A(2):75–88

    Article  Google Scholar 

  • Dahl CA (1978) Consumer adjustment to a gasoline tax. Rev Econ Stat 61(3):427–432

    Article  Google Scholar 

  • Dahl C (1986) Gasoline demand survey. Energy J 7(1):67–82

    Article  Google Scholar 

  • Dahl C, Sterner T (1991) Analysing gasoline demand elasticities: a survey. Energy Econ 13:203–210

    Article  Google Scholar 

  • Efron B, Tibshirani RJ (1993) An introduction to the bootstrap. Chapman and Hall, New York

    Book  Google Scholar 

  • Espey M (1998) Gasoline demand revisited: an international meta-analysis of elasticities. Energy Econ 20:273–295

    Article  Google Scholar 

  • US Energy Information Administration (2012) Estimated US gasoline consumption low compared to five-year average. http://www.eia.gov/todayinenergy/detail.cfm?id=7510

  • Feng Y, Fullerton D, Gan L (2005) Vehicle choices, miles driven, and pollution policies. Working Paper 11553. NBER Working Paper Series, Cambridge, MA

  • Freedman DA, Peters SC (1984) Bootstrapping an econometric model: some empirical results. J Bus Econ Stat 2(2):150–158

    Google Scholar 

  • Fullerton D, Gan L (2005) Cost-effective policies to reduce vehicle emissions. AEA Papers Proc 95(2):300–304

    Article  Google Scholar 

  • Gillingham K, Kotchen M, Rapson DS, Wagner G (2013) Energy policy: the rebound effect is overplayed. Nature 493:475–476

    Article  Google Scholar 

  • Goldberg PK (1998) The effects of the corporate average fuel efficiency standards in the US. J Ind Econ 46(1):1–33

    Article  Google Scholar 

  • Goodwin PB (1992) A review of new demand elasticities with special reference to short- and long-run effects of price changes. J Transp Econ Policy 26:155–164

    Google Scholar 

  • Graham D, Glaister S (2002) The demand for automobile fuel: a survey of elasticities. J Transp Econ Policy 36:1–26

    Google Scholar 

  • Grazi F, van den Bergh CJM, van Ommeren J (2008) An empirical analysis of Urban form, transport, and global warming. Energy J 29(4):97–122

    Article  Google Scholar 

  • Haughton J, Sarkar S (1996) Gasoline tax as a corrective tax: estimates for the United States, 1970–1991. Energy J 17(2):103–126

    Article  Google Scholar 

  • Heckman J (1979) Sample selection bias as a specification error. Econometrica 47:153–161

    Article  Google Scholar 

  • Hu PS, Reuscher TR (2004) Summary of travel trends: 2001 National Household Travel Survey. U.S, Department of Transportation, Federal Highway Administration, Washington, DC

  • Johansson O, Schipper L (1997) Measuring the long-run fuel demand for cars: separate estimations of vehicle stock and mean annual driving distance. J Transp Econ Policy 31:278–292

    Google Scholar 

  • Klier T, Linn J (2010) The price of gasoline and new vehicle fuel economy: evidence from monthly sales data. Am Econ J Econ Policy 2(3):134–153

    Article  Google Scholar 

  • Lee LF (1978) Unionism and wage rates: a simultaneous equations model with qualitative and limited dependent variables. Int Econ Rev 19(2):415–433

    Article  Google Scholar 

  • Li S, Timmins C, von Haefen RH (2009) How do gasoline prices affect fleet fuel economy? Am Econ J Econ Policy 1(2):113–137

    Article  Google Scholar 

  • Lin C and Zeng J (2012) The elasticity of demand for gasoline and the optimal gas tax for China. Working Paper, under review at Energy Policy, accessed 6/7/12

  • McFadden DM (1974) The measurement of urban travel demand. J Public Econ 3:303–328

    Article  Google Scholar 

  • Oum TH (1989) Alternative demand models and their elasticity etimates. J Transp Econ Policy 23:163–187

    Google Scholar 

  • Parry I, Small K (2009) Should urban transit subsidies be reduced? Am Econ Rev 99(3):700–724

    Article  Google Scholar 

  • Reschovsky C (2004) Journey to Work: 2000, Census 2000 Brief. U.S, Department of Commerce, Census Bureau, Washington, D.C.

  • Sipes K, Mendelsohn R (2001) The effectiveness of gasoline taxation to manage air pollution. Ecol Econ 36:299–309

    Article  Google Scholar 

  • Small K and Van Dender K (2005) The effect of improved fuel economy on vehicle miles traveled: estimating the rebound effect using U.S. state data, 1966–2001. Policy and Economics, University of California Energy Institute, UC Berkeley

  • Spiller E (2012) Household vehicle bundle choice and gasoline demand: a discrete-continuous approach, Working Paper. Available at SSRN: http://ssrn.com/abstract=2129396

  • Spiller E, Stephens HM (2012) The distributional concerns of gasoline taxes: why where we live matters. Resources for the Future Discussion Paper No. 12–30

  • Sterner T (2007) Fuel taxes: an important instrument for climate policy. Energy Policy 35(6):3194–3202

    Article  Google Scholar 

  • Sterner T, Dahl C, Franzen M (1992) Gasoline tax policy: carbon emissions and the global environment. J Transp Econ Policy 26:109–119

    Google Scholar 

  • Su Q, DeSalvo JS (2008) The effect of transportation subsidies on urban sprawl. J Reg Sci 48(3):567–594

    Article  Google Scholar 

  • Train K (1980) A structured logit model of auto ownership and mode choice. Rev Econ Stud 47(2):357– 370

    Google Scholar 

  • West Sarah E (2004) Distributional effects of alternative vehicle pollution control policies. J Public Econ 88:735–757

    Article  Google Scholar 

  • Wheaton W (1982) The long-run structure of transportation and gasoline demand. Bell J Econ 13:439– 454

    Google Scholar 

Data Sources

  • U.S. Department of Commerce, Census Bureau. 2000 Census data, 5% sample. Selected variables extracted through University of Minnesota, Minnesota Population Center, Integrated Public Use Microdata Series (IPUMS). Available from http://ipums.org/

  • U.S. Department of Transportation, Federal Highway Administration. 2001 National Household Travel Survey. Available from http://nhts.ornl.gov

  • ACCRA Cost of Living Index data for gasoline prices in the chosen metropolitan areas. Available from http://www.coli.org/

Download references

Acknowledgments

We thank seminar participants at Resources for the Future and Arizona State University; and conference participants from the AERE session at the 2012 Allied Social Science Association Meeting, the 13th UC Santa Barbara Occasional Workshop, the 2012 AERE Summer Meetings, and the 2012 Urban Economics Association Conference.

Author information

Authors and Affiliations

  1. Environmental Defense Fund, 257 Park Avenue, South New York, NY, 10010, USA

    Elisheba Spiller

  2. California State University, Long Beach, CA, USA

    Heather Stephens

  3. Duke University, Durham, NC, USA

    Christopher Timmins

  4. University of Virginia School of Law, Charlottesville, VA, USA

    Allison Smith

Authors
  1. Elisheba Spiller
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Heather Stephens
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Christopher Timmins
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Allison Smith
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Elisheba Spiller.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Spiller, E., Stephens, H., Timmins, C. et al. The Effect of Gasoline Taxes and Public Transit Investments on Driving Patterns. Environ Resource Econ 59, 633–657 (2014). https://doi.org/10.1007/s10640-013-9753-9

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10640-013-9753-9

Keywords

Search

Navigation