Driving the Vote: Gasoline Expenditures and Swing District Voting

Introduction

US political leaders have long been caught in a bind around gasoline prices. Voters are more likely to vote against incumbents when gasoline prices are high¹, yet political leaders have very limited ability to affect gas prices.²

This report finds a strong correlation between household gasoline consumption and expenditures and a decreased vote share for Joe Biden in swing counties in the U.S. in 2020. The finding is based on data on consumers’ annual gasoline use and the distribution of 2020 presidential election votes across census block groups in swing counties.

It posits that political leaders may be able to avoid the gasoline price bind by addressing voters’ overall fuel expenditures rather than gasoline prices.

Methodology

Data on average annual gasoline consumption with census block group level resolution was sourced from Coltura’s Census-Level Gasoline Model.³ Data on voting with block group level resolution comes from the Harvard Dataverse.⁴ Data on the racial distribution, median household income, and population density of census block groups come from the US Census.⁵ Gas price data is sourced from GasBuddy.⁶

US counties were classified as swing counties if they are in one of the eight historic swing states (Arizona, Georgia, Michigan, Minnesota, North Carolina, Nevada, Pennsylvania, and Wisconsin), and if the differences between 2020 Trump and Biden vote shares in the county were less than 5%. These counties were filtered to exclude counties with fewer than 40 census block groups to reduce bias from small sample size, thereby narrowing the analysis to 24 swing counties.

To isolate gasoline consumption, we control for the confounding variables race, income, and population density. Regression analysis was run on the block groups within the 24 swing counties, with average annual gasoline consumption for 2024 (i.e. gasoline consumption per driver) as the independent variable and 2020 Democrat vote share as the dependent variable.⁷ Race, income, and population density were included as control variables. A similar regression analysis was run for all census block groups in “swing” counties in all 50 states (according to the same 5% criteria), which also included gas price as a control variable. A third regression was also run for all US census block groups, regardless of county, including the same inputs.

Findings

In 23 of the 24 swing counties analyzed, we found that an increase in average annual gasoline consumption and expenditure in a block group was associated with a decrease in votes for Biden in those counties, even when controlling for race, income, and population density (detailed findings are in Table A of the Appendix).

For example, in Eaton County, Michigan an increase of 213 gallons in consumers’ average annual gasoline consumption (one standard deviation) in a block group is associated with a decreased Biden vote share of 9.26 percentage points. Block groups in Eaton where annual gasoline consumption was higher were, to a statistically significant degree, associated with fewer Biden votes, even when controlling for income, race, and population density.

This finding holds with gasoline expenditure replacing gasoline consumption. In Eaton, the same decreased Biden vote share of 9.26 percentage points is associated with an average increase of $751 in gasoline expenditures. The block groups where people spent the most on gasoline were the least likely to vote for Biden in 2020.

The following scatterplots illustrate these correlations:

Eaton County, Michigan: Block Group Average Annual Gasoline Consumption vs. Democrat Vote Share

Eaton County, Michigan with each block group represented by a point, the x-axis representing that block group’s average annual gasoline consumption in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

Eaton County, Michigan: Block Group Average Annual Gasoline Expenditure vs. Democrat Vote Share

Eaton County, Michigan with each block group represented by a point, the x-axis representing that block group’s average annual gasoline expenditure in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

Erie County, Pennsylvania: Block Group Average Annual Gasoline Consumption vs. Democrat Vote Share

Erie County, Pennsylvania with each block group represented by a point, the x-axis representing that block group’s average annual gasoline consumption in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

Erie County, Pennsylvania: Block Group Average Annual Gasoline Expenditure vs. Democrat Vote Share

Erie County, Pennsylvania with each block group represented by a point, the x-axis representing that block group’s average annual gasoline expenditure in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

Racine County, Wisconsin: Block Group Average Annual Gasoline Consumption vs. Democrat Vote Share

Racine County, Wisconsin with each block group represented by a point, the x-axis representing that block group’s average annual gasoline consumption in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

Racine County, Wisconsin: Block Group Average Annual Gasoline Expenditure vs. Democrat Vote Share

Racine County, Wisconsin with each block group represented by a point, the x-axis representing that block group’s average annual gasoline expenditure in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

In swing counties with statistically significant results, an increase in average annual gasoline consumption of 100-300 gallons ($315-$1200 in gasoline expenditure) is associated with about a 1-12 percentage point drop in votes for Joe Biden. For all US block groups, the results indicate a 6.6 percentage point decrease in Biden votes where average annual gasoline consumption is 210 gallons higher, and gasoline expenditures are $790 higher.

This trend held for 23 of the 24 swing counties analyzed. The one county where the trend did not hold was Nash County, North Carolina, where Democratic-leaning long-distance knowledge-worker commuters to Raleigh’s tech industries may have confounded the analysis.

There was also a negative association between gasoline consumption and Biden vote share amongst all US census block groups, and all US census block groups in swing counties.

Implications

While the results of this analysis cannot prove causation – that higher gasoline consumption and expenditure caused people to vote against Joe Biden – it does demonstrate a strong correlation between gasoline consumption and expenditure and voting. One possible explanation is that in 2020, voters spending more on gasoline were more likely to perceive that Donald Trump would do more to lower gasoline prices and reduce their fuel expenditures.⁸ There is, however, no statistically significant relationship between the party of the current president and average inflation-adjusted gas prices based on data from 1950-2023.⁹

An alternative approach to appealing to voters with high gasoline expenditures is demonstrating a credible plan to reduce their fuel costs. For example, policies increasing availability and affordability of electric vehicles would tend to reduce voters’ fuel expenditures. Average fuel savings in the US for a gas car driving 30,000 miles a year and consuming 1,191 gallons (towards the upper end of average gasoline consumption observed in this study) if the car were replaced by an EV would be $2,442, and result in fuel savings of 57 percent.¹⁰ The example from Eaton County, MI (see below) illustrates that areas with the highest potential EV fuel savings tended to vote against Joe Biden. Voters with higher fuel expenditures may be more likely to support candidates who they perceive as making it easier to obtain electric vehicles. EV cost savings are the top reason (74%) that EV owners cite for purchasing an EV.¹¹

Eaton County, Michigan: Block Group Average Potential Annual Fuel Savings for EV Switch vs. Democrat Vote Share

Eaton County, Michigan with each block group represented by a point, the x-axis representing that block group’s average fuel savings for switching to an EV (cost of gasoline minus electricity) in 2024, and the y-axis representing its vote share for Joe Biden in 2020. A line of best fit is added to indicate the trend. Source: Coltura Census-Level Gasoline Model.

Conversely, lowering or eliminating federal gasoline taxes would have a relatively small effect on fuel expenditures, given that the $0.184 cents per gallon federal tax on gasoline¹² constitutes only 5% of average US gasoline prices, an amount within normal price fluctuations.

Conclusion

This research highlights a significant correlation between gasoline consumption and expenditure and voting patterns, particularly in swing counties. It reveals that higher average annual gasoline consumption and expenditure was associated with a decreased vote share for Joe Biden in 2020, even after accounting for confounding variables such as race, income, and population density. This correlation underscores the importance of gasoline expenditure as a factor in voter decision-making and suggests that voters may embrace policies that they believe will reduce their overall fuel expenditures. In areas with high levels of gasoline consumption, electric vehicles may save drivers, on average, more than $2000 annually on fuel expenditures. As such, political leaders should consider focusing more on policies to reduce fuel expenditures rather than transitory, short-term cuts in gasoline prices which do little to reduce overall fuel expenditures. While these policies require a longer time period for implementation, they will lead to lower annual fuel expenditures for the drivers that spend the most on gasoline, and should tend to mitigate the influence of fuel expenditures on voting.

Appendix

Table A: Results of OLS regression analysis (robust standard errors)

Full results of OLS regression available here.

County	Sample size (number of block groups)	Coefficient between average annual gasoline consumption and Democratic vote share¹³	Standard deviation of independent variable	P-value
Maricopa County, Arizona	2,475	-3.58701***	127.6075	9.93E-21
Eaton County, Michigan	80	-9.26478***	212.6689	2.54E-21
Grand Traverse County, Michigan	56	-5.22929***	216.3478	3.45E-05
Isabella County, Michigan	43	-9.34602***	254.1253	2.58E-11
Muskegon County, Michigan	136	-6.06093***	179.078	3.73E-15
Saginaw County, Michigan	173	-5.37519***	229.2295	6.81E-16
Anoka County, Minnesota	229	-4.29735***	140.3039	1.28E-19
Blue Earth County, Minnesota	51	-9.67581***	289.9326	1.68E-33
Carver County, Minnesota	55	-8.13353***	158.2235	2.87E-32
Clay County, Minnesota	44	-6.46712***	293.4972	0.000112
Rice County, Minnesota	42	-7.03176**	198.0397	0.03759
Washoe County, Nevada	305	-4.71842***	116.57	6.30E-06
Nash County, North Carolina	63	-3.11318	170.0239	0.121856
New Hanover County, North Carolina	120	-2.85597**	94.2209	0.010232
Wilson County, North Carolina	47	-9.47345***	140.586	0.005886
Bucks County, Pennsylvania	376	-0.81889**	92.31004	0.028586
Centre County, Pennsylvania	99	-12.2308***	323.9844	1.95E-15
Erie County, Pennsylvania	231	-6.5452***	226.6243	4.43E-24
Northampton County, Pennsylvania	199	-4.54524***	104.6126	3.76E-15
Columbia County, Wisconsin	48	-4.02107***	159.0249	0.003953
Kenosha County, Wisconsin	127	-5.25087***	152.8027	8.84E-19
Portage County, Wisconsin	53	-5.69209***	240.1293	1.58E-08
Racine County, Wisconsin	167	-7.0218***	159.6567	6.36E-26
Winnebago County, Wisconsin	131	-4.93735***	155.1352	8.94E-43
All swing block groups	19,018	-3.80457***	185.0445	0
All block groups	198,456	-6.62694***	209.874	0

Endnotes

Sung Eun Kim and Joonseok Yang, “Gasoline in the Voter’s Pocketbook: Driving Times to Work and the Electoral Implications of Gasoline Price Fluctuations,” American Politics Research 50, no. 3 (September 8, 2021), https://doi.org/10.1177/1532673x211043445
Dylan Brewer and Matthew E Oliver, “Of Presidents and Prices at the Pump: Do Economists Have a Blind Spot?,” SSRN Electronic Journal, December 1, 2023, https://doi.org/10.2139/ssrn.4640592.
Til Dietrich et al., “Coltura Methodology: Gasoline Consumption Data and Map” (Coltura, 2024), https://data.coltura.org/methodology.
Michael Bryan, “U.S. Voting by Census Block Groups,” Harvard Dataverse, January 1, 2022, https://doi.org/10.7910/dvn/nknwbx.
United States Census Bureau, “Explore Census Data,” United States Census Bureau, 2022, https://data.census.gov/.
“US Average Regular Gas Price Comparison by State” (GasBuddy), accessed August 7, 2024, https://www.gasbuddy.com/usa.
One limitation of this study is that its gasoline consumption and expenditure data are from 2024 and they are being correlated with 2020 voting patterns, a limitation mitigated by the stability of vehicle miles traveled and gasoline consumption in recent years. See a) US Federal Highway Administration and b) Energy Information Administration.
a) “Travel Trends - December 2023” (U.S. Department of Transportation Federal Highway Administration, 2023), https://www.fhwa.dot.gov/policyinformation/travel_monitoring/23dectvt/page2.cfm.
b) “U.S. Product Supplied of Finished Motor Gasoline (Thousand Barrels per Day)” (U.S. Energy Information Administration, July 31, 2024), https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=MGFUPUS2&f=A.
While presidents have frequently claimed that they will help lower fuel costs, they have very limited ability to control gasoline prices.
Dongfang Nie, Lacia Engels, and Jingbo Zhang, “Presidential Policies’ Effects on Oil and Gas Price and Productions within the United States from 1977-2021,” The Journal of Applied Business and Economics 24, no. 4 (October 9, 2022), https://doi.org/10.33423/jabe.v24i4.5491.
A t-test of inflation-adjusted gas prices from 1950-2023 for Democratic vs. Republican presidents resulted in a t-statistic of 0.67 with p-value 0.51, demonstrating that there is no statistically significant difference in gas prices between Democrat and Republican presidencies. Inflation-adjusted gas prices sourced from the Alternative Fuels Data Center.
“Average Annual Retail Fuel Price of Gasoline,” Alternative Fuels Data Center, January 2024, https://afdc.energy.gov/data/10641.
Xinyi Wu, Yan Zhou, and David Gohlke, “Adoption of Plug-in Electric Vehicles: Local Fuel Use and Greenhouse Gas Emissions Reductions across the U.S. Energy Systems and Infrastructure Analysis Division” (Argonne National Laboratory, February 2024), https://publications.anl.gov/anlpubs/2024/02/187786.pdf.
Amanda Schoenbauer, “Charging Ahead in the Era of Electric Vehicles (EVs),” Numerator, May 3, 2024, https://www.numerator.com/resources/blog/charging-ahead-electric-vehicles/.
“How Much Tax Do We Pay on a Gallon of Gasoline and on a Gallon of Diesel Fuel?” (U.S. Energy Information Administration (EIA), January 9, 2024), https://www.eia.gov/tools/faqs/faq.php?id=10&t=10.
Level of confidence are indicated as such:
* p < 10%
** p < 5%
*** p < 1%

About The Authors

Matthew Metz
Matthew Metz is the founder and co-executive director of Coltura. He led the coalition supporting passage of legislation that made Washington the first state in the nation to set a 2030 target for all new cars to be electric. Matthew writes about legal, political, and cultural strategies to transition away from gasoline. Matthew received an A.B. in Economics from the University of Chicago and a law degree from UCLA.
Ron Barzilay
Ron Barzilay is a Data and Policy Associate at Coltura, specializing in data analysis and visualization to inform and develop public policy. He graduated from the University of California, Berkeley, with bachelor’s degrees in Data Science and Political Science.

Driving the Vote:

Gasoline Expenditures and Swing District Voting