Per capita gasoline usage for transport in the USA has remained almost constant since the 1980s. This is despite advances in fuel efficiency and what seems like ever increasing gas prices. It would appear that US citizens are simply using the gains in efficiency to drive longer and further and the trend seems to indicate this will continue.

Data Set

What is now the US Department of Transportation publishes an annual collection of statistics called, unsurprisingly, "Highway Statistics" and it has been published since 1945. All of the issues are available online starting from 1946 and they contain a wealth of information about US transportation. In particular, table MF-21 contains the gasoline consumption figures and tables MV-12 (1948-1964) and DL-1 (1965-present) contain the number of driver's licenses in force each year. US population statistics are from United Nations, Department of Economic and Social Affairs, Population Division (2022). World Population Prospects 2022, Online Edition. (Retrieved 2023-07-02).

The data for this study can be found here as (ODS, XLSX, or CSV)

Population and Driver's Licenses

The US Population has been increasing for all of the years included in this study (1950-2021). The annual growth rate averages 1.16% (minimum of 0.31% in 2020 and maximum of 1.92% in 1958). Likewise, the number of individuals with an active drivers license has generally increased over time with an average growth rate of 2.00% (minimum of -0.31% in 2020 and maximum of 5.86% in 1949).

The graph above shows the US population and the population of US drivers over time. From this data, the US population is growing by approximately 2,705,000 each year and the number of driver's licenses in force is growing by 2,451,000 (approximately the same slope).

> summary(lm(population..1. ~ Year, data=data))

Call:
lm(formula = population..1. ~ Year, data = data)

Residuals:
     Min       1Q   Median       3Q      Max
-6447067 -2731168  1469385  2640189  3589339

Coefficients:
              Estimate Std. Error t value Pr(>|t|)
(Intercept) -5.130e+09  3.573e+07  -143.6   <2e-16 ***
Year         2.705e+06  1.799e+04   150.4   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 3173000 on 70 degrees of freedom
  (4 observations deleted due to missingness)
Multiple R-squared:  0.9969,    Adjusted R-squared:  0.9969
F-statistic: 2.261e+04 on 1 and 70 DF,  p-value: < 2.2e-16

> summary(lm(drivers..2. ~ Year, data=data))

Call:
lm(formula = drivers..2. ~ Year, data = data)

Residuals:
     Min       1Q   Median       3Q      Max
-8384076 -3156662  -255735  2987836  8332155

Coefficients:
              Estimate Std. Error t value Pr(>|t|)
(Intercept) -4.715e+09  4.621e+07  -102.0   <2e-16 ***
Year         2.451e+06  2.329e+04   105.3   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 4279000 on 72 degrees of freedom
  (2 observations deleted due to missingness)
Multiple R-squared:  0.9935,    Adjusted R-squared:  0.9935
F-statistic: 1.108e+04 on 1 and 72 DF,  p-value: < 2.2e-16

In the graph above, the number of drivers as a percent of the population is shown. The number of drivers as a percentage of population appears to have reached plateau beginning the 1980's and lasting until the end of the study period. In other words, as the population increases, the number of drivers also increases along with it.

Gasoline Consumption

The amount of gasoline consumed for automobile transportation appears to be growing over time. With notable exceptions in 2020 and 2021. The dip there can be intuitively attributed to COVID-19 and the various "lock downs" putting a halt to travel. By 2021, though, we are almost back to pre-pandemic levels of consumption. I suspect 2022 will be even closer to the trend line, but time will tell.

In the graph above, we can observe that generally speaking gasoline consumption is increasing each year. In addition to the pandemic, other events are observable in this graph. The first is the 1973 Oil Crisis caused by the embargo of the US by OPEC. Later that same decade, the 1979 Oil Crisis caused by a sharp spike in oil prices.

The correlation between time and gasoline consumption is strong (R²=0.966), but even stronger is the gasoline consumption as a function of US population show below (R²=0.977).

This is somewhat intuitive: more people means more consumption, but statistics cannot prove causation.

Consumption Per Person

Let's now turn our attention to how many gallons of gasoline an individual in the US consumes per year. The graph below shows the gallons consumer per each individual in the population and per active driver's license. For example, the US consumed 272 gallons of gasoline per person in 1950 rising to 569 gallons per person in 2019. Per driver's license, the equivalent for 1950 is 648 gallons per license in 1950 and 831 gallons per license in 2019.

But, that's not the only story in this chart. The amount of fuel consumed per person rises almost linearly from 1950 to 1978. Beginning in 1978, though, the graph is close to being flat plateau: neither rising nor falling significantly.

The data is split into two parts: "early" covering 1950-1980 and "late" covering 1981 to 2021. For the early data, gasoline consumption increased at a rate of 9.3 gallons per person per year (p < 0.001 on a test of positive slope).

> data <- read.csv('[docs.google.com/spreadshe...](https://docs.google.com/spreadsheets/d/e/2PACX-1vRt9Mp-HynoCNKbjiuWQpz4Vgo8kGzjwiWdkwVeEVaBUGmN0TkBkhtlHiWm75x0hmhvJwiQOvBrdib-/pub?gid=1407062100&single=true&output=csv'))
> data = subset(data, data$Year != "")
> early.data = subset(data, data$Year <= 1980)
> late.data = subset(data, data$Year > 1980)
> summary(lm(Gallons.driver ~ Year, data=early.data))

Call:
lm(formula = Gallons.driver ~ Year, data = early.data)

Residuals:
     Min       1Q   Median       3Q      Max
-109.015  -14.785   -0.328   10.699   82.370

Coefficients:
              Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.759e+04  1.199e+03  -14.67 1.70e-15 ***
Year         9.357e+00  6.106e-01   15.32 5.17e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 33.4 on 31 degrees of freedom
  (2 observations deleted due to missingness)
Multiple R-squared:  0.8834,    Adjusted R-squared:  0.8796
F-statistic: 234.8 on 1 and 31 DF,  p-value: 5.17e-16

For the late data (1981 to 2021), the growth rate is estimated to be 0.53 gallons per person per year, a very significant change, but not far from zero. In fact we cannot reject the null hypothesis that the slope is 0 (p = 0.251) (i.e. since 1981, the consumption of fuel per person is unchanged).

> summary(lm(Gallons.driver ~ Year, data=late.data))

Call:
lm(formula = Gallons.driver ~ Year, data = late.data)

Residuals:
    Min      1Q  Median      3Q     Max
-98.019 -17.147  -5.508  15.536  64.814

Coefficients:
             Estimate Std. Error t value Pr(>|t|)
(Intercept) -239.5875   918.8011  -0.261    0.796
Year           0.5352     0.4591   1.166    0.251

Residual standard error: 34.53 on 38 degrees of freedom
  (1 observation deleted due to missingness)
Multiple R-squared:  0.03452,   Adjusted R-squared:  0.009113
F-statistic: 1.359 on 1 and 38 DF,  p-value: 0.251

Miles of Driving

Let's examine our driving habits as a whole. The Highway Statistics book from 2021 contains a table VM-202 which contains the aggregate number of miles driven by vehicles in the US. From the chart below, we are, as a population, driving more every year (COVID-19 notwithstanding). A linear model captures 91% of the variance and the trend is clearly upwards.

Now, are we driving more per person? The chart below shows the change in time for miles driven per person and per driver. The number of miles driven appears to have reached a peak in the early 2000s and largely leveled off (COVID-19 excepted here again, of course).

Fleet Aggregate MPG

Let's treat the entire fleet of vehicles as an aggregate. We know the number of miles driven on US roads (from table VM-202 in the 2021 Highway Statistics book) and we know the total number of gallons consumed. That's miles per gallon at the national scale!

Through the 1980's our fleet aggregate MPG steadily improves from year to year. In approximately 1991, progress appears to halt. Taken as a whole our vehicles are NOT getting any better gas mileage! We appear to have stalled at an aggregate of approximately 17 MPG.

However, the statistics say we can reject the hypothesis that the slope is zero (p < 0.001), but the estimated improvement to the US fleet MPG since 1991 is 0.03 MPG per year (that is we're getting 158.4 feet per gallon improvement per year). Doesn't seem like much, does it?

> ninetiesup.data = subset(data, data$Year > 1990)
> summary(lm(miles.per.gallon ~ Year, data=ninetiesup.data))

Call:
lm(formula = miles.per.gallon ~ Year, data = ninetiesup.data)

Residuals:
     Min       1Q   Median       3Q      Max
-0.33018 -0.08906 -0.04173  0.07308  0.29468

Coefficients:
              Estimate Std. Error t value Pr(>|t|)
(Intercept) -42.728244   5.684002  -7.517 2.76e-08 ***
Year          0.029676   0.002833  10.474 2.29e-11 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.1411 on 29 degrees of freedom
Multiple R-squared:  0.7909,    Adjusted R-squared:  0.7837
F-statistic: 109.7 on 1 and 29 DF,  p-value: 2.287e-11

Image credit: Photo by Jesse Donoghoe on Unsplash