Home > Consumption & Efficiency >Transportation Surveys > Method for Estimating Fuel Economy

Return to: Transportation Home Page 


 

Change in Method for Estimating Fuel Economy for the Residential Transportation Energy Consumption Survey*


The current method for collecting data on vehicle fuel economy for the Residential Transportation Energy Consumption Survey is to use the Environmental Protection Agency's fuel-economy ratings. The previous method, abandoned because of budgetary restraints, was to use fuel-purchase logs. Although the two methods have to date yielded similar estimates of fuel economy, the current method may yield less reliable estimates over time.

In 1988, the Energy Information Administration (EIA) changed the way it measured vehicle fuel economy for its Residential Transportation Energy Consumption Survey (RTECS). The RTECS, which collects information on household vehicles and the energy they consume, is a national multistage probability sample survey that has been conducted every 3 years since 1985.1   For the 1983 and 1985 RTECS, EIA used fuelpurchase logs kept by selected respondents to estimate fuel economy. For the 1988 RTECS, EIA used the U.S. Environmental Protection Agency's (EPA) fuel-economy ratings. The purpose of this article is to analyze the differences between the two collection metComputing fuel economy in that manner gave accurate on-road monthly estimates. Those monthly estimates were converted to annualized estimates based on the observed way the fuel economy varied for different months of the year. The response rates for completion of the fuel purchase logs were 46 percent (3,526) for the 1983 RTECS and 40 percent (3,413) for the 1985 RTECS. For those vehicles for which no log data were available, the annualized fuel economies were imputed by using the medians of categories defined by vehicle type, make, model, and model year.hods and the resulting estimates of fuel economy.


Fuel-Purchase Logs

For the 1983 and 1985 RTECS, fuel-economy estimates were based on data from fuel-purchase logs kept by selected respondents. During a particular month of the survey year, each respondent recorded the odometer reading and the amount of fuel purchased at the time of each fuel purchase. That information was then used to compute the fuel economy of the vehicles in miles per gallon.

EPA Ratings

Beginning with the 1988 RTECS, EIA stopped collecting the fuel-purchase data, due to budget limitations.2   Instead of fuel-purchase logs, EIA developed a new approach to estimate fuel economies based on EPA's fuel-economy ratings. After applying EPA fuel-economy ratings (in miles per gallon) to RTECS sample vehicles, EIA multiplied the derived fuel-economy ratio by the annual vehicle miles traveled to derive annual consumption.

EPA's fuel-economy ratings are part of EPA's annual certification files, which contain extensive information on vehicles. The EPA files provide three fuel-economy ratings: city, highway, and composite. The composite rating is formed by combining the city and highway ratings, assuming a "typical" vehicle-use pattern of 55-percent city driving and 45-percent highway driving.

EPA bases its ratings on the fuel economy of test vehicles under simulated driving conditions, adjusted for on-road use. This adjustment uses discount factors, which EPA developed in the early 1980's. The factors reduce the highway ratings by 22 percent and the city ratings by 10 percent. They were developed based on vehicles of late 1970's and early 1980's vintage.

EIA further adjusts the EPA ratings for individual driving circumstances before reporting them in the RTECS. This adjustment takes into consideration such factors as urban versus rural driving patterns, traffic congestion, seasonal temperature variations, humidity levels, geographic variations, altitude, wind, and road gradient and surface conditions. This adjustment uses a regression model unique to each vehicle, based on the vehicle's average miles per day and its geographic location.3

Although these adjustments are the best available for this estimation methodology, they are still unlikely to capture many of the inherent differences among vehicles and drivers that may affect fuel economy.

Comparing the Two Methods

To compare the two methods for estimating fuel economy, estimates from both methods were needed. Therefore, either log-based fuel-economy estimates for the 1988 and 1991 RTECS vehicles or adjusted EPA fuel-economy ratings for the 1983 and 1985 RTECS vehicles had to be obtained. Accordingly, three steps were taken:

  1. Base EPA fuel-economy ratings for the 1983 and 1985 RTECS vehicles were obtained by matching them with vehicles in EPA's certification files, which list extensive information on each make and type of vehicle of model years 1975 and later. Some of the RTECS vehicles had to be excluded, either because they were of model years earlier than 1975, or because their make or model names could not be determined. As a result, 2,799 of the 1983 RTECS vehicles and 2,297 of the 1985 RTECS vehicles were excluded from the analysis, leaving 4,866 and 6,194 usable vehicle records in the 1983 and 1985 RTECS databases, respectively.
  2. The EPA fuel-economy ratings for these vehicles were adjusted for on-road use and individual driving circumstances, using the same method as had been used in the 1988 and 1991 RTECS
  3. Estimates of average fuel economy for the U.S. vehicle stock and various subpopulations were computed using both methods, and differences were analyzed.
Results of Comparison of Methods

A comparison of the results of the two methods showed that for the total vehicle stock, the EPA-based estimates were slightly higher than log-based estimates in both 1983 and 1985 (1 percent higher for 1983 and 4 percent higher for 1985) (Tables 1 and 2). However, the EPA-based estimates were noticeably lower than the log-based estimates for vehicles in certain categories:

  • For vehicles that traveled fewer than 5,000 miles per year, the EPA-based estimates were 6 percent lower in 1983 and 3 percent lower in 1985.
  • For vehicles whose primary drivers were 66 years of age or older, the EPA-based estimates were 4 percent lower in 1983 and 3 percent lower in 1985.
  • For vehicles with model years of 1975 or 1976, the EPA-based estimates were 3 percent lower in l 983 and 2 percent lower in 1985.
  • For vehicles with rotary engines, the EPA-based estimates were 3 percent lower in 1983 and 8 percent lower in 1985. However, this difference was likely due not to the method used but to the fact fhat rotary engine vehicles were older and relatively rare (only 12 out of 4,866 sample vehicles in 1983 and 9 out of 6,194 in 1985).

Analysis of the data from the 1983 and 1985 RTECS indicates that the EPA-based method for computing fuel economy yielded consistent results when compared with the log-based method. However, for the total vehicle stock, the difference between the two estimates was only l percent in 1983 but 4 percent in 1985, suggesting the possibility that the differences between the two types of estimates may increase over time.

This possible trend is most apparent when analyzing the data disaggregated by vehicle model year. Within the 1983 and 1985 data, the ratio of the log-based estimates to the EPA-based estimates is close to 1 for early model years (1975 to 1980). However, for newer vehicles (early to mid-1980's), the ratio decreases, indicating that the EPA-based estimates tend to be even higher than the log-based estimates for newer vehicles.

Furthermore, the EPA adjustment for on-road use has not been updated since it was developed in the early 1980's. Because the adjustment uses factors based on technology features and driving conditions that pertained to vehicles used in the early 1980's, it may be inappropriate for evaluating vehicles used today. For example, increased urbanization, higher speed limits in nonurban areas, and increased traffic congestion in recent years could mean that the difference between the EPA fuel-economy ratings and actual on-road fuel economy has increased to the point that current adjustment procedures are no longer adequate.

Due to the errors associated with the data, the above observations cannot be stated definitively. The standard errors indicate that the ratios of the differences between the two estimates are not statistically significant. Also, more than two data points are necessary before any apparent trend can be said to be meaningful.

The Argonne National Laboratory, which conducted an analysis similar to EIA's on 1985 vehicles, found a much larger shortfall between the log-based estimates of fuel economy and the EPA-based estimates.4   The reason for the difference, however, is that Argonne adjusted the EPA ratings for on-road vehicle use but not for individual driving circumstances.

Summary

In 1988, EIA changed the way it measured vehicle fuel economy for the RTECS. For the 1983 and 1985 RTECS, EIA used fuel-purchase logs kept by selected respondents to estimate fuel economy. For the 1988 RTECS, EIA used a method based on EPA's fuel-economy ratings.

A comparison of the fuel-economy ratings in the 1983 and 1985 surveys with the fuel-economy ratings that would have resulted had the current method been used to determine those ratings shows that the two methods yield similar results. Therefore, the current method of estimating fuel economy, using adjusted EPA ratings, seems adequate.

However, EPA-based ratings were slightly higher than the log-based ratings, and the differences were higher for 1985 than for 1983. This possible trend suggests that the current method for estimating fuel economy may have to be further adjusted to take account of changes in driving conditions, such as increased traffic congestion, that have taken place in recent years.

End Note(s)

   *The author, Jennifer Reichert, was a mathematical statistician in the Energy End Use Division, Office of Energy Markets and End Use, Energy Information Administration, however, she has since left the agency.

1.   EIA conducted the RTECS every month from June 1979 to September 1981. In 1983, EIA conducted the first RTECS covering a calendar year and conducted subsequent RTECS in 1985, 1988, and 1991.

2.   However, EIA continued to collect the annual vehicle-miles traveled, using beginning-of-year and end-of-year odometer readings.

3.   Both adjustments are described in detail in both the 1988 and 1991 RTECS publications. See Energy Information Administration, Household Vehicles Energy Consumption 1988, DOE/EIA-0464(88) (Washington, DC, February 1990), pp. 101-107; or Household Vehicles Energy Consumption 1991, DOE/EIA 0464(91) (Washington. DC, December 1993), pp. 131-137.

4.   Marianne Mintz, et. al., "Differences Between EPA-Test and In-Use Fuel Economy: Are the Correction Factors Correct?" Transportation Research Record 1416 (Washington, DC, 1993),pp. 124-130.