Step on It: Driving Behavior and Vehicle Fuel Economy Ashley Langer and Shaun McRae University of Arizona and University of Michigan November 1, 2014
How do we decrease gasoline use? Drive more efficient cars (CAFE standards, Feebates, Cash for Clunkers etc) Busse, Knittel, and Zettelmeyer (2012), Sallee (2011), Goldberg (1998), Allcott and Wozny (2012), many, many others Drive less (Gasoline taxes, VMT taxes etc.) Hughes, Knittel, and Sperling (2008), Levin, Lewis, and Wolak (2012), Gillingham (2012) among others Drive better (Speed limits...?) van Benthem (2013) Burger and Kaffine (2010), Wolff (2014)
Could policy improve fuel economy on the road? In this paper, we use a dataset of individual-level real-world driving behavior to better understand: How does fuel economy vary across drivers? Where is there room for improvement? In real-world driving, how does driver behavior affect fuel economy? What policies would be effective to improve driver fuel economy?
Outline of the talk 1 Introduction 2 Data 3 Evidence on the variation in fuel economy 4 Model of driver and vehicle 5 Trade-off between fuel consumption and time 6 Policy simulations
Data: The IVBSS Experiment IVBSS (Integrated Vehicle-Based Safety System) was a $32 million field test of advanced crash-warning technology by the USDOT, industry partners, and the UM Transportation Research Institute (UMTRI) Sixteen identical passenger cars were fitted with the technology
Participants in the study 108 drivers from southeast Michigan were given the vehicles to use for approximately six weeks Had to have clean driving records and drive a substantial amount. Sample stratified by age group and gender 20 30 years, 40 50 years, 60 70 years Experiment ran from April 16, 2009 to May 11, 2010 approx. 213,000 miles of driving in 23,000 trips
What data was collected during the experiments? Each car had a computer installed that recorded 600 variables at a rate of 10 times per second Vehicle location, speed, acceleration, fuel use, etc Detailed data from the crash warning systems Each car included five cameras (two in-car, three exterior)
Outline of the talk 1 Introduction 2 Data 3 Evidence on the variation in fuel economy 4 Model of driver and vehicle 5 Trade-off between fuel consumption and time 6 Policy simulations
Variation in fuel economy across trips 0.10 EPA highway (26 mpg) EPA city (18 mpg) 0.08 Fraction of sample 0.06 0.04 0.02 0.00 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 Average fuel economy (L/100 km)
Variation in fuel economy across drivers 0.30 0.25 EPA highway fuel economy (26 miles per gallon) EPA city fuel economy (18 miles per gallon) Fraction of sample 0.20 0.15 0.10 0.05 0.00 8.0 9.0 10.0 11.0 12.0 13.0 14.0 Average fuel economy (L/100 km)
Decomposing average driver fuel use (L/100km) All independent variables are z-scores Female (0/1) 0.73*** 0.15 (0.20) (0.09) Age 40-50 (0/1) -0.69** -0.01 (0.26) (0.11) Age 60-70 (0/1) -0.064** 0.05 (0.25) (0.11) Speed given Speed > 100km/h 0.20*** (0.05) Idle time (0-1) 0.35*** (0.10) Acceleration (m/s 2 ) 0.17*** (0.05) Acceleration events per km 0.71*** (0.10) Adjusted R 2 0.211 0.897
Outline of the talk 1 Introduction 2 Data 3 Evidence on the variation in fuel economy 4 Model of driver and vehicle 5 Trade-off between fuel consumption and time 6 Policy simulations
Modeling driver behavior and fuel use In order to empirically understand the effect of policies intended to improve fuel economy, we need a model that connects driver choices to fuel consumption Our model has two main components: 1 A driver maximizes utility given a trip, by choosing the optimal route and, conditional on the route, minimizing a combination of trip time and fuel consumption 2 Fuel consumption is a function of driver behavior (speed, acceleration, stops, etc), given by a physical model of a car that we estimate
Modeling driver behavior and fuel use (2/2) Behavioral model: Tradeoff between trip time and fuel use is determined by the driver s value of time. Requires the physical model to connect driver behavior to fuel use. Physical model: Physical model of the forces acting on the cars gives a non-linear equation for fuel consumption as a function of acceleration, speed, and grade. We estimate a polynomial approximation of this function using the interaction of sixth-order polynomials in velocity and fourth-order polynomials in acceleration, separately modelling positive and negative acceleration.
Fuel consumption per 100 meters for one trip Actual Fuel consumption (milliliters) 35 30 25 20 15 10 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Cumulative trip distance (kilometers)
Estimated model of fuel consumption fits the observed data well Actual Predicted Fuel consumption (milliliters) 35 30 25 20 15 10 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Cumulative trip distance (kilometers)
Outline of the talk 1 Introduction 2 Data 3 Evidence on the variation in fuel economy 4 Model of driver and vehicle 5 Trade-off between fuel consumption and time 6 Policy simulations
Can information or gasoline taxes improve fuel economy? The EPA website provides tips for driving more fuel efficiently in order to save money. Drive more slowly on the freeway Accelerate less aggressively Our model allows us to calculate the value of time that would be required for the fuel economy savings to be worthwhile.
Relationship between speed and fuel consumption is almost flat for a large range of speeds 25 Fuel consumption (liters/100 kilometers) 20 15 10 5 0 0 25 50 75 100 125 150 Speed (kilometers per hour)
You should drive faster (ignoring safety) Over 100 kilometers Constant speed (km/h) 90 100 110 120 130 Fuel consumption (L) 7.16 7.25 7.63 8.27 9.08 Fuel cost ($) 6.62 6.71 7.05 7.65 8.40 Time (minutes) 66.7 60.0 54.5 50.0 46.2 Effect of increasing speed Fuel cost ($). 0.09 0.34 0.60 0.75 Time (minutes). -6.7-5.5-4.5-3.8 Cost of time ($/hour). 0.77 3.78 7.86 11.76
Higher rates of acceleration get to a given speed using less fuel Accelerating from 2-15 m/s 60 Fuel consumption (milliliters) 50 40 30 20 10 Acceleration 0.5.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 Acceleration (m/s2)
Little effect of acceleration on fuel usage over a given distance Accelerating from 2-15 m/s over 250 meters Fuel consumption (milliliters) 60 50 40 30 20 10 Constant speed Acceleration 0.5.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 Acceleration (m/s2)
Model predictions fit the acceleration data very well Accelerating from 2-15 m/s over 250 meters Fuel consumption (milliliters) 60 50 40 30 20 10 Constant speed Acceleration 0.5.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 Acceleration (m/s2)
You should drive more aggressively (ignoring safety) Accelerating from 2-15 m/s over 250m Acceleration (m/s 2 ) 1.0 1.5 2.0 2.5 3.0 2 15 m/s over 250 m Fuel consumption (ml) 45.26 45.16 45.66 46.20 46.74 Fuel cost (cents) 4.18 4.18 4.22 4.27 4.32 Time (seconds) 22.3 20.4 19.5 18.9 18.5 Cost of time ($/hour). -0.18 1.77 3.18 4.76
You should drive more aggressively (ignoring safety) Accelerating from 15-25 m/s over 500m Acceleration (m/s 2 ) 1.0 1.5 2.0 15 25 m/s over 500 m Fuel consumption (ml) 76.71 79.80 80.96 Fuel cost (cents) 7.09 7.38 7.49 Time (seconds) 22.0 21.3 21.0 Cost of time ($/hour). 15.35 11.63
Effect of Varying Speed on the Freeway 1 kilometer at 67 mph Variation in speed (m/s) 0 4 6 8 10 Fuel consumption (L) 0.075 0.095 0.105 0.117 0.129 L/100 km 7.46 9.50 10.54 11.65 12.91
Outline of the talk 1 Introduction 2 Data 3 Evidence on the variation in fuel economy 4 Model of driver and vehicle 5 Trade-off between fuel consumption and time 6 Policy simulations
Simulate gasoline consumption with optimal route choice and driving style Start City route: Shorter distance (10 km) Slower speed (max 40 mph) Intermediate stops (5) End Highway route: Longer distance (15.7 km) Faster speed (max 70 mph) No intermediate stops
Population representative of working adults in Michigan Distribution of personal hourly earnings in 2012 from CPS Value of time assumed to be half the hourly earnings (standard assumption in literature) Each driver chooses both the optimal route (city or highway) and the optimal speed and acceleration along that route Drivers assumed to minimize total cost of travel from start to end locations, subject to constraints on speed: TC = P gas Q gas + VOT t
Slightly less than half of the drivers would choose the city route Base % city route 45.6 Fuel consumption (L) 1.23 Time (minutes) 9.53 Fuel economy (L/100km) 9.41 City 10.84 Highway 8.65 Total cost to driver $2.92 Social cost $3.03
Reduction in fuel consumption from higher gasoline prices more than offset by the increase in trip times Base +$0.50 +$1.00 % city route 45.6 68.4 78.8 Fuel consumption (L) 1.23 1.17 1.14 Time (minutes) 9.53 10.02 10.30 Fuel economy (L/100km) 9.41 9.91 10.16 City 10.84 10.82 10.80 Highway 8.65 8.65 8.65 Total cost to driver $2.92 $3.52 $4.10 Social cost $3.03 $3.04 $3.06
Lower speed limits in the city would increase overall fuel consumption (by shifting more drivers to highway) Base Hwy 65 City 35 % city route 45.6 51.7 29.9 Fuel consumption (L) 1.23 1.19 1.26 Time (minutes) 9.53 9.91 9.59 Fuel economy (L/100km) 9.41 9.30 9.03 City 10.84 10.84 10.55 Highway 8.65 8.25 8.62 Total cost to driver $2.92 $2.97 $2.94 Social cost $3.03 $3.09 $3.05
Eliminating stops along city routes reduces fuel consumption and improves welfare Base 4 stops 3 stops % city route 45.6 56.5 68.4 Fuel consumption (L) 1.23 1.17 1.10 Time (minutes) 9.53 9.63 9.72 Fuel economy (L/100km) 9.41 9.41 9.33 City 10.84 10.34 9.82 Highway 8.65 8.65 8.65 Total cost to driver $2.92 $2.89 $2.85 Social cost $3.03 $3.00 $2.95
Conclusion There is substantial variation in fuel economy across drivers in identical vehicles Most variation in fuel use comes from frequency of acceleration events, not the rate of acceleration or highway speed This suggests that gasoline taxes will be fairly ineffective at improving fuel economy for vehicles on the road Measures to reduce variation in speed (such as vehicle-infrastructure communication) could lead to substantial reductions in fuel use