Download presentation
Presentation is loading. Please wait.
Published byGervase Hunt Modified over 8 years ago
1
Vehicle Size Choice and Automobile Externalities: A Dynamic Analysis Clifford Winston Jia Yan Brookings Institution Washington State Univ. Cwinston@brookings.eduCwinston@brookings.edu Jiay@WSU.eduJiay@WSU.edu
2
Broader Implications of Larger Vehicles on the Road In 2000, 20% of new vehicles sold in the US were SUVs; in 2012, that percentage climbed to 33% This could increase the risk of a serious accident to occupants of smaller vehicles Larger vehicles also consume more fuel and thus produce greater emissions
3
What we do in this paper The purpose of this study is to provide a disaggregate analysis of the effects of various factors including price, operating cost and congestion on vehicle size choice and to explore the implications for optimal polices to address automobile externalities Preliminary results suggest congestion may notably affect vehicle sizes and thus automobile externalities. Optimal policies will then be explored
4
Why does congestion matter for vehicle size choice? Consumers are willing to pay a safety premium for larger vehicles. Vehicle collisions increase with traffic congestion.
5
Policy implications from our study Gasoline tax is used to address the externalities, excessive fuel consumption, of large vehicles. Our study suggests that policies reducing congestion such as congestion pricing have impacts on vehicle size, which in turn affects fuel consumption and safety.
6
Data On Individuals, 2004-2009 A random sample of 475 motorists who lived and worked in the Seattle, Washington metro area and owned a car as of 2009 Information on: ◦ Car ownership, residential and employment location, and personal profiles such as income, age, and education etc. of the individuals ◦ The route and time of day usually taken by the individuals for their trips to work. ◦ Congestion (travel delay) on the routes used by the individuals—determined as the difference between actual and free flow travel time About 1/3 of the motorists changed their vehicles between 2004 and 2009.
7
Commuting in Seattle The Seattle Metropolitan Area is adjacent to the Puget Sound and motorists must often travel over a body of water to get to their destinations. There are many bottlenecks that contribute to congestion that are created by the bridges that people must cross when they drive into Seattle
10
Defining Choice Set We consider motorists’ choice among 13 combinations of vehicle class and size: ◦ compact domestic; ◦ compact luxury imported; ◦ compact pickup; ◦ full SUV; ◦ full size domestic; ◦ full size luxury imported; ◦ midsize SUV; ◦ midsize domestic; ◦ midsize luxury imported; ◦ passenger van; ◦ standard pickup; ◦ subcompact domestic; ◦ subcompact luxury imported Given current vehicle ownership, a motorist chooses a vehicle class and size combination from the most recent 10 model years.
11
Developing A Dynamic Model of Vehicle Holding and Replacement Most studies on vehicle choice use data on new vehicle purchase decisions. Because we have a short panel, we identify motorists’ preferences for vehicle attributes including price, operating cost and vehicle size under different congestion environments from ◦ motorists’ vehicle purchase decisions to replace currently owned vehicles AND ◦ motorists’ decisions to keep currently owned vehicles.
12
Why do we need a dynamic model? Motorists do not change their vehicles frequently and when they do change, they sell their vehicle in the used-vehicle market. The price of a motorist’s vehicle depreciates and the maintenance costs of the vehicle increases with vehicle age. Vehicle operating costs, which depend on both mileage-per-gallon (mpg) of vehicles and gasoline price, evolve with time because of the fluctuations in gasoline prices Vehicle attributes evolve over time because of technological advances in automobiles
13
Plausible Assumptions to reduce dimensionality Consumers do not predict the evolution of the automobile industry’s vehicle offerings. Motorists make reasonable predictions of gasoline prices and base their vehicle replacement decisions on those predictions (Busse, Knittle and Zettelmeyer (2013)). Motorists monitor the depreciation of their vehicles. Congestion is persistent and stable.
14
The Dynamic Vehicle-size Choice Model
18
We specify the key congestion variable as an interaction Delay is determined as the difference between actual travel time and free flow travel time. We define it as a dummy of a long commute (>= 1 hour) × dummy for excessive delay (15% or more of total commute time) We interact the delay faced by a motorist with three vehicle dummies – full size SUV, medium size SUV and standard pickup, and vehicle operating cost. ◦ Motorists are more likely to buy larger vehicles because of the disutility from traveling in congestion. ◦ But motorists are more likely to buy more fuel- efficient vehicles when congestion is higher.
20
Identification strategy for endogenous vehicle attributes In baseline models, we use (12) vehicle class and size combination dummies to control for omitted vehicle attributes. The underlying assumption is that omitted attributes of a vehicle class-size combination do not vary across model years. This assumption is justified to the extent that vehicle design is stable over time.
21
Identification strategy for endogenous delay Possible strategies ◦ Modeling location and vehicle choice jointly: infeasible because of the complications of modeling location as a disaggregate dynamic choice. ◦ Using motorist dummies to control for unobserved preferences for housing and location: infeasible because of the incidental parameter problem caused by nonlinearity in parameters and a short panel. Our strategy is to follow the assumption of “selection on observables” to use several observed housing and location characteristics as control variables and capture any unobserved influences related to location choice
23
Observed housing and location characteristics House square footage Zillow home value index of the zip code Median household income of the zip code School index of the zip code Personal crime index of the zip code Property crime index of the zip code
30
Results from myopic choice
31
Results from dynamic choice
32
Results from dynamic choice without using housing and location characteristics as control variables
33
Summary of baseline results Consumers are more likely to buy full-size SUV and standard pickups when facing congestion. Offset effect of congestion: consumers are also more likely to buy fuel-efficient vehicles when facing congestion. Simulations using the dynamic model show that removing congestion would reduce the market shares of large SUV by 13% and reduce the market shares of standard pickups by 15% respectively. Additional policy simulations will be performed
34
Comparison between dynamic and myopic model Compared with myopic model, dynamic model fits the data better (log-likelihood values of myopic and dynamic model are -3083.45 and -3042.78 respectively) Myopic model overestimates the responsiveness to price and underestimates the effect of congestion on motorists’ willingness-to-pay for more fuel efficient vehicles.
35
Robustness Checks Check1: Split the sample into two subsamples: one contains zip codes with median household income > 50k, and one contains zip codes with median household income <= 50k. Results from the two subsamples are in line with the baseline results. Check 2: Change the price depreciation from 15% to 25%. Results are still in line with the baseline findings.
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.