Responses to Gas Prices in Knoxville, TN

1
Responses to Gas Prices in Knoxville, TN

• Vince Bernardin, Jr., Ph.D. Bernardin,
  Lochmueller Associates
• Mike Conger, P.E.Knoxville Regional
  Transportation Planning Organization

2
Background

• Gas price fluctuations prompt the question
• How are changes in gas prices affecting travel?

3
More Less VMT

• Various studies have attempted to estimate the
  elasticity of VMT to gas prices
• Short term elasticities -0.07 to -0.17
• Long term elasticities -0.22 to 0.33

4
Components of Travelers Response

• Travelers can reduce gas consumption in various
  ways, some easier than others,
• More carpooling
• Destinations closer to each other
• Destinations closer to home
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)
• Lower vehicle ownership (long term)

5
ModelingTravelers Response

• Travelers can reduce gas consumption in various
  ways, some easier than others,
• More carpooling
• Destinations closer to each other
• Destinations closer to home
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)
• Lower vehicle ownership (long term)

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other
• Destinations closer to home
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)
• Lower vehicle ownership (long term)

6
Challenges

• Models have faced two key problems in
  incorporating additional sensitivity to fuel
  prices
• Data limitations
• Structural limitations

7
Data Limitations

• Travel models have traditionally been estimated
  from cross-sectional household survey data
• The resulting lack of variation in fuel prices
  with observed travel behavior has generally
  precluded the incorporation of fuel prices as a
  variable

8
Structural Limitations

• The traditional four-step model design does not
  allow the incorporation of many effects
• Changes in mode and car-pooling can be captured
  in mode choice, but
• The agglomeration of destinations cannot be
  reflected as the gravity model treats all
  destination choices as independent
• Activity participation and touring rates cannot
  respond because cross-classification trip
  production models cannot incorporate fuel price
  as a variable
• Vehicle ownership is generally not modeled at all

9
Overcoming the Challenges

• In Knoxville, we are attempting to overcome both
  challenges
• Travel survey data was collected in both
  2000-2001 and again in 2008, yielding data with
  significant variation in fuel prices
• A new hybrid trip-based/tour-based model design
  has been adopted which overcomes the structural
  limitations of the four-step model

10
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other
• Destinations closer to home
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)
• Lower vehicle ownership (long term)

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other ??
• Destinations closer to home ??
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)
• Lower vehicle ownership (long term)

11
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling

12
Carpooling

• In the Knoxville model, as in activity-based
  models, vehicle occupancy is determined by trip
  mode choice models, distinct from tour mode
  choice

13
Variables
Models
Population Synthesizer
TAZ
Vehicle Availability Choice
Disaggregate Models
Activity / Tour Generation
Accessibility
Tour Mode Choice
Network
Stop Location Choice
Travel Times
Stop Sequence Choice
Aggregate Models
Trip Mode Choice
Flow Averaging
Departure Time Choice
Link Flows
Traffic Assignment
14
Carpooling

• In the Knoxville model, as in activity-based
  models, vehicle occupancy is determined by trip
  mode choice models, distinct from tour mode
  choice
• NL and MNL models of trip mode choice were
  estimated using the combined 2000-2001 2008
  datasets
• The models show a combined elasticity of vehicle
  occupancy with respect to fuel price of 0.128.

15
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other ??
• Destinations closer to home ??

16
Destination Choices

• The new Knoxville model does incorporate
  trip-chaining effects reflecting the fact that
  travelers group their stops into convenient tours
• However, we were unable to directly estimate the
  effect of fuel prices on trip-chaining or
  destination choice due to the limitations of our
  estimation technique

17
Destination Choice

• Analysis of the data using regression did show
  fuel price effects on destination choice
• Trip-based perspective
• Home-based trip length elasticity -0.114
• Non-home-based trip length elasticity -0.064
• Tour-based perspective
• Direct travel time from home to stop elasticity
  -0.036
• Elasticity of destination accessibility 0.042

18
Destination Choice

• Elasticities from regression analysis may be
  incorporated in stop location choice models
  through a heuristic calibration effort
• Time labor intensive process
• Contingent on schedule and budget feasibility

19
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other ??
• Destinations closer to home ??
• More transit/walking

20
Mode Shifts

• Shifts from driving to bus and walking are
  primarily reflected in tour mode choice
• Nested logit models of combined tour mode and
  stop location choice were estimated sequentially
  from household travel on-board survey data
• Modeled elasticity of bus ridership 0.853
• Observed elasticity of bus ridership from KATS
  weekly counts for 2006 vs. 2008 0.318

21
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other ??
• Destinations closer to home ??
• More transit/walking
• Fewer tours (more stops/tour)

22
Tour-making

• Conceptually, it seems reasonable that travelers
  may respond to increased fuel prices by reducing
  travel costs by combining/eliminating tours
• However, the Knoxville data showed no evidence of
  this sort of behavior

23
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other ??
• Destinations closer to home ??
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)

24
Activity Participation

• Travelers can also respond by decreasing their
  participation in out-of-home activities
• This effect was observed in the Knoxville data
  and incorporated in stop generation
• Low income travelers (lt 25k/yr) and
  discretionary activities were primarily affected
• Range of elasticities for various income groups
  and stop types -0.155 to -0.233

25
ModelingTravelers Response

• Traditional models have represented some of these
  responses, but neglected others,
• More carpooling
• Destinations closer to each other ??
• Destinations closer to home ??
• More transit/walking
• Fewer tours (more stops/tour)
• Lower activity participation (fewer stops)
• Lower vehicle ownership (long term)

26
Vehicle Ownership

• Over the long term, travelers can also respond by
  owning fewer (or more efficient) vehicles
• An ordered response logit model for vehicle
  ownership choice was estimated
• Elasticity of household vehicles with respect to
  fuel price -0.067

27
Ongoing Work

• Currently, estimation is complete for the new
  Knoxville model, but work is on-going to
  calibrate the component models
• Hope to estimate total elasticity of VMT to fuel
  price as part of model validation

28
Thank You!