discrete choice analysis: hydrogen fcv demand potential · d. dist. trips: no adv planning driving...
TRANSCRIPT
Discrete Choice Analysis:H2 FCV Demand Potential
Cory Welch
H2 Scenario Analysis WorkshopWashington, D.C. , January 31, 2007
2
Overview
• Motivation for work
• Methodology
• Relative Attribute Strengths
• Insights and Recommendations
• Responses to Key Questions
3
Acknowledgments
Special thanks toAndy Goett (1), Bob Baumgartner (1), Eric Rambo (1), Sergei Rodkin (2) for close collaboration
and to
(1) PA Consulting, (2) Knowledge Networks
for insightful feedback
Klaus Bonhoff -- DaimlerChryslerSig Gronich -- DOEFred Joseck -- DOE
Craig Stephan -- FordBritta Gross -- GM
Moshe Ben-Akiva -- MITJeroen Struben -- MIT
Kenneth Train -- NERAMaggie Mann -- NREL
Keith Wipke -- NRELTonya Huyett -- NREL
Trish Bross -- NRELDon Jones -- RCF
4
Motivation
• Quantify sensitivity of H2 FCV market potential to station convenience
• Move beyond “% stations” or “distance to nearest station” metrics
• Input quantified parameters into HyDIVE™(other models?) for data driven analysis– quantify “chicken-and-egg” barrier– develop robust, high leverage policies/strategies
5
Methodology
• Discrete Choice Analysis– consumers “choose” among vehicles w/diff. attributes– through repeated choices, sensitivities are determined
• Responses being obtained from 500 households– Knowledge Networks’ Knowledge PanelSM
– 10 choice tasks per respondent, 2 decisions per task– 500 x 10 x 2 = 10,000 choice observations
• NREL used BIOGEME* for parameter estimation, model development (logit, nested logit, etc.)
* Bierlaire, M. (2003). BIOGEME: A free package for the estimation of discrete choice models , Proceedings of the 3rd Swiss Transportation Research Conference, Ascona, Switzerland.
6
Target Region: Southern California
∝
∝
7
Local Trips (w/in 20 mi. of
home)
Driving Radius (e.g., 150 miles)
Medium Distance
Trips (20-150 mi. from home)
Long Distance Trips (>150 mi.
from home)
Home
Local Trips (w/in 20 mi. of
home)
Driving Radius (e.g., 150 miles)
Medium Distance
Trips (20-150 mi. from home)
Long Distance Trips (>150 mi.
from home)
Home
Spatial Characterization
8
Example: “Choice Task”
9
Why are results “preliminary”
• Data recently received (Jan. 25)
• Subset of total dataset (451/500)
• Doesn’t yet consider consumer heterogeneity– via segmentation and/or “random” coefficients
• Interactions/non-linearity not fully explored
• Demographic balance to be quantified
However, results are considered to be directionally accurate, and all estimated coefficients are statistically significant.
Tornado Chart: H2 FCV Market Share Potential
0.8 1 1.2 1.4 1.6 1.8 2
Extra Time to Local Stations
d. Dist. Trips: No Adv Planning
Driving Radius
Price Difference
Fuel Cost Difference
Possible Long Distance Trips
0 min3 min10 min
Reference Case 2
0 % 50 % 90 % 100 %
Reference Case 2
100 mi 150 mi 200 mi
Reference Case 2
-15 %0 %15 %
Reference Case 2
-50 %0 %50 %
Reference Case 2
0 % 50 % 90 % 100 %
Reference Case 2
Market Share Potential -- Relative to Reference
DescriptionNonLinMod14-450Responses.matee next slide.
0.8 1 1.2 1.4 1.6 1.8 2
Extra Time to Local Stations
d. Dist. Trips: No Adv Planning
Driving Radius
Price Difference
Fuel Cost Difference
Possible Long Distance Trips
0 min3 min10 min
Reference Case 2
0 % 50 % 90 % 100 %
Reference Case 2
100 mi 150 mi 200 mi
Reference Case 2
-15 %0 %15 %
Reference Case 2
-50 %0 %50 %
Reference Case 2
0 % 50 % 90 % 100 %
Reference Case 2
Market Share Potential -- Relative to Reference
DescriptionNonLinMod14-450Responses.matee next slide.
Med. Distance Trips: No Advance Planning
*see previous slide
700/
350
bar?
Sta
tion
loca
tion,
tim
ing
stra
tegy
Veh
icle
vs.
fuel
, st
atio
n in
cent
ives
How NOT to read this chart: widest is “most important”
?
11
HyDIVE Example – “Static”
L.A. L.A.
Gas stationH2 station B: 400 StationsA: 40 Stations
12
HyDIVE Example – “Static”
Greater demand(1) potential due to improved station coverage/convenience
L.A. L.A.
B: 400 StationsA: 40 Stations
(1) All else equal (e.g., net vehicle price, availability, etc.)
Vehicle Demand(1)
13
HyDIVE Example – “Static”
Vehicles/Station(1) (B) Vehicles/Station (A)<< ?
However,
(1) An indicator of station profitability
L.A. L.A.
B: 400 StationsA: 40 Stations
Vehicles per Station(1)
14
What is your most important consideration when choosing where to refuel your vehicle?
0%
10%
20%
30%
40%
50%
60%
70%
Low-pricedfuel
Convenientlocation
Access to apreferred
brand of fuel
Avoiding awaiting line to
refuel
Availability ofadditionalservices
Station Selection Considerations
Helps explain relative strengths of “time to
station” and “fuel cost”
15
Convenient Time to Station
What is the longest time you would drive to a refueling station and still consider it to be “convenient” ?
0%
10%
20%
30%
40%
50%<2 2-5
5-10
10-1
515
-20
20-2
525
-30
>30
Drive Time (minutes)
0%
20%
40%
60%
80%
100%
Cum
ulat
ive
(%)
Average = 6.4 minutes
What is the longest time you would drive to a refueling station and still consider it to be “convenient” ?
0%
10%
20%
30%
40%
50%<2 2-5
5-10
10-1
515
-20
20-2
525
-30
>30
Drive Time (minutes)
0%
20%
40%
60%
80%
100%
Cum
ulat
ive
(%)
Average = 6.4 minutes
Helps explain low sensitivity of “time to
station” below 10 min.
16
“Comfort Level” – No Stations NearbyWould you be comfortable estimating whether you had
enough fuel in your tank to travel to a destination with no refueling stations nearby?
0%
10%
20%
30%
40%
50%
DefinitelyNot
ProbablyNot
ProbablyYes
DefinitelyYes
Would you be comfortable estimating whether you had enough fuel in your tank to travel to a destination with no
refueling stations nearby?
0%
10%
20%
30%
40%
50%
DefinitelyNot
ProbablyNot
ProbablyYes
DefinitelyYes
Helps explain why some are
sensitive to “medium distance
trip” coverage
17
Insights and Recommendations
• Diminishing returns of local coverage once local time to station is within 5-10 minutes
• Adding more stations induces demand, but depending on the starting point,– can reduce overall station utilization/profitability.– In such a case, adding more stations is not high leverage.
• Vehicle demand would need to increase via other means (e.g., price incentives, vehicle availability, fuel cost reductions, etc.)
• The spatial, temporal behavior of this highly complex, non-linear system needs rigorous modeling/analysis considering uncertainty in various inputs to develop robust policies and strategies that are likely to be effective.
18
Additional Slides
19
“Market Share Potential” Calculation
• Can be thought of as an equilibrium, upper bound market share, assuming:– H2 FCV “option” universally available (all makes/models)– All consumers “aware” of H2 FCV technology option– All consumers are “aware” of actual levels of station coverage
(i.e., no difference b/w reality and perception)
• The dynamics of growth in “availability” and “awareness” would lower actual forecast market share relative to the “potential”– To be incorporated in future analyses (beyond Apr 07)
20
Refueling LocationWhere do you most frequently refuel your vehicle?
0%
10%
20%
30%
40%
50%
60%
70%
Nearest home It dependsmainly on
where I amwhen my tank
is low
B/w work &home (e.g.,
along freewayor other road
to work)
Nearest work Other
Where do you most frequently refuel your vehicle?
0%
10%
20%
30%
40%
50%
60%
70%
Nearest home It dependsmainly on
where I amwhen my tank
is low
B/w work &home (e.g.,
along freewayor other road
to work)
Nearest work Other
21
Importance of Vehicle BenefitsWhich of these vehicle benefits is most important
to you?
0%
10%
20%
30%
40%
50%
Virtuallyeliminate oilimports for
vehicle
Reducegreenhouse
gas emissions
Eliminatesmog-causing
emissions
None areespeciallyimportant
Which of these vehicle benefits is most important to you?
0%
10%
20%
30%
40%
50%
Virtuallyeliminate oilimports for
vehicle
Reducegreenhouse
gas emissions
Eliminatesmog-causing
emissions
None areespeciallyimportant
22
Global WarmingHow concerned are you about global warming?
0%
10%
20%
30%
40%
VeryUnconcerned
VeryConcerned
How concerned are you about global warming?
0%
10%
20%
30%
40%
VeryUnconcerned
VeryConcerned
23
Pollution/SmogHow concerned are you about local air pollution or
smog?
0%
10%
20%
30%
40%
VeryUnconcerned
VeryConcerned
How concerned are you about local air pollution or smog?
0%
10%
20%
30%
40%
VeryUnconcerned
VeryConcerned